JPH11146410A - Video camera - Google Patents

Abstract

PROBLEM TO BE SOLVED: To eliminate noise generated by crosstalk in a camera provided with a CCD sensor in which the three kinds of picture elements (first, second and third picture elements) are disposed in a grid shape and the first picture elements are disposed in a checkered pattern. SOLUTION: This video camera is composed of a first subtraction means 26 for calculating the signal level difference of the noticed first picture element on a noticed horizontal line and the picture element adjacent to it, a second subtraction means 26 for calculating the level difference of signals outputted from the two picture elements corresponding to the above-mentioned two picture elements on a line adjacent to the noticed line and a means 28 for correcting the signal level of the noticed first picture element to a value obtd. by multiplying the signal levels of the noticed first picture element and the four first picture elements obliquely adjacent to the noticed first picture element by a prescribed coefficient and adding them in the case that the level difference between the absolute values of the outputs of the first and second subtraction means 26 is more than a prescribed value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a video camera having a charge-coupled device (hereinafter referred to as "CCD") in which each pixel generates a signal corresponding to one of three primary colors as an image pickup device.

[0002]

2. Description of the Related Art In a camera, noise is generated in an image signal due to various factors. Eliminating this noise is indispensable for obtaining the clearest possible image. In a camera using a CCD configured to generate a signal of a different color between adjacent pixels as an image sensor, a major cause of noise generation is crosstalk between adjacent pixels. The principle of noise generation due to crosstalk will be described.

If there is a difference between the output levels of image signals between adjacent pixels that are successively read out, in the analog processing circuit system, the image signals affect each other, and their signal levels deviate from the original signal levels. Would. This phenomenon is referred to as crosstalk in this specification. This crosstalk becomes more remarkable as the signal level difference between adjacent pixels increases.

[0004] The occurrence of noise due to crosstalk means that a difference in signal level difference between adjacent pixels causes a difference in the magnitude of the influence of crosstalk, and this difference causes noise. The generation of noise due to crosstalk will be described using a specific example with reference to FIGS.

FIG. 7 shows a mosaic filter adhered to a CCD. GREEN (G) of primary colors, BLUE
(B) and RED (R) color filters are alternately arranged in a grid pattern. Regarding the horizontal line, RG lines and GB lines are alternately arranged. The G pixels are arranged in a checkered pattern. Each pixel of the CCD generates a signal of the color of the corresponding filter. The charges of the pixels of the CCD are sequentially read out for each horizontal line, that is, RG lines and GB lines are alternately read. According to such a primary color filter, the signal level difference between adjacent pixels is particularly large, and crosstalk is likely to occur.

An RG line which is read continuously and a GB line
When the image signal of the line is as shown in FIG. 8A, the signal level difference between the adjacent pixels of the RG line and the signal level difference of the adjacent pixels of the BG line are substantially the same, and the crosstalk received by the G pixel Is almost the same for both lines. Therefore, the deviation W of the G pixel from the black signal level 80 is substantially the same in both lines. FIG. 8B shows a case where G pixels are interpolated by averaging adjacent pixels for each of these two lines.

On the other hand, when the difference between the signal level difference between the adjacent pixels on the RG line and the signal level difference between the adjacent pixels on the GB line is large, and the respective image signals are as shown in FIG. The effect of crosstalk on the two lines is different. That is, in the GB line, since the signal level difference between the adjacent pixels is large, the G pixel is greatly influenced by the crosstalk by the B pixel and is pulled upward, and the deviation W from the black signal level 80 increases. On the other hand, the signal level shift W of the G pixel on the RG line is smaller than that on the GB line.

Therefore, when G interpolation is performed on these two lines by averaging adjacent pixels, that is, R pixels (in the case of RG lines) and B pixels (in the case of GB lines) existing between G pixels. Is interpolated by the G signal (for example, by interpolating the average of the G pixels of the lines adjacent above and below each line), dot noise is generated as shown in FIG. 8D. This noise is generated due to crosstalk.

