JP4253265B2 - Shadow detection apparatus, shadow detection method and shadow detection program, image processing apparatus using shadow detection apparatus, image processing method using shadow detection method, and image processing program using shadow detection program - Google Patents

Description

  The present invention uses a shadow detection apparatus, a shadow detection method and a shadow detection program, an image processing apparatus using the shadow detection apparatus, and a shadow detection method that can be applied to a road monitoring system and an intrusion monitoring system that use image data obtained by a camera. The present invention relates to an image processing method and an image processing program using a shadow detection program.

  Conventionally, shadows are generated around an object in a time zone in which sunlight irradiation is strong. In such a case, for an image obtained by one camera, it is difficult to remove a shadow region by image processing using inter-frame difference, background difference, or edge extraction processing, and an object and shadow region is extracted, and only the object is extracted. It was extremely difficult to detect. This is because the luminance of the pixels in these regions changes greatly compared to the luminance of the corresponding pixel in the background image, whether it is an object or a shadow.

JP, 2001-229389, A There is technology called stereo image processing which performs processing by obtaining images from different directions using two cameras, but in a normal monitoring system, there is a camera for one target. Is a single unit, and performing stereo image processing is not realistic. Even if stereo image processing is possible, there is a problem that the calculation load is heavy and a dedicated image processing board is required.

  Conventionally, attempts have been made to perform shadow detection using an image obtained by one camera. For example, in the case of the one described in Patent Document 1, images obtained before and after time series from the same camera. The intensity ratio is calculated using the data, and the shadow area is detected for the pixel area having the intensity ratio obtained by the above calculation using the standard intensity ratio of the shadow area.

  In the above-described conventional example, it is necessary to devise in order to obtain an appropriate value as a standard such as a standard intensity ratio of the shadow region, and much of the processing is spent. In short, the accuracy of shadow area detection depends on the quality of the referenced intensity ratio.

  The problem to be solved is that the shadow area cannot be accurately detected by a simple calculation using an image obtained from one camera.

  The shadow detection apparatus according to the present invention obtains a difference image by a calculation for obtaining an absolute value of a difference between a background image and an input image, binarizes the difference image, and detects a white pixel region as a calculation region. And a specifying means for specifying, as the predetermined value attenuation pixel, a pixel in which the pixel luminance of the input image has a luminance attenuation of a predetermined value as compared with the pixel luminance of the background image. The attenuation pixel region is detected as a shadow.

  The shadow detection apparatus according to the present invention obtains a difference image by a calculation for obtaining an absolute value of a difference between a background image and an input image, binarizes the difference image, and detects a white pixel region as a calculation region. A means for specifying, as the predetermined value attenuation pixel, a pixel in which the pixel luminance of the input image has a luminance attenuation of a predetermined value compared to the pixel luminance of the background image with respect to the calculation area; and the predetermined value attenuation pixel A first average luminance calculating means for calculating a first average luminance that is an average luminance of the pixel luminance in the input image; and a second average luminance that is an average luminance of the pixel luminance in the difference image for the predetermined value attenuation pixel. Based on the difference between the first average brightness and the second average brightness, the brightness of the predetermined value attenuation pixel in the difference image is calculated based on the difference between the first average brightness and the second average brightness. An offset value calculating means for calculating an offset value for approaching one average luminance, a difference between the first average luminance and the second average luminance, and a luminance in the difference image, for pixels in the calculation area; An arithmetic means for calculating the luminance to be a corrected luminance close to the first average luminance, and an absolute value of a difference between the pixel luminance in the input image and the corrected luminance for a pixel in the calculation area in the input image; The offset value is compared, and a sorting means for classifying the shadow area pixel into another area based on the size of the offset value is provided, and the area of the shadow area pixel is detected as a shadow.

  An image processing apparatus according to the present invention includes the shadow detection apparatus according to claim 1, and performs image processing using an area obtained by excluding the area of the predetermined value attenuation pixel from the calculation area as an object area. .

  An image processing apparatus according to the present invention includes the shadow detection apparatus according to claim 2, and performs image processing using an area defined as an area other than a shadow area pixel by the classification unit as an object area.

  The image processing apparatus according to the present invention further compares the ratio of the number of pixels in the shadow region and the number of pixels in the region other than the shadow region with a predetermined threshold value, and removes the region determined as the shadow region from the object region. It is characterized by comprising a shadow removal judging means for judging whether or not.

