KR20110079317A

KR20110079317A - Method of detection and correction of bad pixel

Info

Publication number: KR20110079317A
Application number: KR1020090136335A
Authority: KR
Inventors: 박동영
Original assignee: 주식회사 동부하이텍
Priority date: 2009-12-31
Filing date: 2009-12-31
Publication date: 2011-07-07

Abstract

PURPOSE: A method for detecting and correcting a bad pixel is provided to accurately detect only defected pixels. CONSTITUTION: An edge is compared at a kernel edge detection stage and an edge line detection stage. The kernel edge detection stage detects edge which is expected to be damaged and creates an edge signal(S100). The edge line detecting stage detects the bad pixel by the edge signal by grasping the linearity and distribution tendency(S105). The pattern which detects bad pixel are compared(S110). The detected bad pixel is compensated(S115).

Description

Method of detection and correction of bad pixel

Embodiments relate to a method for detecting and correcting defective pixels.

Common defective pixel processing methods are as follows. If the reference pixel is multiplied by a predetermined ratio and there is a difference of a certain threshold or ratio or more compared with the peripheral pixel, the peripheral pixel is regarded as a defective pixel, and the peripheral pixel is corrected by reflecting the threshold.

As described above, the method using the threshold difference or the ratio difference has a loss in the sharpness of the image due to the correction. In particular, when the peripheral pixels are distributed in the edge region, the average value of the peripheral pixels increases the error of detecting the defective pixels due to the influence of the edges. Is likely to be set. Therefore, even though the peripheral pixels are defective pixels, there is a problem in that the quality of the product is lowered or the yield due to the defective products is lowered.

On the other hand, if the detection criterion is raised to reduce the detection error of the defective pixels in the edge area, the average value of the peripheral pixels reflected in the detection criteria of the defective pixels is not optimized and the normal pixels are identified as defective pixels. There is a problem.

Accordingly, the defective pixel detection and correction algorithm is more complicated, but there are limitations in solving the above problems, and circuits such as a calculator and a comparator are added to increase the size of the semiconductor chip for image processing and increase the production cost.

In addition, a method of minimizing image deterioration by determining whether a peripheral pixel exists in an edge region and restricting the application of a bad pixel correction method to a peripheral pixel of an edge region has been proposed, but this also does not solve a fundamental problem. I can't.

According to an embodiment, even if a defective pixel exists in an image or an edge region including a large number of high frequency patterns, only defective pixels can be detected accurately while maintaining the characteristics of the input image to the maximum, and minimization of deterioration of the high frequency component of the input image is minimized. By providing a high sharpness and sharp images, and provides a method for detecting and correcting defective pixels that can minimize image degradation due to the correction of defective pixels.

The bad pixel detection and correction method according to the embodiment includes a kernel edge detection step of detecting an edge of a center pixel suspected as a bad pixel among kernels configured with a Bayer pattern and generating edge signals for a plurality of edge directions and the edge signal. An edge comparison step of detecting a defective pixel by detecting a linearity and a distribution tendency between the center pixel and peripheral pixels located in a series of kernel lines; A pattern comparison step of detecting defective pixels by identifying a linearity and a distribution tendency with peripheral pixels located in the peripheral area around the central pixel; And a correction step of performing correction on the detected defective pixels, wherein the Bayer pattern is formed by forming a matrix structure of a plurality of pixels having color information of red, red green, blue, and blue green.

According to the embodiment, the following effects are obtained.

First, even if there is a bad pixel in the image or edge region that contains a lot of high frequency patterns, only bad pixels can be detected accurately and accurately in order to maintain the characteristics of the input image. have.

Second, by minimizing the deterioration of the high frequency component of the input image, it is possible to maintain a sharp image with high sharpness and to minimize image degradation due to the correction of defective pixels.

Third, it is possible to solve the problem of deterioration of product quality and production yield caused by defective pixels.

Hereinafter, a method for detecting and correcting defective pixels according to an embodiment will be described in detail with reference to the accompanying drawings.

Hereinafter, in describing the embodiments, detailed descriptions of related well-known functions or configurations are deemed to unnecessarily obscure the subject matter of the present invention, and thus only the essential components directly related to the technical spirit of the present invention will be referred to. .

The defective pixel detection and correction method according to the embodiment may be applied to an image sensor, for example, and may be implemented through an external image signal processor (ISP) embedded in the image sensor chip or connected to the image sensor.

1 is a diagram illustrating a form of a Bayer pattern used in the bad pixel detection and correction method according to the embodiment.

Referring to FIG. 1, in the Bayer pattern, some pixels having color information of red (R), red green (GR), blue (B), and blue green (GB) are formed in a matrix structure.

For example, assuming that the center pixel is a red green pixel, the upper and lower pixels adjacent to the center pixel are blue pixels, the left and right pixels adjacent to the center pixel are red pixels, and the four pixels adjacent to the diagonal direction are blue green pixels. Are arranged.

In this case, the red and green pixels are arranged in the vertical and horizontal directions by one pixel.

Defective pixel detection and correction method according to the embodiment is largely divided into the method through the edge comparison and the method through the pattern comparison, these methods are used together according to the characteristics of the image.

