CN103795943B - Image processing apparatus and image processing method - Google Patents

Abstract

The present disclosure provides an image processing device and an image processing method for removing flicker noise in an image. The image processing device comprises: a division unit, which is used to divide the pixel value domain into at least two intervals; a noise removal unit, which is used to perform filtering with different smoothness on pixels whose pixel values in the image respectively fall into at least two intervals, so as to flicker noise in the image is removed, wherein the noise removal unit performs smoother filtering on pixels falling in intervals with higher pixel values and smoothes pixels falling in intervals with lower pixel values lower filtering.

Description

Image processing device and image processing method

technical field

The present disclosure generally relates to the field of image processing, and in particular, relates to an image processing device and an image processing method for removing flicker noise in an image.

Background technique

In the field of image and video processing, there are several forms of flicker noise. The flicker noise of the video is random in time and changes in image pixel values in space. Usually caused by degradation of the storage medium, conversion to different standards, or differences in exposure time for each image. In CMOS imaging, flicker noise is a periodic horizontal band caused by the AC signal of the light source. Whereas in the current linear sensors that are more and more used, due to the operating mechanism of the line sensor such as the top-mounted line sensor (motor drive of the line sensor), the horizontal band of the flicker noise in the scanned image obtained by the line sensor may be Unequally spaced. An image obtained by scanning with an overhead line sensor is schematically shown in FIG. 1 . For the sake of clarity, content such as documents and the like are omitted in FIG. 1 , and only flicker noise N _f in the shape of a horizontal band with unequal intervals is shown.

Traditional flicker noise reduction methods require multiple consecutive images; or use a single image, but the flicker noise pattern is periodic and the period is known. Alternatively, flicker noise removal can be performed by hardware processing.

Contents of the invention

However, processing multiple consecutive images increases the computational load on the system. Moreover, in some cases, it is difficult to obtain multiple consecutive images, so it is difficult to remove noise. In addition, for non-equally spaced flicker noise, the method suitable for equal spaced flicker noise is applied to remove the noise, and the efficiency and accuracy of noise removal become lower. In addition, noise removal by hardware processing greatly increases the running cost.

In view of the above problems, it is desired to provide an image processing device and an image processing method capable of removing flicker noises at equal or unequal intervals only by processing a single image. Noticing that the flicker noise is more obvious in the image where the pixel value is high than in the position where the pixel value is low, the inventors filter the positions with different pixel values with different degrees of smoothness according to the degree of obviousness of the noise.

According to an aspect of the present invention, there is provided an image processing device for removing flicker noise in an image, including: a division unit for dividing the pixel value range into at least two intervals; a noise removal unit for Pixels with middle pixel values falling into at least two intervals are filtered with different degrees of smoothness to remove flicker noise in the image, wherein the noise removal unit performs smoothing on pixels falling in intervals with higher pixel values. , and less smooth filtering for pixels that fall in the interval with lower pixel values.

In one embodiment, the noise removal unit may filter the pixels by applying a Gaussian filter, wherein the noise removal unit may apply a Gaussian filter with a larger variance to pixels falling in an interval with a higher pixel value, And apply a Gaussian filter with smaller variance to the pixels that fall in the bin with lower pixel values.

In one embodiment, the noise removal unit can filter the pixels in each row by processing the cumulative histogram of each row of the image, so as to remove the noise in the image. The noise removal unit may include: a cumulative histogram generating unit for generating a cumulative histogram for each row in the image as an original cumulative histogram; a weighting processing unit for combining the original cumulative histograms of adjacent rows of each row with The original cumulative histogram of each row is weighted together to obtain the target cumulative histogram of each row; the regularization unit is used to perform histogram regularization from the original cumulative histogram to the target cumulative histogram to obtain the noise-removed image. For each part of the original cumulative histogram corresponding to at least two intervals divided by the dividing unit, the weighting processing unit uses different weights to perform weighting processing.

In one embodiment, the weighting processing unit may perform weighting processing by applying a Gaussian function. Wherein, the weighting processing unit may apply a larger Gaussian variance to a portion corresponding to an interval with a higher pixel value; apply a smaller Gaussian variance to a portion corresponding to an interval with a lower pixel value.

In one embodiment, the noise removal unit may filter the pixels in each row to remove noise in the image by processing the cumulative histogram of each row of the image, each row of the image has one or more pixels the height of. The noise removal unit may include: a histogram generating unit for generating a histogram for each row in the image as an original histogram; a weighting processing unit for combining the original histograms of adjacent rows of each row with each row The original histograms are weighted together to obtain the target histogram of each row; the cumulative histogram generation unit is used to generate the original cumulative histogram and target cumulative histogram for each row according to the original histogram and target histogram of each row ; A regularization unit for obtaining a noise-removed image by performing histogram regularization from the original cumulative histogram to the target cumulative histogram, wherein the original histogram corresponds to at least two intervals divided by the division unit For each cell, the weighting processing unit uses different weights to perform weighting processing.

In one embodiment, the specification unit may include: a lookup table generation unit, configured to generate a lookup table from an original pixel value to a target pixel value of a pixel according to the original cumulative histogram and the target cumulative histogram; and a specification processing unit, configured to Correcting the corresponding target pixel value based on the original pixel value of the specific pixel and the corresponding target pixel value read from the lookup table according to a predetermined rule, and making the pixel value of the specific pixel in the specified result image equal to the corrected pixel value.

In one embodiment, when the absolute value of the difference between the original pixel value of a specific pixel and the corresponding target pixel value is less than or equal to a predetermined threshold, the stipulation processing unit makes the corrected target pixel value the corresponding target pixel value itself; when the specific pixel When the absolute value of the difference between the original pixel value and the corresponding target pixel value is greater than the predetermined threshold, the stipulation processing unit makes the absolute value of the difference between the original pixel value and the corrected target pixel value equal to the predetermined threshold.