  The shadow detection method according to the present invention obtains a difference image by calculating an absolute value for a difference between a background image and an input image, binarizes the difference image, and detects a white pixel region as a calculation region. And a step of specifying, as the predetermined value attenuation pixel, a pixel in which the pixel luminance of the input image has a luminance attenuation of a predetermined value compared to the pixel luminance of the background image, as the predetermined value attenuation pixel. The attenuation pixel region is detected as a shadow.

  The shadow detection method according to the present invention obtains a difference image by calculating an absolute value for a difference between a background image and an input image, binarizes the difference image, and detects a white pixel region as a calculation region. A step of identifying, for the calculation area, a pixel in which a pixel luminance of the input image is compared with a pixel luminance of a background image, and a pixel having a predetermined value attenuation as a predetermined value attenuation pixel, and the predetermined value attenuation pixel A first average luminance calculation step of calculating a first average luminance that is an average luminance of pixel luminance in the input image; and a second average luminance that is an average luminance of pixel luminance in the difference image for the predetermined value attenuation pixel The predetermined average attenuation pixel in the difference image based on a second average luminance calculation step for calculating the difference between the first average luminance and the second average luminance. An offset value calculating step for calculating an offset value for bringing the luminance of the first average luminance closer to the first average luminance, a difference between the first average luminance and the second average luminance, and a luminance in the difference image, A calculation step for calculating the luminance of the pixel in the calculation area in the difference image as a corrected luminance close to the first average luminance; and a pixel luminance and a difference in the input image for the pixel in the calculation area in the input image A step of comparing the absolute value of the difference with the corrected brightness in the image and the offset value, and classifying the offset area pixel into a shadow area pixel and another area based on the magnitude, It is characterized by detecting as.

  An image processing method according to the present invention is characterized in that the shadow detection method according to claim 6 is used, and image processing is performed using an area obtained by excluding the area of the predetermined value attenuation pixel from the calculation area as an object area. .

  An image processing method according to the present invention is characterized in that the shadow detection method according to claim 7 is used, and image processing is performed using an area determined as an area other than a shadow area pixel by the classification step as an object area.

  The image processing method according to the present invention further compares the ratio of the number of pixels in the shadow region and the number of pixels in the region other than the shadow region with a predetermined threshold value, and removes the region determined as the shadow region from the object region. A shadow removal determining step is provided.

  A shadow detection program according to the present invention is a program executed by a computer, obtains a difference image by an operation for obtaining an absolute value of a difference between a background image and an input image, binarizes the difference image, and generates white pixels. A calculation area detecting step for detecting an area as a calculation area; and for the calculation area, a pixel in which the luminance of the pixel of the input image is compared with the luminance of the pixel of the background image is identified as a predetermined value attenuation pixel. And a specific step, wherein the region of the predetermined value attenuation pixel is detected as a shadow.

  A shadow detection program according to the present invention is a program executed by a computer, obtains a difference image by an operation for obtaining an absolute value of a difference between a background image and an input image, binarizes the difference image, and generates white pixels. A calculation area detecting step for detecting an area as a calculation area; and for the calculation area, a pixel in which the luminance of the pixel of the input image is compared with the luminance of the pixel of the background image is identified as a predetermined value attenuation pixel. A specifying step; a first average luminance calculating step for calculating a first average luminance that is an average luminance of pixel luminance in the input image for the predetermined value attenuation pixel; and a pixel luminance in the difference image for the predetermined value attenuation pixel. A second average luminance calculation step of calculating a second average luminance that is an average luminance of the first average luminance, and a difference between the first average luminance and the second average luminance An offset value calculating step for calculating an offset value for bringing the luminance of the predetermined value attenuation pixel in the difference image close to the first average luminance; the first average luminance and the second average luminance; And a calculation step for calculating the luminance of the pixel in the calculation area in the difference image as a corrected luminance close to the first average luminance, based on the difference in the difference image and the luminance in the difference image, and the calculation area in the input image A step of comparing the absolute value of the difference between the pixel luminance in the input image and the corrected luminance in the difference image with the offset value and classifying the pixel into a shadow region pixel and other regions based on the magnitude And the area of the shadow area pixel is detected as a shadow.

  An image processing program according to the present invention includes the shadow detection program according to claim 11, and performs image processing using an area obtained by removing the area of the predetermined value attenuation pixel from the calculation area as an object area. .

  An image processing program according to the present invention includes the shadow detection program according to claim 12, and performs image processing using an area determined as an area other than a shadow area pixel by the classification step as an object area.