[0009] Even when pixel interpolation is not performed, there is a difference between the signal levels of the G pixels on the RG line and the GB line (almost the same level unless the boundary of the color of the subject changes). Therefore, for example, when this data is displayed on the display means, the image becomes a noise-carrying image in which the shading changes alternately for each line.
Such noise is also generated due to crosstalk.

As a method for preventing the occurrence of noise as described above, a method has been proposed in which the degree of color modulation of the entire image is reduced by filtering. According to this method, the signal level difference between adjacent pixels is reduced, and the influence of crosstalk is also reduced. Therefore, the occurrence of noise in the interpolated data due to the difference in the degree of influence of crosstalk is reduced.

[0011]

However, if the above method is adopted to prevent the generation of noise, the sharpness of the entire image will be reduced. In addition, if the resolution of the image is reduced, it is considered that the influence of crosstalk is reduced and the occurrence of noise is also reduced.

As described above, in the conventional method, there is a problem in that the resolution and sharpness of the entire image are reduced due to the removal of noise generated by crosstalk.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a video camera which can reduce the color reproducibility of an image and remove noise generated by crosstalk without deteriorating the resolution.

[0014]

In order to achieve the above object, the present invention provides first, second, and third pixels for generating signals corresponding to one of three colors, respectively. A first line in which pixels are alternately arranged in the horizontal direction, and a second line in which first and third pixels are alternately arranged in the horizontal direction are alternately arranged in the vertical direction. In a video camera having, as an image sensor, a charge-coupled device repeatedly arranged so as to be located every other pixel, first subtraction means for calculating a level difference between a first pixel of interest on a line of interest and a signal output from a pixel adjacent thereto. And a second pixel which is a pixel on a line adjacent to the line of interest and has a level difference between signals output from two pixels corresponding to the two pixels.
When the level difference between the absolute values of the outputs of the subtraction means and the first and second subtraction means is equal to or greater than a predetermined value, the level of the signal output from the first pixel of interest is changed to the signal output from the first pixel of interest. Means for correcting a level and a signal level output from a plurality of first pixels around the noted first pixel by multiplying each by a predetermined coefficient and correcting the sum to a value;
The predetermined coefficient is such that the coefficient K ₀ of the first pixel of interest and the coefficients K ₁ of the plurality of first pixels have a relationship of K ₀ > K ₁ .

In the above configuration, it is desirable that the first pixel is a pixel that outputs a G signal closest to human visibility among the three primary colors of RGB.

[0016]

FIG. 1 is a block diagram showing an electronic still video camera according to the present embodiment. The light incident from the photographing lens 1 is subjected to light quantity control by the optical diaphragm 2 and then formed on the CCD sensor 3. Note that the flash 22 operates according to the amount of imaging light. The CCD sensor 3 is an area sensor of a type that reads out all pixels for each horizontal line, and the same sensor as the mosaic filter shown in FIG. The minimum unit (R, G, B) of the filter is the CCD sensor 3.
Are bonded so as to correspond to one pixel. Hereinafter, R,
The pixels of the CCD sensor 3 to which the G and B filters are attached are referred to as R pixels, G pixels, and B pixels, respectively.

Exposure control data of the aperture value of the aperture 2 and the accumulation time (electronic shutter speed) of the CCD sensor 3 are as follows:
It is calculated by the camera control CPU 16 based on the light quantity data measured by the photometric sensor 21. Camera control CP
U16 is based on the calculated exposure data, and the aperture driver 1 is adjusted so that the amount of exposure to the CCD sensor 3 becomes appropriate.
4 and the timing generator 15 are controlled. The aperture driver 14 controls the aperture value of the aperture 2, and the timing generator 15 controls the accumulation time of the CCD sensor 3.

The image signal photoelectrically converted by the CCD sensor 3 is sampled by the CDS 4 to remove noise generated under the influence of the reset pulse. This image signal is
After the sensitivity is corrected by the C circuit 5, the signal is converted from an analog signal to a digital signal by the A / D converter 6, and the image is processed by the CPU 1.
3 Then, the electric charge is written to the image memory 20 in synchronization with the electric charge reading of the CCD sensor 3.
The following processing is performed by accessing the image data in the image memory 20.