In one embodiment, the prescriptive processing unit makes the corrected target pixel value equal to a weighted sum of the original pixel value and the corresponding target pixel value.

In one embodiment, the image processing device may further include: a color separation and combination unit, configured to separate the color image to be processed into separate single-channel images, and combine the single-channel images that have undergone noise removal processing into color image.

According to another aspect of the present invention, there is provided an image processing method for removing flicker noise in an image, comprising: a division step, dividing the pixel value domain into at least two intervals; a noise removal step, aiming at the pixel value in the image Pixels falling into at least two intervals are filtered with different degrees of smoothness to remove flicker noise in the image, wherein, pixels falling into the interval with a higher pixel value are filtered with a higher degree of smoothness, and pixels falling into Pixels in the interval with lower pixel values are filtered less smoothly.

By using the image processing device and the image processing method described in the present disclosure, it is possible to efficiently implement equal-spaced and non-equal-spaced flicker noise by only processing a single image without increasing the amount of calculation and running cost of the system. remove.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory in nature and are not restrictive of the invention, as claimed.

Description of drawings

The above and other objects, features and advantages of the present invention will be more easily understood with reference to the following description of the embodiments of the present invention in conjunction with the accompanying drawings. In the drawings, the same or corresponding technical features or components will be indicated by the same or corresponding reference numerals. The dimensions and relative positions of elements are not necessarily drawn to scale in the drawings.

FIG. 1 is a schematic diagram showing an image obtained by scanning using an overhead line sensor.

FIG. 2 is a block diagram showing the configuration of an image processing apparatus according to an embodiment of the present invention.

FIG. 3 is a flowchart illustrating an image processing method according to an embodiment of the present invention.

FIG. 4 is a block diagram showing the structure of an image processing apparatus according to another embodiment of the present invention.

FIG. 5 is a flowchart illustrating an image processing method according to another embodiment of the present invention.

FIG. 6 is a block diagram showing the structure of an image processing apparatus according to another embodiment of the present invention.

FIG. 7 is a flowchart illustrating an image processing method according to another embodiment of the present invention.

FIG. 8 is a block diagram showing the structure of an improved specification unit according to an embodiment of the present invention.

FIG. 9 is a block diagram showing an exemplary structure of a computer realizing the image processing apparatus of the present invention.

detailed description

Embodiments of the present invention will be described below with reference to the drawings. It should be noted that representation and description of components and processes that are not related to the present invention and known to those skilled in the art are omitted from the drawings and descriptions for the purpose of clarity.

For the scanned image with non-equally spaced flicker noise N _f shown in FIG. 1 , whether it is a grayscale image or a color image, the image processing device and image processing method described in the present invention can be used for noise removal. The difference is that for a color image, it is first necessary to separate the color image into single-channel images such as R, G, and B channels. The processing taught in this disclosure is performed on the images of each channel separately. Then, the processed images of the three channels are combined to obtain a color image with flicker noise removed. In the following, the processing of a grayscale image or a single-channel image of one of the images of R, G, and B channels will be described. In addition, needless to say, the present invention is also applicable to images with equally spaced flicker noise.

FIG. 2 is a block diagram showing the configuration of an image processing apparatus 200 according to an embodiment of the present invention. As shown in FIG. 2 , the image processing device 200 includes: a division unit 210 and a noise removal unit 220 .

The dividing unit 210 is used for dividing the pixel value range into at least two intervals. The noise removal unit 220 is configured to perform filtering with different degrees of smoothness on pixels whose pixel values in the image respectively fall into at least two intervals divided by the division unit 210 , so as to remove flicker noise in the image. Specifically, the noise removal unit 220 performs smoother filtering on pixels falling in an interval with a higher pixel value in the pixel value domain, and performs smoother filtering on pixels falling in an interval with a lower pixel value. low filtering. Each unit will be described in detail below.

First, the concept of "pixel range" will be described. In the field of computer image processing, binary numbers are used to quantize single-channel grayscale images, so that different pixel values are used to represent the grayscale of pixels. For example, a grayscale image can be quantized to 8 bits, resulting in 256 pixel values (0 to 255). Typically in the art, and by default in the context of this disclosure, a pixel value of 0 represents the darkest shade of black, and a pixel value of 255 represents the lightest shade of white. The integer value from 0 to 255 (inclusive) is the pixel value range when 8-bit quantization is used, which will be represented by [0, 255] in the following. Similarly, if 10-bit quantization is used, the integer value of the pixel value range from 0 to 1023 (inclusive) is the pixel value range during 10-bit quantization, which can be represented by [0, 1023]. In the following, the case of 8-bit quantization is taken as an example for description.

The division unit 210 can perform different divisions on the pixel value range as required. For example, the pixel value range is divided according to one or more preset dividing points (pixel values serving as interval boundaries). Alternatively, the pixel value range is equally divided according to a preset number of copies. The division points or the number of copies can be set according to different standards and depending on different design requirements. For example, in one embodiment of the present invention, one or more divisions are set considering that the flicker noise in the scanned image is less noticeable at positions with lower pixel values and more noticeable at positions with higher pixel values. point. In one embodiment, only two intervals can be divided, and in the pixel value range of 0 to 255, a pixel value in the range of 170 to 200 can be taken as the division point. Of course, in other embodiments, two or more dividing points may also be used in the entire pixel value range, so as to divide the pixel value range into three or more intervals. In addition, the division unit 210 may also perform adaptive division according to the characteristics of the image to be processed.

The noise removal unit 220 processes the image to be processed to remove flicker noise in the image. Specifically, the noise removal unit 220 performs filtering with different degrees of smoothness on the pixels whose pixel values in the image respectively fall into at least two intervals divided by the division unit 210 , so as to remove the flicker noise in the image. In one embodiment, considering that the flicker noise is more obvious at the position of higher pixel value than at the position of lower pixel value, the noise removal unit 220 smoothes the pixels falling in the interval with higher pixel value to a higher degree. , and less smooth filtering for pixels that fall in the interval with lower pixel values.