  The image processing program according to the present invention further compares the ratio between the number of pixels in the shadow area and the number of pixels in the area other than the shadow area with a predetermined threshold value to determine whether or not to remove the area as the shadow area from the object area. A shadow removal determining step is provided.

  In the present invention, a difference image is obtained by an operation for obtaining an absolute value of a difference between a background image and an input image, the difference image is binarized, and a white pixel region is set as a calculation region. Since the background image pixel in which the luminance attenuation of the predetermined value is seen compared with the luminance is specified as the predetermined value attenuation pixel and the region of the predetermined value attenuation pixel is detected as a shadow, the calculation region constituted by the shadow and the object It is possible to easily distinguish a shadow from an object simply by detecting a predetermined value of luminance attenuation.

  In the present invention, with respect to an area of a predetermined value attenuation pixel that is considered to be a shadow area, attention is paid to an offset based on the difference between the luminance of the input image and the luminance of the difference image, and the luminance correction is performed on the pixels in the calculation area in the difference image. If the absolute value of the difference between the pixel brightness in the input image and the pixel brightness in the input image exceeds the offset, the shadow area and target area are attenuated by a predetermined value. It is possible to obtain a shadow or a region of an object well.

  In the present invention, the ratio between the number of pixels in the shadow area and the number of pixels in the area other than the shadow area is compared with a predetermined threshold value to determine whether or not to remove the area designated as the shadow area from the object area. In such a case, it is possible to appropriately detect the object region by removing the shadow.

  The present inventor has discovered that the object of appropriately dividing the area between the shadow and the object can be easily realized by the following principle. That is, the following differences are observed between the target object and its shadow in the input image. First, the pattern (texture) of the background portion is lost in the object area. On the other hand, in the shadow area, the pattern (texture) of the background portion remains relatively good. Further, in the difference image obtained by taking the difference between the input image and the background image, the following differences are observed in the object and the shadow portion. First, the pattern (texture) of the background portion is lost in the object area. On the other hand, in the shadow area, the pattern (texture) of the background portion remains relatively good. Therefore, the brightness of the shadow pattern area where the background pattern (texture) remains relatively good and the area of the target object where the background pattern (texture) has been lost are represented by the brightness of the background pattern (texture). Shadow detection (from the opposite, object detection) and image processing are performed. Hereinafter, with reference to the accompanying drawings, a shadow detection apparatus, a shadow detection method and a shadow detection program according to the present invention, an image processing apparatus using the shadow detection apparatus, an image processing method using the shadow detection method, and an image using the shadow detection program A processing program will be described. In each figure, the same components are denoted by the same reference numerals, and redundant description is omitted.

  FIG. 1 is a block diagram of an image processing apparatus according to the present invention. The image processing device incorporates the function of a shadow detection device. A camera 1 is provided as image input means or image acquisition means. An image signal (for example, NTSC system) obtained by the camera 1 is sent to the image conversion unit 2, converted into digital luminance data consisting of i × j pixels in the image conversion unit 2, and taken into the image processing unit 10. It is processed. The image processing unit 10 is connected to an output unit 3 that is a display device or the like and an input unit 4 that is a keyboard or a mouse.

  The image processing unit 10 includes a background image creation / storage unit 11. For example, an instruction is input from the input unit 4 for a desired image among a plurality of images sent from the camera 1 and displayed on the output unit 3. Then, it is stored in the background image creation / storage unit 11 as a background image. An example of the background image stored in the background image creation / storage unit is a road image that does not include a vehicle as shown in FIG.

  Further, the image processing unit 10 is provided with a calculation area detecting unit 12 and a specifying unit 13. The calculation area detection unit 12 obtains a difference image by calculation for obtaining an absolute value of the difference between the background image and the input image, binarizes the difference image, and detects a white pixel area as the calculation area. When the input image includes the vehicle C and its shadow S as shown in FIG. 3, the difference image is as shown in FIG. A binarized image obtained by binarizing the difference image is as shown in FIG. A region that appears white in the binarized image of FIG. 5 is a calculation region.