The image data in the image memory 20 is transferred to a crosstalk elimination circuit 7, a pixel interpolation circuit 47, and a band correction circuit 2.
3. Image processing is performed by each circuit of the color balance control circuit 8 and the gamma correction circuit 9 and written into the image memory 20 again. Hereinafter, the processing of each circuit will be described.

The image data is first subjected to image processing by a crosstalk removing circuit 7. Here, if crosstalk is expected to generate more noise than allowed,
Filtering is performed to reduce the difference in level difference between the left and right pixel signals between adjacent horizontal lines.

FIG. 2 shows a control flowchart of the crosstalk removing circuit 7. Data of 3 × 3 pixels is input to the data input unit 24 in order with each pixel data centered on the pixel. Next, the G pixel extraction unit 25 determines whether or not the data p22 of the center pixel is G pixel data.

If the data is not G pixel data (determination 2), it can be considered that noise due to the influence of crosstalk does not occur. Therefore, the corrected data p'22 is corrected to p'22 = p22 at 29 without correcting the data of p22. Output as
In the case of the G pixel data (determination 1), the process proceeds to the modulation factor calculating unit 26. Here, signal level differences S1 and S2 of adjacent pixels are calculated. S1 = | p22-p23 |, S2 = | p
32-p33 |.

Here, S1 and S2 will be described with reference to specific examples. Let p22 be the signal level of the G pixel 91 shown in FIG. 9, for example. And the signal level is p23, p
Pixels 32 and p33 are respectively a B pixel 92 and an R pixel 9
3, G pixels 94. At this time, as shown in FIG.
Is the signal level difference between the G pixel 91 and the B pixel 92, and S2 is the signal level difference between the R pixel 93 and the G pixel 94.

Returning to FIG. 2, in the present embodiment, the means for calculating S1 is a first subtraction means, and the means for calculating S2 is a second subtraction means.
It is a subtraction means. After calculating S1 and S2, the process proceeds to the determination unit 27. Here, the crosstalk reference level REF is set in advance, and the determination is performed based on the values of S1 and S2. |
If S1−S2 | ≧ REF, determination A is made. | S1-
If S2 | <REF, determination B is made. Note that the judgment A is
Since the difference between the signal levels of the left and right adjacent pixels in the adjacent horizontal line is equal to or greater than a predetermined value (REF), it is predicted that noise exceeding an allowable amount will occur. Therefore, correction is performed by the correction unit 28 so that noise does not occur (correction filter A
To correct). On the other hand, in the case of the determination B, since it is predicted that no noise exceeding the allowable amount is generated, the correction unit 28 does not correct the data (correction is performed by the correction filter B).

The correction by the correction filter is to multiply each data of 3 × 3 pixels by a value of a corresponding pixel of the correction filter as a coefficient, and add these values. Then, the sum is used as the correction data p ′ for the central pixel.
22. That is, according to the correction filter A, p′22 = 0.125 × p11 + 0.125 × p
13 + 0.5 × p22 + 0.125 × p31 + 0.12
It becomes 5 × p33.

When such a correction is performed, for example, in the case of FIG. 9, the signal level of the G pixel 91 is changed from that before the correction to the G pixel 9
The values are close to the signal levels of 4, 95, 96, and 97.
In addition, unless the image signal is at the boundary where the color of the subject changes,
As shown in FIG. 9, the signal levels of the G pixels of two lines adjacent to one line (in the case of FIG. 9, two RG lines 1 and 2 adjacent to the GB line 1) are substantially the same (the two signal lines). This is because the effect of the crosstalk on the G pixels in one line is almost the same.) Therefore, the correction filter A
Correction reduces the difference in signal level of G pixels between adjacent lines.

When the signal level difference between G pixels between adjacent lines becomes smaller, the difference between the signal levels of adjacent G pixels becomes smaller when performing G pixel interpolation in a pixel interpolation process described later. That is, generated noise is reduced. Further, even when the pixel interpolation processing is not performed, the difference between the signal levels of the G pixels, which should be the same level originally, becomes small.
Therefore, generated noise is reduced.

According to the correction filter B, p'22 = p
22. Then, the calculated p′22 is corrected to p2
Output as 2. The above processing is performed for all pixels.