For example, in the example of dividing the pixel value range into 3 intervals (for example, the two division points may be 160 and 200 respectively), the noise removal unit 220 may perform a pixel value ranging from 0 to 160 (for example, [0,160), without The pixels on the interval containing 160) are filtered with low smoothness, the pixels on the interval with pixel values from 160 to 200 (such as [160,200)) are filtered with medium smoothness, and the pixels with values from 200 to 255 ( For example [200,255]), the pixels on the interval are filtered with a higher degree of smoothness. Hereinafter, for the sake of simplicity, the embodiment of dividing two intervals will be taken as an example for detailed description.

In the example of dividing the pixel value range into two intervals (for example, the division point is 180), the noise removal unit 220 can perform the Filtering with a lower degree of smoothness, and filtering with a higher degree of smoothness for pixels on the interval of pixel values from 180 to 255 (such as [180,255]). Further specific examples are described below.

In one embodiment, the noise removal unit 220 may filter the pixels by applying an image processing filter such as a Gaussian filter. The noise removal unit 220 may apply a Gaussian filter with a larger variance to pixels falling in an interval with higher pixel values (eg, [180,255]), and apply a Gaussian filter with a larger variance to pixels falling in an interval with lower pixel values (eg, [0,180]). ) to apply a Gaussian filter with a small variance. In this embodiment, the Gaussian filter used may be a two-dimensional Gaussian filter, as shown in formula (1):

Among them, σ ² is the variance of the two-dimensional Gaussian function.

In this embodiment, a larger variance σ ² is selected for pixels whose pixel values fall into the interval with higher pixel values, and a smaller variance σ 2 is selected for pixels whose pixel values fall into the interval with lower pixel values ² . For example, for pixels whose pixel values fall within the interval [180, 255], σ ² =2 may be set; for pixels whose pixel values fall within the interval [0,180), σ ² =0.2 may be set.

It should be noted that, in addition to the Gaussian filter, various image processing filters known in the art can also be used to perform the smoothing process according to the present invention. For example, using an image processing filter such as a bilateral filter performs smoother filtering at positions with higher pixel values and less smooth filtering at positions with lower pixel values.

FIG. 3 is a flowchart showing an image processing method corresponding to the processing performed by the image processing apparatus shown in FIG. 2 .

As shown in FIG. 3, in step S301, the pixel value range is divided into at least two intervals. The pixel value range can be divided differently according to needs. For example, the pixel value range is divided according to one or more preset dividing points. Alternatively, the pixel value range is equally divided according to a preset number of copies. The division points or the number of copies can be set according to different standards and depending on different design requirements. For example, in one embodiment of the present invention, one or more divisions are set considering that the flicker noise in the scanned image is less noticeable at positions with lower pixel values and more noticeable at positions with higher pixel values. point. In one embodiment, in the pixel value range from 0 to 255, one of the pixel values from 170 to 200 may be used as the dividing point, thereby dividing two intervals.

In step S302, filtering with different degrees of smoothness is performed on pixels whose pixel values in the image respectively fall into at least two intervals, so as to remove flicker noise in the image. Specifically, pixels falling in an interval with a higher pixel value may be filtered with a higher degree of smoothness, and pixels falling in an interval with a lower pixel value may be filtered with a lower degree of smoothness.

One embodiment is to filter the pixels by applying various image processing filters known in the art. For example, a Gaussian filter can be applied to filter pixels in an image. Specifically, a Gaussian filter with a larger variance is applied to pixels falling in an interval with a higher pixel value, and a Gaussian filter with a smaller variance is applied to pixels falling in an interval with a lower pixel value. A detailed example is described in conjunction with FIG. 2 , and will not be repeated here.

FIG. 4 is a block diagram showing the structure of an image processing apparatus 400 according to an embodiment of the present invention. As shown in FIG. 4 , the image processing device 400 may include a division unit 410 and a noise removal unit 420 . Wherein, the division unit 410 has the same function and structure as the division unit 210 shown in FIG. 2 , and thus, its detailed description is omitted here. The noise removal unit 420 can filter the pixels in each row of the image to be processed by processing the cumulative histogram of the row, so as to remove the noise in the image. The noise removal unit 420 may include: a cumulative histogram generation unit 422 , a weighting processing unit 424 and a specification unit 426 . Hereinafter, detailed description will be given unit by unit.

The cumulative histogram generating unit 422 can generate a cumulative histogram for each row in the image as an original cumulative histogram.

In one example, when receiving an image to be processed, the cumulative histogram generation unit 422 of the noise removal unit 420 first generates a histogram for each row of the image to be processed. The height of "each row" mentioned here can be one pixel or several pixels. The cumulative histogram generating unit 422 can generate a histogram using various existing methods.

For example, assuming that the width and height of a single-channel image are expressed as w and h respectively, and the pixel value coordinates at the image point (x, y) are I(x, y), then for the histogram of the i-th row of the single-channel image The graph calculation formula can be shown in formula (2):

u＝0,1...255 (2)

Among them, u represents the number of the cell (bin). Please note: Equation (2) regards a pixel value as a small area, and adopts an 8-bit quantized grayscale image, so there are 256 small areas in total. In formula (2), the δ operator represents such a mapping relationship: δ(0)=1 (or non-zero value) and δ(non-zero)=0.

Equation (2) shows an example of obtaining a histogram when one pixel value is regarded as one small area. According to the definition of a histogram, multiple pixel values may also be corresponding to one small area. For example, u=0,...,7, as shown in the following formula (3):

u=0,1...7 (3)

Among them, B represents the mapping from pixel value I to cell u, when the mapped value B of pixel value I falls into cell u, δ=1 (or non-zero value), otherwise δ=0.