  The specifying unit 13 uses, as the predetermined value attenuation pixel, a pixel in which the pixel luminance of the input image shows a luminance attenuation of a predetermined value (attenuation of the luminance of the road surface by the object) compared to the pixel luminance of the background image in the calculation area. It is something to identify. Here, the predetermined value is 60%, for example. The specifying unit 13 compares the pixel luminance of the input image with the pixel luminance of the background image for the calculation region appearing white in the binarized image of FIG. 5, and the pixel luminance of the input image is 60% of the pixel luminance of the background image. Are extracted, and these pixels are specified as predetermined value attenuation pixels. If the luminance of the specified predetermined value attenuation pixel region is white, and the other region is black, a shadow region appearing as a white region in FIG. 6 is detected. The object region obtained by removing the shadow region in FIG. 6 from the calculation region shown in FIG. 5 can be labeled and processed to extract the vehicle region.

  The image processing apparatus other than the camera 1 in FIG. 1 is configured by a personal computer or a workstation having a configuration as shown in FIG. 7, for example. That is, the computer shown in FIG. 2 has a CPU 51 that performs overall control of the apparatus, and a main storage device 52 that stores information such as programs and data used by the CPU 51 is connected to the CPU 51. Further, a keyboard control unit 54, a display control unit 55, a printer control unit 56, an input interface 57, a mouse control unit 58, and a magnetic disk control unit 59 are connected to the CPU 51 via a system bus 53. The keyboard control unit 54 is connected to a keyboard input device 60 capable of inputting various information, the display control unit 55 is connected to a CRT display device 61 for displaying information, and the printer control unit 56 prints information. A printer device 62 for output is connected, a processing unit 63 for capturing an image signal from the monitoring camera 1 and performing A / D conversion and multi-value conversion is connected to the input interface 57, and a pointing unit is connected to the mouse control unit 58. A mouse 64 as a device is connected, and a magnetic disk device 65 as an auxiliary storage device is connected to the magnetic disk control unit 59. A flexible disk drive, a magnetic card or IC card reader, an MO (magneto-optical disk) drive, etc. are provided as necessary.

  In the above computer, for example, an image processing program corresponding to the flowchart shown in FIG. 8 is provided in the magnetic disk device 65, and this is read into the main storage device 52 and executed, so that it is shown in FIG. Since image processing is performed as each means, the operation will be described based on the above flowchart.

  First, an image signal is captured from the processing unit 63 (S1), and a background image is created (S2). As described above, this processing is performed, for example, by inputting an instruction from the keyboard input device 60 serving as the input unit 4 for a desired image in a plurality of images sent from the camera 1 and displayed on the CRT display device 61 serving as the output unit 3. Then, it is stored in the main storage device 52 as a background image.

  Next, an image signal is acquired from the processing unit 63 for the currently obtained input image, and difference processing is performed by calculating an absolute value of the difference between the input image and the previously obtained background image (S3). As a result, when the input image is as shown in FIG. 3 and the background image is as shown in FIG. 2, the difference image is as shown in FIG. Next, the binarization threshold value is automatically calculated (S4). The binarization threshold value can be obtained, for example, as an average value of luminance at pixels located on both sides of the boundary by paying attention to the boundary where luminance greatly varies.

  Further, using the threshold value obtained in step S4, a binarization process is performed on the difference image shown in FIG. 4 to create a binarized image, and a pixel in which a luminance change has occurred is extracted. The configured area is set as a calculation area (S5: calculation area detection step). A region that appears white in the binarized image of FIG. 5 is a calculation region.

  Next, for the calculation area, the pixel luminance of the input image is compared with the pixel luminance of the background image, and a pixel in which luminance attenuation of a predetermined value is seen is specified as the predetermined value attenuation pixel (S6: specifying step). For example, for the calculation area appearing white in FIG. 5, the pixel luminance of the input image is compared with the pixel luminance of the background image, and a pixel in which the pixel luminance of the input image is 60% of the pixel luminance of the background image is extracted. These pixels are specified as predetermined value attenuation pixels.

  Next, in the binarized image of FIG. 5, a shadow area is obtained by setting the remaining area in the calculation area as a black pixel while keeping the predetermined value attenuation pixel white (S7). In contrast to the above, in the binarized image of FIG. 5, the predetermined value attenuation pixel is black, and the remaining area in the calculation area is left as a white pixel, thereby obtaining an object (vehicle) area by the white pixel area. (S8).

  FIG. 9 shows a configuration example of an image processing apparatus incorporating the function of the shadow detection apparatus according to the second embodiment. In this image processing apparatus, in addition to the configuration shown in FIG. 1, an image having a first average luminance calculation means 14, a second average luminance calculation means 15, an offset value calculation means 16, a calculation means 17 and a sorting means 18. A processing unit 10A is provided.