Next, the pixel interpolation circuit 47 performs a pixel interpolation process. The pixel interpolation process is a process performed to form image data (total three images) in which all the pixels are filled with each signal because the R, G, and B signals are present only at discrete intervals. By this pixel interpolation processing, three image data of an R image, a G image, and a B image are formed from one image data. FIG. 3 shows a flowchart of the processing by the pixel interpolation circuit 47.

First, after inputting image data in step # 5, G (R, B) masking processing is performed in step # 10. G (R, B) masking processing
This is a process of setting signals of pixels other than (R, B) to 0. After the masking process, an interpolation process is performed in step # 15. The interpolation process is performed for each pixel. First, for each pixel, a 3 × 3 pixel signal centered on that pixel is read. Then, the processing is performed by the interpolation filter.

The G pixel is processed by a median filter. The processing by the median filter means that the value obtained by adding the signal value of the center pixel to the average value of the intermediate binary values of the signal values of the four pixels adjacent to the center pixel of the read 3 × 3 pixel in the horizontal and vertical directions This is the process of setting the value of the central pixel later.

That is, assuming that 30 is a read signal of 3 × 3 pixels, and that the value in the pixel represents the signal value, g'22 = (g1
2, the average value of intermediate values of g21, g23, and g32)
+ G22. In step # 20, g′22 is output as the signal value of the central pixel of the 3 × 3 pixel 30 after the interpolation processing.

Interpolation of R and B pixels (step # 1)
5) is performed by the average interpolation filter 31. Since the processing is the same, only the R pixel will be described. As the processing content, a numerical value in a corresponding pixel of the average interpolation filter 31 is set as a coefficient to be applied to the signal value of each pixel, and the signal value of each pixel is multiplied by a coefficient and then added. Then, the value after the calculation is set to the value r′22 of the center pixel after the interpolation. That is, assuming that 32 is the read 3 × 3 pixel and that the value in the pixel represents the signal value, r′22 = １ ／ × r11 + ／ × r1
2 + ／ × r13 + ／ × r21 + r22 + ／
× r23 + ／ × r31 + ／ × r32 + ／ ×
r33.

The above-described interpolation processing is performed on all pixels of each image after the masking processing. Then, the three image data after interpolation are output (step # 25). Next, the band correction circuit 23 performs band correction processing on the image data. The band correction process is a process performed to correct the output response as the frequency becomes higher due to the frequency characteristics.

The principle of band correction will be described with reference to FIG.
Unless special processing is performed, the output response of the image data decreases as the frequency increases, as indicated by the curve 35 due to the frequency characteristics. In order to obtain a clear image, it is ideal that the output response is uniformly high at any frequency. The band correction process of the present embodiment is a process performed to make the output responsiveness more ideal as a result.

According to the band correction process, for example, a curve of 35 is corrected to a curve of 36. In the present embodiment, the band correction processing is performed on the G image. In the processing of the R and B images, since only low-frequency components are included as compared with the G image, the middle and high-frequency components of the G image are added to obtain clearer image data with clearer edges.

FIG. 5 is a control block diagram of the band correction circuit 23. At 37, the GL is calculated from the input G image data.
The same band limitation as that for the R and B images is performed by a low-pass filter to extract a low-frequency signal. Then, the low band signal is subtracted from the original data to extract the middle band signal. At 41, only a high-frequency signal is extracted from the original data of the G image by an enhancer filter.

The level of the mid-range signal extracted at 37 is adjusted at 38 by the level adjustment Amp, and then 39, 40, and 42.
Are added to the G, R, and B image data, respectively. Also,
The high-frequency signal extracted at 41 is level-adjusted at 43 by a level adjustment Amp, base-clipped at 44 by a CLIP circuit, and then G, 39, 45 and 46 respectively.
It is added to the R and B image data. The middle band signal and the high band signal of the G image are added to the image data of each color and output from the band correction circuit 23.