Then, the cumulative histogram generation unit 422 generates a cumulative histogram for each row based on the histogram for each row of the image. For example, the cumulative histogram generating unit 422 can generate a cumulative histogram based on the following equation (4):

The weighting processing unit 424 obtains the target cumulative histogram of each row by weighting the original cumulative histogram of adjacent rows of each row together with the original cumulative histogram of each row. For each part in the original cumulative histogram generated by the cumulative histogram generation unit 422 corresponding to at least two intervals divided by the division unit 410, for example, the original cumulative histogram corresponds to the intervals [0, 180) and [180, 255] respectively The weighting processing unit 424 performs weighting processing using different weights.

In one example, the weighting processing unit 424 performs weighting processing on the original cumulative histogram H _i±k of the adjacent row of the i-th row (i±k-th row) together with the original cumulative histogram H _i of the i-th row, to Get the target cumulative histogram for row i. Wherein, k can take a value of 1, 2, 3, or more.

In one embodiment of the present invention, various image processing filter functions known in the art can be used to perform the weighting process. For example, Gaussian functions can be used to weight Hi and _Hi±k to obtain the target cumulative histogram for row _i As shown in formula (5):

in, k represents the number of adjacent rows in the up and down direction, σ ² is the Gaussian variance, and, there are:

Where σ ₁ ² <σ ₂ ² , T is the threshold separating the low pixel value interval from the high pixel value interval, that is, the dividing point. In one example, T can take on a value of 170-200. Here, only one dividing point is used to divide two intervals. In other embodiments, it is also possible to divide more than two intervals by using more than one dividing point. Also, different variances are employed depending on the pixel values in each section.

The normalization unit 426 obtains a noise-removed image by performing histogram normalization from the original cumulative histogram to the target cumulative histogram.

After getting the target cumulative histogram Afterwards, the specification unit 426 in the noise removal unit 420 passes from the original cumulative histogram H _i to the target cumulative histogram Perform histogram normalization to obtain a noise-removed image. For example, the normalization unit 426 can use a traditional normalization method to transform the original cumulative histogram H _i into the target cumulative histogram by adjusting the pixel values of the image Cumulative histogram vs converted goal The corresponding image is the image from which the noise has been removed.

FIG. 5 is a flowchart showing an image processing method corresponding to the processing performed by the image processing apparatus shown in FIG. 4 . In step S501, the pixel value range is divided into at least two intervals. Its processing is the same as that described in connection with step S301 in FIG. 3 , and will not be repeated here.

In step S502, a cumulative histogram is generated for each row in the image as an original cumulative histogram.

In one example, when an image to be processed is received, a histogram is first generated for each row of the image to be processed. Various existing methods can be used to generate the histogram. A cumulative histogram for each row is then generated based on the histogram for each row of the image. The generation formulas of histogram and cumulative histogram are shown in formula (2) or (3) and (4) above.

In step S503, the target cumulative histogram of each row is obtained by weighting the original cumulative histogram of adjacent rows of each row together with the original cumulative histogram of each row. Specifically, for the part of the original cumulative histogram corresponding to at least two intervals of the pixel value range divided in step S301, different weights are used for weighting processing.

For example, in one embodiment, in step S503, weighting processing may be performed by applying a Gaussian function. Specifically, a larger Gaussian variance is applied to the portion of the original cumulative histogram corresponding to the interval with higher pixel values; a smaller Gaussian variance is applied to the portion corresponding to the interval with lower pixel values.

In step S504, a noise-removed image is obtained by performing histogram specification from the original cumulative histogram to the target cumulative histogram. The traditional prescriptive method can be used to transform the original cumulative histogram into the target cumulative histogram by adjusting the pixel values of the image, and the image corresponding to the converted target cumulative histogram is the image from which the noise has been removed.

In the above embodiments, the flicker noise of the image is removed by weighting the cumulative histogram of each row of the image. In another embodiment, the flicker noise of the image may also be removed by directly performing a weighting process on the histogram of the image, and then obtaining a cumulative histogram for prescribing.

FIG. 6 is a block diagram showing the structure of an image processing apparatus 600 according to another embodiment of the present invention. As shown in FIG. 6 , the image processing device 600 may include a division unit 610 and a noise removal unit 620 . Wherein, the division unit 610 has the same function and structure as the division units 210 and 410 shown in FIGS. 2 and 4 , and thus, its detailed description is omitted here. The noise removal unit 620 can filter the pixels in each row of the image to be processed by processing the cumulative histogram of the row, so as to remove the noise in the image. The noise removal unit 620 may include: a histogram generation unit 622 , a weighting processing unit 624 , a cumulative histogram generation unit 626 and a specification unit 628 . Hereinafter, detailed description will be given unit by unit.

The histogram generation unit 622 may generate a histogram for each row in the image to be processed as an original histogram. Here, the height of each row of the image may be one or more pixels. In other words, each row of the image has a height of one or more pixels. For histogram generation, reference may be made to the example described above with reference to formula (2) or (3).

The weighting processing unit 624 can obtain the target histogram of each row by weighting the original histograms of adjacent rows of each row together with the original histogram of each row. The weighting process is performed on the original histograms of each row and its adjacent rows, that is, the weighting process is performed on the number of pixels corresponding to the corresponding sub-districts in the original histogram.

Taking the division into two intervals [0,180) and [180,255] as an example, assuming that on the interval [0,180), the weighting coefficient corresponding to the original histogram of the i-th row image is Q _i , and the weighting coefficient corresponding to its adjacent row i±k is Q _i±k ; and on the interval [180,255], the weighting coefficient corresponding to the original histogram of the i-th row image is P _i , and the weighting coefficient corresponding to the adjacent row i±k is P _i±k . Among them, the settings of Q _i , Q _i±k , P _i , P _i±k make it possible to filter the pixels of the i-th row image falling in the interval [0,180) with a relatively low degree of smoothness, while the pixels falling into The pixels of the i-th row of image in the interval [180, 255] are filtered with a higher degree of smoothness.