  Also in the second embodiment, the calculation area detecting means 12 obtains a difference image by calculating the absolute value of the difference between the background image and the input image, and binarizes the difference image to calculate the white pixel area. The area is detected, and the specifying unit 13 specifies, as the predetermined value attenuation pixel, the pixel in which the pixel luminance of the input image is compared with the pixel luminance of the background image in the calculation area. The processes performed by the first average luminance calculation unit 14, the second average luminance calculation unit 15, the offset value calculation unit 16, the calculation unit 17, and the sorting unit 18 are all performed on the calculation area.

Here, terms used in the following image processing are defined.
(i, j): Index indicating the pixel position in the image
I (i, j): Input image
BG (i, j): Background image
D (i, j): Difference image
Bin (i, j): Binary image obtained by binarizing the difference image
Coef1: Decay rate of luminance in the background image
Coef2: Coefficient for determining the luminance offset value when extracting the object area from the input image and the difference image
Bright1: Average brightness of pixels in the input image in the area estimated to be a shadow
Bright2: Average brightness of pixels in the difference image in the area estimated to be a shadow
offset: Brightness correction value to bring the brightness of the pixels in the calculation area of the difference image closer to Bright1
D _mod (i, j): Difference image after brightness correction
Bin _final (i, j): Final binarized image extracted mainly from pixels of the region estimated to be the target in the input image
WP num: Total number of white pixels in the binarized image obtained by binarizing the difference image
SP num: Total number of pixels in the shadow area
Threshold: Threshold value that determines the necessity of shadow removal processing

When the definition is performed as described above, the process performed by the specifying unit 13 is a process of specifying a pixel that satisfies a condition according to the following (Expression 1) for the calculation region. Note that the default value of Coef1 in (Equation 1) is, for example, 0.6.
I (i, j) <Coef1 × BG (i, j) (Formula 1)

  Then, the first average luminance calculation means 14 obtains the average luminance Bright1 for all the pixels obtained by the above (Equation 1). The second average luminance calculation means 15 obtains the second average luminance Bright2 of the pixels in the area estimated as a shadow for the calculation area in the background image. The second average luminance calculation means 15 compares the obtained second average luminance Bright2 with the luminance of the pixels in the calculation area in the difference image, and determines the luminance of the pixels whose luminance is higher than the second average luminance Bright2 as the second average luminance. Replace the brightness Bright2 and smooth the part with the projected brightness value.

Based on the difference between the first average brightness Bright1 and the second average brightness Bright2, the offset value calculation means 16 is an offset value for bringing the brightness of a predetermined value attenuation pixel in the difference image closer to the first average brightness Bright2. The offset is calculated by the following (Expression 2). Coef2 is, for example, about “2”.
offset = | (Coef2 × (Bright1-Bright2)) | ... (Formula 2)

Based on the difference between the first average brightness Bright1 and the second average brightness Bright2 and the brightness in the difference image, the calculation means 17 brings the brightness of the pixels in the calculation area in the difference image closer to the first average brightness Bright1. By performing the calculation to obtain the corrected luminance, the difference image D _mod (i, j) after performing the luminance correction shown in the following (Equation 3) is obtained.
D _mod (i, j) = D (i, j) − (Bright2-Bright1) (Equation 3)

The sorting unit 18 calculates the absolute value of the difference between the pixel luminance I (i, j) in the input image and the corrected luminance D _mod (i, j) obtained by the calculating unit 17 for the pixels in the calculation area in the input image. The offset value offset is compared and divided into shadow area pixels and other areas based on the magnitude. The classifying unit 18 determines that a pixel satisfying the following (Expression 4) is a pixel belonging to the object region, and other pixels are pixels belonging to the shadow region.
| I (i, j) −D _mod (i, j) |> offset (Expression 4)

Then, the image processing unit 10A replaces the pixels belonging to the object area with white pixels, and replaces the other pixels with black pixels. The final binarized image Bin _final (i, j) ( A binarized image in which pixels in a region estimated as an object in the input image are extracted. A labeling process is performed on the final binarized image Bin _final (i, j) obtained by this process to extract an object (vehicle) region.

  In the computer shown in FIG. 7, for example, the magnetic disk device 65 includes an image processing program corresponding to the flowcharts shown in FIGS. 10 and 11 in the computer shown in FIG. Since the image processing is performed as each unit shown in FIG. 9 by reading and executing, the operation will be described based on the above flowchart.