The image signal subjected to the band correction is subjected to color balance control processing by a color balance control circuit 8. The color balance control is to control a color signal of a video signal in accordance with a color temperature of imaging light so as to reproduce white as correct white. Specifically, first, in the camera control CPU 16, the R, G, and B of the entire imaging light transmitted from the colorimetric sensor 21 are read.
G / R and G / B are calculated from the above data. And
The correction gain for the R and B image data is determined. On the basis of the correction gain, the color balance control means 8 sets R,
The B image data is corrected.

After the color balance control, the gamma correction circuit 9 performs gamma correction on the image data of each color. The gamma correction circuit 9 performs non-linear conversion on each color data, and performs gradation conversion suitable for an output CRT. FIG. 6 shows the signal level conversion by the gamma correction circuit 9 in detail. When the gamma correction is not performed, the input value and the output value are in a proportional relationship as shown by a dotted line 48, but when the gamma correction is performed, the relationship is as shown by a curve 49. With this correction, the image displayed on the monitor 18 has high gradation reproducibility. The three image data after gamma correction are stored in the image memory 2
0 is stored.

When the shutter 17 is in the half-pressed state (preview state), the three image data subjected to the series of image processing described above are read out from the image memory 20 and encoded by the video encoder 11 into NTSC / PAL. Output to the built-in monitor 18. The image data is updated at a predetermined frame cycle and displayed on the monitor 18 as a moving image.

When the shutter 17 is in the pressed state (camera release state), the same processing as in the above-described half-pressed state is performed, video data is displayed on the monitor 18, and the image compression circuit 10 performs compression processing. The data is recorded on the memory card 19 by the memory card driver 12.

[0043]

According to the present invention, it is possible to suppress noise due to crosstalk occurring in a specific area without lowering the resolution of an image. Therefore, a beautiful video with high reproducibility is recorded as image data.
In addition, the images viewed in the preview mode are also beautiful with little noise.

Furthermore, since the control method is simple without having complicated calculation processing and the like, it can be configured at low cost.

[Brief description of the drawings]

FIG. 1 is a block diagram of an electronic still video camera according to an embodiment.

FIG. 2 is a control flowchart of a crosstalk removing circuit.

FIG. 3 is a flowchart of processing by a pixel interpolation circuit.

FIG. 4 is a diagram showing frequency characteristics before and after band correction of image data.

FIG. 5 is a control block diagram of a band correction circuit.

FIG. 6 is a diagram showing a relationship between input and output signal levels in a gamma correction circuit.

FIG. 7 is a diagram showing a mosaic filter adhered to a CCD.

FIG. 8 is a diagram for explaining the principle of noise generation due to crosstalk.

FIG. 9 is a diagram showing a specific example of image data for describing control by a crosstalk removing circuit.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Shooting lens 2 Aperture 3 CCD sensor 4 CDS 5 AGC circuit 6 A / D converter 7 Crosstalk removal circuit 8 Color balance control circuit 9 Gamma correction circuit 10 Image compression circuit 13 Image processing CPU 16 Camera control CPU 18 Monitor 19 Memory card Reference Signs List 20 image memory 26 modulation degree operation unit (consisting of first and second subtraction means) 27 determination unit 28 correction unit 47 band correction circuit

Claims (1)

[Claims] 1. A first line in which first and second pixels are alternately arranged in the horizontal direction for first, second, and third pixels that respectively generate signals corresponding to one of three colors; An image pickup device comprising a charge-coupled device in which first and third pixels are alternately arranged in a horizontal direction, and second lines are alternately arranged in a vertical direction and the first pixels are repeatedly arranged so as to be arranged every other pixel in a vertical direction. A first subtraction means for calculating a level difference between a first pixel of interest on a line of interest and a signal output from a pixel adjacent thereto, and a pixel on a line adjacent to the line of interest, and A second subtraction unit that calculates a level difference between signals output from two pixels corresponding to the two pixels; and a level difference between absolute values of outputs of the first and second subtraction units is equal to or more than a predetermined value. Level of the signal output from the first pixel of interest Means for correcting the signal level output from the first pixel of interest and the signal levels output from a plurality of first pixels surrounding the first pixel of interest by a predetermined coefficient. Wherein the predetermined coefficient is a coefficient K ₀ of the first pixel of interest and a coefficient K ₁ of the plurality of first pixels is in a relation of K ₀ > K ₁ .