The cumulative histogram generating unit 626 may generate an original cumulative histogram and a target cumulative histogram for each row from the original histogram and the target histogram of each row, respectively. Those skilled in the art can know the method of generating the histogram, which will not be described in detail here.

The normalization unit 628 may obtain a noise-removed image by performing histogram normalization from the original cumulative histogram to the target cumulative histogram. For example, the regularization unit 628 can use a traditional regularization method to transform the original cumulative histogram into a target cumulative histogram by adjusting the pixel values of the image, and the image corresponding to the converted target cumulative histogram is the image with noise removed .

Although the embodiment described with reference to FIG. 6 and the embodiment described with reference to FIG. 4 perform weighting processing on the histogram and the cumulative histogram respectively, since they all perform different degrees of filtering on pixels falling into different pixel value intervals, they can all be achieve a good noise removal effect.

FIG. 7 is a flowchart showing an image processing method corresponding to the processing performed by the image processing apparatus 600 according to FIG. 6 . In step S701, the pixel value range is divided into at least two intervals. Its processing is the same as that described in conjunction with steps S301 and S501 in FIG. 3 and FIG. 5 , and will not be repeated here.

In step S702, a histogram is generated for each row in the image as an original histogram. Then, in step S703, the target histogram of each row is obtained by weighting the original histogram of the adjacent row of each row together with the original histogram of each row. Specifically, different weights are used to carry out weighting processing on the original histograms of the row and adjacent rows corresponding to at least two intervals divided by the division unit, so as to filter the pixel values of the image. For example, pixel values falling in the interval with higher pixel values are filtered more smoothly, and pixel values falling in the interval with lower pixel values are filtered less smoothly. For an example method, reference may be made to the description made when describing the image processing device in FIG. 6 .

In step S704, an original cumulative histogram and a target cumulative histogram for each row are generated respectively according to the original histogram and the target histogram of the row. In step S705, a noise-removed image is obtained by performing histogram specification from the original cumulative histogram to the target cumulative histogram. This histogram normalization can be performed using conventional normalization methods.

The image processing apparatus and method according to the present invention have been described above by way of example. Using the scheme above, non-equally spaced flicker noise removal can be performed on color images with large differences in adjacent row histograms or cumulative histograms. Next, in order to obtain a better noise removal effect, further improvements will be discussed in conjunction with other examples. In the above scheme, the pixels have been directly filtered, or the histogram or cumulative histogram is smoothed to indirectly filter the pixels, and the positions with different pixel values are filtered with different degrees of smoothness. Image with flicker noise removed. Hereinafter, with reference to FIG. 8 , a further improvement solution capable of avoiding problems such as linear distortion or overflow caused by noise removal is described.

FIG. 8 is a block diagram showing the structure of an improved specification unit 800 according to an embodiment of the present invention. As shown in FIG. 8 , the specification unit 800 in the image processing apparatus according to this embodiment includes: a lookup table generation unit 810 and a specification processing unit 820 .

The lookup table generation unit 810 may generate a lookup table from the original pixel value to the target pixel value of the pixel according to the original cumulative histogram and the target cumulative histogram. Then, the specification processing unit 820 can modify the corresponding target pixel value according to a predetermined rule based on the original pixel value of the specific pixel and the corresponding target pixel value read from the lookup table, and make the specific pixel in the specified result image The pixel value of is equal to the corrected pixel value.

The rules for correcting target pixels can be designed according to different needs. In one embodiment of the present invention, when the absolute value of the difference between the original pixel value of a specific pixel and the corresponding target pixel value is less than or equal to a predetermined threshold α, the specification processing unit 820 can make the corrected target pixel value the corresponding target The pixel value itself; and when the absolute value of the difference between the original pixel value and the corresponding target pixel value of the specific pixel is greater than the predetermined threshold α, the stipulation processing unit 820 can make the absolute value of the difference between the original pixel value and the corrected target pixel value The value is equal to the predetermined threshold α. As shown in formula (6):

According to the original cumulative histogram and the target cumulative histogram, the lookup table generating unit 810 obtains a lookup table F(), which maps the original pixel value I(x,y) to the target pixel value F(I(x,y)). But in order to further avoid special situations such as overflow, as shown in formula (6), the specification processing unit 820 determines whether to use the target pixel value itself in the lookup table as the pixel of the corresponding pixel in the specification result image according to the preset threshold α value When |F(I(x,y))-I(x,y)|>α, the specification processing unit 820 discards the target pixel value in the lookup table, and uses to generate corrected target pixel values for specifying the resulting image. Here, "±" to α depends on the sign of F(I(x,y))-I(x,y). Specifically, when F(I(x,y))-I(x,y)>0, it is "+"; when F(I(x,y))-I(x,y)<0, for"-". In addition, in some cases, when I(x,y)±α<0 or I(x,y)±α>255, I(x,y)±α may be made equal to 0 or 255, respectively.

For example, assuming that the original pixel value is 120 and its corresponding target pixel value is 80 in the lookup table, the absolute value of the difference between the original pixel value and the target pixel value is 40. In the case where the preset threshold value is 20, the stipulation processing unit 820 corrects the target pixel value so that the corrected target pixel value is equal to 100 (120−20). Then, the specification processing unit 820 writes the corrected pixel value (100) into the corresponding position of the image as the processing result.