  Steps S1 to S6 are the same as those in the first embodiment. Next, an average luminance (first average luminance Bright1) of all pixels satisfying (Expression 1) is obtained in the calculation area (S11: first average luminance calculation step). Next, an average luminance (second average luminance Bright2) of pixels in an area estimated as a shadow is calculated for the calculated image in the difference image (S12: second average luminance calculation step).

Following step S12, smoothing is performed to replace the luminance in the pixels having a luminance higher than the second average luminance Bright2 with the second average luminance Bright2 in the calculation area of the difference image (S13). Further, an offset offset is obtained from (Expression 2) (S14: offset value calculation step). Next, a difference image D _mod (i, j) corrected for luminance based on (Equation 3) is obtained (S15: calculation step). By this luminance correction, the original difference image (in the calculation region) is brought close to the average luminance of the pixels in the input image in the region estimated as a shadow.

  Next, a pixel satisfying (Expression 4) is set as a pixel belonging to the object region (S16: classification step). This (Equation 4) is the difference between the difference image related to the luminance correction brought close to the average luminance of the pixels in the input image and the pixels in the input image in the region where the original difference image (in the calculation region) is estimated as a shadow. It is detected whether the difference exceeds the offset value offset (the brightness correction value for bringing the brightness of the pixels in the calculation area in the difference image closer to Bright1). Extremely speaking, by replacing the luminance of the region S1 (shadow region) of the difference image shown in FIG. 4 with the luminance of the shadow of the input image shown in FIG. Since the area has a value in the input image shown in FIG. 3 and the object area has a value in the difference image in FIG. 4, the value in the absolute value of Equation 4 is the offset value offset (the shadow in FIG. 3). The difference between the brightness of the area and the shadow area in FIG. 4 is equal to or less than that, but the object area is predicted to take a value much larger than the offset value offset. The object region is detected using this principle.

Subsequent to step S16, binarization is performed by replacing pixels belonging to the object area with white pixels and the remaining pixels with black pixels (S17), and final binarization obtained by the above binarization. The image Bin _final (i, j) is labeled to extract an object (vehicle) region (S18). Thus, it is possible to extract a target (vehicle) region with higher accuracy.

  FIG. 12 shows a configuration example of an image processing apparatus according to the third embodiment incorporating a shadow removal determination function. In addition to the configuration shown in FIG. 9, the image processing apparatus includes an image processing unit 10 </ b> B having a shadow removal determination unit 19. The shadow removal determining means 19 compares the ratio of the number of pixels in the shadow area and the number of pixels in the area other than the shadow area with a predetermined threshold value, and determines whether or not to remove the area determined as the shadow area from the object area. To do.

Specifically, the total number WP num of white pixels in the binarized image with respect to the difference image shown in FIG. 5 and the total number SP of pixels in the shadow area obtained by the sorting unit 18. Use num to detect the presence or absence of shadows based on whether the following conditions are met:
(SP num / WP num)> Threshold (0.2 by default) ... (Formula 5)

  If the above (Formula 5) is satisfied, the shadow area and the object area are extracted by the process shown in the second embodiment, assuming that there is a shadow area. On the other hand, if (Equation 5) is not satisfied, the white area of the binarized image of FIG. 5 which is the result of binarization of the background difference image shown in FIG. Extract as

  In the computer shown in FIG. 7, for example, the above-described image processing apparatus has an image processing program corresponding to the flowcharts shown in FIGS. 10, 11, and 13 in the magnetic disk device 65. Since the image processing is performed as each unit shown in FIG. 12 by reading and executing, the operation will be described based on the above flowchart.

The processing from step S1 to step S16 in the flowcharts of FIGS. 10 and 11 is not changed. Following step S16, the process proceeds to the process shown in the flowchart of FIG. The total number WP of white pixels in the binarized image for the difference image shown in FIG. num and the total number SP of shadow region pixels obtained by the sorting means 18 Using num, the presence or absence of a shadow is detected based on whether or not the condition of (Expression 5) is satisfied (S21, S22: shadow removal determination step).

  If it is detected in the above step S22 that (Equation 5) is satisfied, the shadow area and the object area are extracted by the processing shown in steps S17 and S18 in the second embodiment, assuming that there is a shadow area. Do. On the other hand, if (Equation 5) is not satisfied, the white area of the binarized image of FIG. 5 which is the result of binarization of the background difference image shown in FIG. (S23).

  According to this embodiment, it is determined whether or not it is necessary to remove the shadow, and the object area from which the shadow is removed is detected when necessary.