In another embodiment of the present invention, when the absolute value of the difference between the original pixel value of a specific pixel and the corresponding target pixel value is less than or equal to a predetermined threshold, the specification processing unit 820 can make the corrected target pixel value corresponding to the target pixel value itself; and when the absolute value of the difference between the original pixel value of a specific pixel and the corresponding target pixel value is greater than a predetermined threshold, the stipulation processing unit 820 can modify the target pixel value so that the corrected target pixel value is equal to the original pixel value and the weighted sum of the corresponding target pixel values. As shown in the following formula (7), a way to generate a corrected target pixel value by calculating the weighted sum:

Among them, 0≤m≤1.

Of course, it is also possible not to set any threshold, and to make the specification processing unit 820 directly correct the target pixel value. For example, as shown in formula (7), the corrected target pixel value is equal to the weighted sum of the original pixel value and the corresponding target pixel value.

The improved prescriptive processing method used by the prescriptive unit 800 corresponds to the prescriptive processing steps performed by the prescriptive unit 800 described above in detail in conjunction with FIG. 8 , and will not be described in detail here.

Flicker noise removal for color images. The image processing apparatus according to the present invention may further include a color separation and combination unit. The color separation and combination unit can separate the color image to be processed into separate single-channel images (such as R, G, B), and combine the single-channel images that have undergone noise removal processing respectively into a color image. In the corresponding processing, the color image is separated into individual single-channel images before noise removal is performed on the color image; and, after noise removal is performed on each single-channel image, each single-channel image is combined into a color image.

Hereinafter, an exemplary structure of a computer implementing the image processing apparatus of the present invention is described with reference to FIG. 9 . FIG. 9 is a block diagram showing an exemplary structure of a computer implementing the present invention.

In FIG. 9 , a central processing unit (CPU) 901 executes various processes according to programs stored in a read only memory (ROM) 902 or programs loaded from a storage section 908 to a random access memory (RAM) 903 . In the RAM 903, data required when the CPU 901 executes various processes is also stored as necessary.

The CPU 901 , ROM 902 , and RAM 903 are connected to each other via a bus 904 . An input/output interface 905 is also connected to the bus 904 .

The following components are connected to the input/output interface 905: an input section 906 including a keyboard, a mouse, etc.; an output section 907 including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage section 908 , including a hard disk, etc.; and a communication section 909, including a network interface card such as a LAN card, a modem, and the like. The communication section 909 performs communication processing via a network such as the Internet.

A drive 910 is also connected to the input/output interface 905 as needed. A removable medium 911 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 910 as necessary, so that a computer program read therefrom is installed into the storage section 908 as necessary.

In the case of implementing the above-described steps and processing by software, the programs constituting the software are installed from a network such as the Internet or a storage medium such as the removable medium 911 .

Those skilled in the art should understand that such a storage medium is not limited to the removable medium 911 shown in FIG. 9 in which the program is stored and distributed separately from the method to provide the program to users. Examples of the removable medium 911 include magnetic disks, optical disks (including compact disk read only memory (CD-ROM) and digital versatile disk (DVD)), magneto-optical disks (including MiniDisc (MD)), and semiconductor memories. Alternatively, the storage medium may be the ROM 902, a hard disk contained in the storage section 908, or the like, in which programs are stored and distributed to users together with a method containing them.

In the above description of specific embodiments of the present invention, features described and/or illustrated for one embodiment can be used in one or more other embodiments in the same or similar manner, and features in other embodiments Combination, or replacement of features in other embodiments.

It should be emphasized that the term "comprising/" when used herein refers to the presence of features, elements, steps or components, but does not exclude the presence or addition of one or more other features, elements, steps or components.

In addition, the methods in the various embodiments of the present invention are not limited to being executed in the time sequence described in the description or shown in the drawings, and may also be executed in other time sequences, in parallel or independently. Therefore, the execution order of the methods described in this specification does not limit the technical scope of the present invention.

To sum up, in the embodiments according to the present invention, the present invention provides the following solutions:

1. An image processing device for removing flicker noise in an image, comprising:

A division unit, configured to divide the pixel value range into at least two intervals;

A noise removal unit, configured to perform filtering with different degrees of smoothness on pixels whose pixel values in the image respectively fall into the at least two intervals, so as to remove flicker noise in the image, wherein,

The noise removal unit performs smoother filtering on pixels falling in a section with a higher pixel value, and performs smoother filtering on pixels falling in a section with a lower pixel value.

2. The image processing device according to item 1, wherein the noise removal unit filters pixels by applying a Gaussian filter, wherein the noise removal unit filters pixels falling in a section with a higher pixel value Apply a Gaussian filter with a larger variance, and apply a Gaussian filter with a smaller variance to pixels that fall in an interval with lower pixel values.

3. The image processing device according to item 1, wherein the noise removal unit filters pixels in each row of the image by processing a cumulative histogram of each row of the image to remove noise in the image noise, each row of the image has a height of one or more pixels, and the noise removal unit includes:

A cumulative histogram generating unit, configured to generate a cumulative histogram for each row in the image as an original cumulative histogram;

a weighting processing unit, configured to perform weighting processing on the original cumulative histogram of adjacent rows of each row and the original cumulative histogram of each row to obtain the target cumulative histogram of each row;

A regularization unit for obtaining a noise-removed image by normalizing the histogram from the original cumulative histogram to the target cumulative histogram, wherein,

For each part of the original cumulative histogram corresponding to the at least two intervals divided by the dividing unit, the weighting processing unit uses different weights to perform the weighting process.

4. The image processing device according to item 3, wherein the weighting processing unit performs the weighting processing by applying a Gaussian function, wherein

The weighting processing unit applies a larger Gaussian variance to a portion corresponding to the interval with a higher pixel value; applies a smaller Gaussian variance to a portion corresponding to the interval with a lower pixel value.

5. The image processing device according to item 1, wherein the noise removal unit filters pixels in each row of the image by processing a cumulative histogram of each row of the image to remove noise in the image noise, each row of the image has a height of one or more pixels, and the noise removal unit includes:

A histogram generating unit, configured to generate a histogram for each row in the image as an original histogram;

a weighting processing unit, configured to obtain the target histogram of each row by weighting the original histogram of the adjacent row of each row and the original histogram of each row;

a cumulative histogram generating unit, configured to generate an original cumulative histogram and a target cumulative histogram for each row according to the original histogram and the target histogram for each row;

A specification unit for obtaining a noise-removed image by performing histogram specification from the original cumulative histogram to the target cumulative histogram, wherein,

The weighting processing unit uses different weights to perform the weighting processing for each cell in the original histogram corresponding to the at least two intervals divided by the dividing unit.

6. The image processing device according to any one of clauses 3 to 5, wherein the specifying unit comprises:

a lookup table generation unit, configured to generate a lookup table from the original pixel value of the pixel to the target pixel value according to the original cumulative histogram and the target cumulative histogram; and

A prescriptive processing unit, configured to modify the corresponding target pixel value according to a predetermined rule based on the original pixel value of a specific pixel and the corresponding target pixel value read from the look-up table, and make the specified result image The pixel value of the particular pixel is equal to the corrected pixel value.

7. The image processing device according to item 6, wherein when the absolute value of the difference between the original pixel value of a specific pixel and the corresponding target pixel value is less than or equal to a predetermined threshold, the specification processing unit makes the corrected target pixel value is the corresponding target pixel value itself; when the absolute value of the difference between the original pixel value of the specific pixel and the corresponding target pixel value is greater than a predetermined threshold, the specification processing unit makes the original pixel value and the corrected target pixel value The absolute value of the difference in pixel values is equal to said predetermined threshold.

8. The image processing device according to item 6, wherein the specification processing unit makes the corrected target pixel value equal to a weighted sum of the original pixel value and the corresponding target pixel value.

9. The image processing device according to any one of items 1 to 8, further comprising: a color separation and combination unit, which is used to separate the color image to be processed into separate single-channel images, and perform noise removal processing respectively The single-channel images are combined into color images.

10. An image processing method for removing flicker noise in an image, comprising:

A division step, dividing the pixel value range into at least two intervals;

The noise removal step is to perform filtering with different degrees of smoothness on pixels whose pixel values in the image respectively fall into the at least two intervals, so as to remove flicker noise in the image, wherein,

Pixels falling in bins with higher pixel values are filtered more smoothly, and pixels falling in bins with lower pixel values are filtered less smoothly.

11. The image processing method according to item 10, wherein, in the noise removal step, pixels are filtered by applying a Gaussian filter, wherein pixels falling in intervals with higher pixel values are applied with Gaussian filter with larger variance and applies a Gaussian filter with smaller variance to pixels that fall in the interval with lower pixel values.

12. The image processing method according to item 10, wherein, in the noise removal step, the pixels in each row of the image are filtered by processing the cumulative histogram of each row to remove Noise in the image, each row has a height of one or more pixels, and the noise removal step includes:

A cumulative histogram generation step, generating a cumulative histogram for each row in the image as an original cumulative histogram;

A weighting processing step, by weighting the original cumulative histogram of the adjacent row of each row and the original cumulative histogram of each row to obtain the target cumulative histogram of each row;

A regularization step, by performing histogram regularization from the original cumulative histogram to the target cumulative histogram, to obtain a noise-removed image, wherein,

In the weighting processing step, different weights are used to perform the weighting processing on the parts of the original cumulative histogram corresponding to the at least two intervals.

13. The image processing method according to item 12, wherein, in the weighting processing step, the weighting processing is performed by applying a Gaussian function, wherein,

A larger Gaussian variance is applied to portions corresponding to said intervals with higher pixel values; a smaller Gaussian variance is applied to portions corresponding to said intervals with lower pixel values.

14. The image processing method according to item 10, wherein, in the noise removal step, the pixels in each row of the image are filtered by processing the cumulative histogram of each row to remove Noise in the image, each row of the image has a height of one or more pixels, and the noise removal step includes:

A histogram generating step, generating a histogram for each row in the image as an original histogram;

A weighting processing step, by weighting the original histogram of the adjacent row of each row and the original histogram of each row together to obtain the target histogram of each row;

A cumulative histogram generation step, generating an original cumulative histogram and a target cumulative histogram for each row according to the original histogram and the target histogram of each row respectively;

A regularization step, by performing histogram regularization from the original cumulative histogram to the target cumulative histogram, to obtain a noise-removed image, wherein,

In the weighting processing step, different weights are used to perform the weighting processing for each of the cells in the original histogram corresponding to the at least two intervals divided by the dividing unit.

15. The image processing method according to any one of clauses 10 to 14, wherein the step of specifying comprises:

A look-up table generation step, generating a look-up table from the original pixel value of the pixel to the target pixel value according to the original cumulative histogram and the target cumulative histogram; and

A prescriptive processing step, based on the original pixel value of a specific pixel and the corresponding target pixel value read from the look-up table, correcting the corresponding target pixel value according to a predetermined rule, and making the specified pixel value described in the prescriptive result image The pixel value of a particular pixel is equal to the corrected pixel value.

16. The image processing method according to item 15, wherein, when the absolute value of the difference between the original pixel value of a specific pixel and the corresponding target pixel value is less than or equal to a predetermined threshold, in the prescribed processing step, the corrected The target pixel value is the corresponding target pixel value itself; and when the absolute value of the difference between the original pixel value of the specific pixel and the corresponding target pixel value is greater than a predetermined threshold value, in the prescribing processing step, the original The absolute value of the difference between the pixel value and the corrected target pixel value is equal to the predetermined threshold.

17. The image processing method according to item 15, wherein, in the prescribing processing step, the corrected target pixel value is made equal to the weighted sum of the original pixel value and the corresponding target pixel value.

18. The image processing method according to any one of items 10 to 17, further comprising:

The color separation step is to separate the color image to be processed into individual single-channel images; and the color combination step is to combine the single-channel images respectively subjected to noise removal processing into a color image.