  When performing various types of monitoring such as intrusion monitoring as well as detection of a vehicle region on a road, shadow removal can be performed and applied to detection of an appropriate target region.

1 is a block diagram of an image processing apparatus that employs a shadow detection apparatus according to a first embodiment of the present invention. The figure which shows the background image used for an image process in the Example of this invention. The figure which shows the input image used for an image process in the Example of this invention. The figure which shows the difference image obtained in the Example of this invention. The figure which shows the binarized image which binarized the difference image of FIG. The figure which shows the image which replaced the target object area | region with the black pixel about the binarized image of FIG. The figure which shows the structure of the computer which implement | achieves the image processing apparatus of this invention. 5 is a flowchart for explaining the operation of the image processing apparatus employing the shadow detection apparatus according to the first embodiment of the present invention. The block diagram which shows the image processing apparatus which employ | adopted the shadow detection apparatus which concerns on 2nd Example of this invention. 9 is a flowchart for explaining the operation of an image processing apparatus employing a shadow detection apparatus according to a second embodiment of the present invention. 9 is a flowchart for explaining the operation of an image processing apparatus employing a shadow detection apparatus according to a second embodiment of the present invention. The block diagram which shows the image processing apparatus which employ | adopted the shadow detection apparatus which concerns on 3rd Example of this invention. 10 is a flowchart for explaining the operation of an image processing apparatus employing a shadow detection apparatus according to a third embodiment of the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Camera 2 Image conversion part 3 Output part 4 Input part 10, 10A, 10B Image processing part 11 Background image preparation / memory | storage part 12 Calculation area | region detection means 13 Specification means 14 1st average brightness calculation means 15 2nd average brightness calculation Means 16 Offset value calculation means 17 Calculation means 18 Classification means 19 Shadow removal determination means

Claims (15)