Claims (10)

1. An image processing apparatus for removing flicker noise in an image, comprising:

a dividing unit for dividing the pixel value range into at least two sections;

a noise removing unit, configured to perform filtering with different degrees of smoothness on pixels in the image whose pixel values respectively fall into the at least two sections to remove flicker noise in the image,

the noise removing unit performs filtering with a higher degree of smoothing on pixels falling in an interval with a higher pixel value, and performs filtering with a lower degree of smoothing on pixels falling in an interval with a lower pixel value.

2. The image processing apparatus according to claim 1, wherein the noise removal unit filters the pixels by applying a gaussian filter, wherein the noise removal unit applies a gaussian filter having a large variance to pixels falling in an interval having a higher pixel value and applies a gaussian filter having a small variance to pixels falling in an interval having a lower pixel value.

3. The image processing apparatus according to claim 1, wherein the noise removing unit filters pixels in each line of the image by processing a cumulative histogram of the each line to remove noise in the image, each line of the image having a height of one or more pixels, the noise removing unit includes:

a cumulative histogram generating unit for generating a cumulative histogram for each row in the image as an original cumulative histogram;

a weighting processing unit, configured to perform weighting processing on the original cumulative histogram of the adjacent row of each row and the original cumulative histogram of each row to obtain a target cumulative histogram of each row;

a prescribing unit for obtaining an image from which noise is removed by histogram specification from an original cumulative histogram to a target cumulative histogram, wherein,

the weighting processing unit performs the weighting processing using different weights for respective portions of the original cumulative histogram corresponding to the at least two bins divided by the dividing unit.

4. The image processing apparatus according to claim 1, wherein the noise removing unit filters pixels in each line of the image by processing a cumulative histogram of the each line to remove noise in the image, each line of the image having a height of one or more pixels, the noise removing unit includes:

a histogram generating unit configured to generate a histogram for each line in the image as an original histogram;

a weighting processing unit, configured to perform weighting processing on the original histograms of the adjacent rows of each row and the original histogram of each row to obtain a target histogram of each row;

a cumulative histogram generating unit configured to generate an original cumulative histogram and a target cumulative histogram for each line from the original histogram and the target histogram of each line, respectively;

a regularization unit for obtaining an image from which noise is removed by performing histogram regularization from an original cumulative histogram to a target cumulative histogram, wherein,

the weighting processing unit performs the weighting processing using different weights for each cell of the original histogram corresponding to the at least two bins divided by the dividing unit.

5. The image processing apparatus according to claim 3 or 4, wherein the prescribing unit includes:

a look-up table generating unit for generating a look-up table of original pixel values to target pixel values of pixels according to the original cumulative histogram and the target cumulative histogram; and

a prescribing processing unit for correcting a corresponding target pixel value read from the lookup table according to a predetermined rule based on an original pixel value of a specific pixel and the corresponding target pixel value, and making a pixel value of the specific pixel equal to the corrected pixel value in a prescribing result image.

6. An image processing method for removing flicker noise in an image, comprising:

a dividing step of dividing the pixel value domain into at least two sections;

a noise removing step of performing filtering of different degrees of smoothness on pixels of which pixel values respectively fall into the at least two sections in the image to remove flicker noise in the image,

pixels falling in an interval having a higher pixel value are subjected to filtering with a higher degree of smoothing, and pixels falling in an interval having a lower pixel value are subjected to filtering with a lower degree of smoothing.

7. The image processing method according to claim 6, wherein in the noise removal step, the pixels are filtered by applying a gaussian filter, wherein a gaussian filter having a larger variance is applied to pixels falling in an interval having a higher pixel value, and a gaussian filter having a smaller variance is applied to pixels falling in an interval having a lower pixel value.

8. The image processing method according to claim 6, wherein in the noise removing step, the pixels in each line of the image are filtered by processing a cumulative histogram of each line to remove noise in the image, the each line having a height of one or more pixels, the noise removing step includes:

a cumulative histogram generating step of generating a cumulative histogram for each line in the image as an original cumulative histogram;

a weighting processing step of obtaining a target cumulative histogram of each line by weighting the original cumulative histograms of the adjacent lines of each line together with the original cumulative histogram of each line;

a prescribing step of obtaining an image from which noise is removed by performing histogram specification from an original cumulative histogram to a target cumulative histogram, wherein,

in the weighting processing step, the weighting processing is performed using different weights for portions of the original cumulative histogram corresponding to the at least two bins.

9. The image processing method according to claim 6, wherein in the noise removing step, the pixels in each line of the image are filtered by processing a cumulative histogram of the each line to remove noise in the image, each line of the image having a height of one or more pixels, the noise removing step includes:

a histogram generation step of generating a histogram for each line in the image as an original histogram;

a weighting processing step of performing weighting processing on the original histograms of the adjacent rows of each row and the original histogram of each row to obtain a target histogram of each row;

a cumulative histogram generating step of generating an original cumulative histogram and a target cumulative histogram for each line from the original histogram and the target histogram for each line, respectively;

a prescribing step of obtaining an image from which noise is removed by performing histogram specification from an original cumulative histogram to a target cumulative histogram, wherein,

in the weighting processing step, the weighting processing is performed using different weights for each cell of the original histogram corresponding to the at least two bins divided by the dividing unit.

10. The image processing method according to claim 8 or 9, wherein the prescribing step includes:

a look-up table generating step of generating a look-up table of original pixel values to target pixel values of pixels according to the original cumulative histogram and the target cumulative histogram; and

a specification processing step of correcting, according to a predetermined rule, a corresponding target pixel value read from the lookup table based on an original pixel value of a specific pixel and the corresponding target pixel value, and making the pixel value of the specific pixel in a specification result image equal to the corrected pixel value.