A calculation area detecting means for obtaining a difference image by calculating an absolute value for a difference between the background image and the input image, binarizing the difference image, and detecting a white pixel area as a calculation area;
A specifying means for specifying, as the predetermined value attenuation pixel, a pixel in which the pixel luminance of the input image has a luminance attenuation of a predetermined value compared to the pixel luminance of the background image with respect to the calculation region;
A shadow detection apparatus that detects a region of the predetermined value attenuation pixel as a shadow. A calculation area detecting means for obtaining a difference image by calculating an absolute value for a difference between the background image and the input image, binarizing the difference image, and detecting a white pixel area as a calculation area;
For the calculation area, a specifying unit that specifies, as a predetermined value attenuation pixel, a pixel in which the pixel luminance of the input image is compared with the pixel luminance of the background image and a predetermined value of luminance attenuation is seen
A first average luminance calculating means for calculating a first average luminance that is an average luminance of the pixel luminance in the input image for the predetermined value attenuation pixel;
A second average luminance calculating means for calculating a second average luminance that is an average luminance of the pixel luminance in the difference image for the predetermined value attenuation pixel;
Based on the difference between the first average luminance and the second average luminance, offset value calculation for calculating an offset value for bringing the luminance of the predetermined value attenuation pixel in the difference image closer to the first average luminance. Means,
Based on the difference between the first average luminance and the second average luminance and the luminance in the difference image, the corrected luminance in which the luminance of the pixels in the calculation area in the difference image is close to the first average luminance An arithmetic means for performing an arithmetic operation,
For the pixels in the calculation region in the input image, the absolute value of the difference between the pixel luminance in the input image and the corrected luminance in the difference image is compared with the offset value, and the shadow region pixel and the other region are based on the magnitude And classifying means for classifying
A shadow detection apparatus that detects a region of the shadow region pixel as a shadow. The shadow detection apparatus according to claim 1,
An image processing apparatus that performs image processing using an area obtained by excluding the area of the predetermined value attenuation pixel from the calculation area as an object area. A shadow detection apparatus according to claim 2,
An image processing apparatus that performs image processing using an area defined as an area other than a shadow area pixel by the classification unit as an object area. A shadow removal determination unit is provided that compares the ratio of the number of pixels in the shadow region with the number of pixels in the region other than the shadow region and a predetermined threshold to determine whether or not to remove the region made the shadow region from the object region. The image processing apparatus according to claim 4. A calculation area detection step of obtaining a difference image by calculating an absolute value for a difference between the background image and the input image, binarizing the difference image, and detecting a white pixel area as a calculation area;
A step of specifying, as the predetermined value attenuation pixel, a pixel in which the pixel luminance of the input image has a luminance attenuation of a predetermined value compared to the pixel luminance of the background image with respect to the calculation region;
A shadow detection method, wherein the predetermined value attenuation pixel region is detected as a shadow. A calculation area detection step of obtaining a difference image by calculating an absolute value for a difference between the background image and the input image, binarizing the difference image, and detecting a white pixel area as a calculation area;
A specific step of specifying, as the predetermined value attenuation pixel, a pixel in which the pixel luminance of the input image is compared with the pixel luminance of the background image and a predetermined value of luminance attenuation is seen for the calculation region;
A first average luminance calculating step for calculating a first average luminance which is an average luminance of the pixel luminance in the input image for the predetermined value attenuation pixel;
A second average luminance calculating step of calculating a second average luminance that is an average luminance of the pixel luminance in the difference image for the predetermined value attenuation pixel;
Based on the difference between the first average luminance and the second average luminance, offset value calculation for calculating an offset value for bringing the luminance of the predetermined value attenuation pixel in the difference image closer to the first average luminance. Steps,
Based on the difference between the first average luminance and the second average luminance and the luminance in the difference image, the luminance of the pixel in the calculation area in the difference image is adjusted to be close to the first average luminance. An operation step for performing an operation to be performed;
For the pixels in the calculation area in the input image, the absolute value of the difference between the pixel luminance difference in the input image and the corrected luminance in the image is compared with the offset value, and the shadow area pixel and the other area are compared based on the magnitude. A classification step for classifying, and
A shadow detection method for detecting a region of the shadow region pixel as a shadow. Using the shadow detection method according to claim 6,
An image processing method, wherein image processing is performed using an area obtained by excluding the area of the predetermined value attenuation pixel from the calculation area as an object area. Using the shadow detection method according to claim 7,
An image processing method characterized in that image processing is performed using an area defined as an area other than a shadow area pixel by the classification step as an object area. A shadow removal determining step is provided for determining whether or not to remove the area that is the shadow area from the object area by comparing a predetermined threshold with a ratio between the number of pixels in the shadow area and the number of pixels in the area other than the shadow area. The image processing method according to claim 9. A program executed on a computer,
A calculation area detection step of obtaining a difference image by calculating an absolute value for a difference between the background image and the input image, binarizing the difference image, and detecting a white pixel area as a calculation area;
A step of specifying, as the predetermined value attenuation pixel, a pixel in which the pixel luminance of the input image has a luminance attenuation of a predetermined value compared to the pixel luminance of the background image with respect to the calculation region;
A shadow detection program for detecting a region of the predetermined value attenuation pixel as a shadow. A program executed on a computer,
A calculation area detection step of obtaining a difference image by calculating an absolute value for a difference between the background image and the input image, binarizing the difference image, and detecting a white pixel area as a calculation area;
A specific step of specifying, as the predetermined value attenuation pixel, a pixel in which the pixel luminance of the input image is compared with the pixel luminance of the background image and a predetermined value of luminance attenuation is seen for the calculation region;
A first average luminance calculating step for calculating a first average luminance which is an average luminance of the pixel luminance in the input image for the predetermined value attenuation pixel;
A second average luminance calculating step of calculating a second average luminance that is an average luminance of the pixel luminance in the difference image for the predetermined value attenuation pixel;
Based on the difference between the first average luminance and the second average luminance, offset value calculation for calculating an offset value for bringing the luminance of the predetermined value attenuation pixel in the difference image closer to the first average luminance. Steps,
Based on the difference between the first average luminance and the second average luminance and the luminance in the difference image, the luminance of the pixel in the calculation area in the difference image is adjusted to be close to the first average luminance. An operation step for performing an operation to be performed;
For the pixels in the calculation area in the input image, the absolute value of the difference between the pixel brightness in the input image and the corrected brightness in the difference image is compared with the offset value, and the shadow area pixel and other areas are compared based on the magnitude. A classification step for classifying, and
A shadow detection program for detecting an area of the shadow area pixel as a shadow. A shadow detection program according to claim 11,
An image processing program for performing image processing using an area obtained by excluding the area of the predetermined value attenuation pixel from the calculation area as an object area. A shadow detection program according to claim 12,
An image processing program for performing image processing using an area determined as an area other than a shadow area pixel by the classification step as an object area. A shadow removal determining step is provided for determining whether or not to remove the area that is the shadow area from the object area by comparing a predetermined threshold with a ratio between the number of pixels in the shadow area and the number of pixels in the area other than the shadow area. The image processing program according to claim 14.

Images