At this time, the method through the edge comparison is preceded by the method through the pattern comparison.

2 is a flowchart illustrating a bad pixel detection and correction method according to an exemplary embodiment.

As shown in FIG. 2, if the center pixel of the current kernel is detected as a bad pixel and the center pixel is corrected according to the methods S100 and S105 through the edge comparison, the center pixel is determined by the method S110 through pattern comparison. does not apply.

This is because the defective pixels corrected through the method S100 and S105 through the edge comparison are corrected to have a similarity with the neighboring pixels, and thus no longer need to be recognized as defective pixels.

Defective pixel detection and correction method according to the embodiment is a method for solving the problem that the conventional bad pixel correction method is vulnerable to the edge or the image degradation near the edge, to minimize the sharpness loss of the image and to detect and restore the defective pixel It can be done.

In the edge comparison method, a kernel edge detection step (S100) of detecting an edge of a pixel suspected to be a bad pixel among kernels configured with the Bayer pattern, and the kernel edge information and the bad pixel outputted through the kernel edge detection step are included in a series of kernel lines. Edge line detection step S105 of determining linearity and distribution tendency with the located peripheral pixels is performed.

The kernel edge detection step S100 may be performed with respect to the vertical edge, the horizontal edge, the first diagonal edge of 45 degrees, and the second diagonal edge of 135 degrees.

In particular, in the case of the diagonal edge, a method of setting a threshold value in the diagonal pixel and determining the diagonal edge of the corresponding edge when the vertical edge and the horizontal edge occur at the same time may be applied.

The kernel edge detection step S100 includes firstly detecting an edge with respect to a defective pixel and secondly generating an edge signal for the plurality of edge directions.

When the edge signal is generated, the edge line detection step S105 is performed.

The edge line detecting step S105 analyzes a plurality of pixels, for example, three pixel distributions along the edge direction according to the edge signal, and finally determines whether the pixel detected as the defective pixel is defective.

The edge line detecting step (S105) may be performed by correcting defective pixels according to edge line detection by considering red green and blue green as the same color and correcting the red and blue green.

3 is a view illustrating a Bayer pattern to which an edge line detection step in which red green and blue green are regarded as the same color is applied, and FIG. 4 is an edge line detection step in which red green and blue green are divided according to an embodiment. It is a figure which illustrates the applied Bayer pattern.

For reference, a method of detecting / correcting a defective pixel by considering red green and blue green as the same color has a larger number of pixels than a red pixel and a blue pixel in a green pixel, so the logic size is increased by using an additional detector and a compensator. Due to the large number of pixels, accurate detection and correction are possible.

In addition, the method of detecting / correcting defective pixels by dividing red green and blue green can minimize the logic configuration by reducing the number of detectors.

The edge line detecting step S105 may be performed in three steps.

First, a first deviation of a plurality of pixels arranged on a center pixel (Py) line, for example, three pixels except for the center pixel (for example, pixels 4 and 9 of FIG. 4) according to the edge signal is obtained. The second deviation between the average value of the remaining pixels and the center pixel is obtained.

Subsequently, when the first deviation is smaller than the first threshold value and the second deviation is larger than the second threshold value, the center pixel may be determined as a defective pixel, and the correcting pixel correction step S115 may be performed. In this case, all four color channels may be recognized as independent channels, and an edge line detection step may be performed for each pixel.

5 is a diagram illustrating a Bayer pattern when the first edge line detection step is performed by the edge signal in the diagonal direction according to the embodiment.

The first edge line detection step may be performed by an edge signal in a vertical direction, or may be performed by an edge signal in a horizontal direction or an edge signal in a diagonal direction as shown in FIG. 5.

For reference, this step has the following principle.

Defective pixels can be classified into hot pixels that maintain brighter values than surrounding pixels and dark pixels that maintain darker values than surrounding pixels.If the central pixel to be corrected is defective pixels of hot pixels, Therefore, the defective pixel is corrected by the larger value of two neighboring pixels, or the defective pixel value is corrected by the average value of the two pixels.

In contrast, when the center pixel is a dark pixel, the defective pixel is corrected by using a smaller value of two neighboring pixels or the averaged pixel value is corrected by an average value of two pixels according to the characteristics of the input image. This is because the minimum pixel value is corrected to maintain similarity with the surrounding pixels even if the normal pixel is not the actual bad pixel. Therefore, even if the normal pixel of the edge region is corrected, it is naturally applied to the human visual characteristics having the characteristics of the low pass filter. This is to minimize image degradation so that it can be seen.

Second, the blue-green and red-green channels are regarded as the same channel, and according to the edge signal, a predetermined number of first peripheral pixel groups adjacent to the center pixel, for example, pixels having intermediate values among the 4, 5, and 6 pixels of FIG. A pixel having a median value among the pixels of the second peripheral pixel group, for example, 7,8, and 9 of FIG. 4 is selected.

Subsequently, a third deviation of the two selected pixels is obtained, and an average value of the two selected pixels and a fourth deviation of the center pixel are obtained.

When the third deviation is smaller than the first threshold value and the fourth deviation is larger than the second threshold value, the center pixel may be determined as a defective pixel, and the correction pixel correction step S115 may be performed.

That is, the second edge line detection step may be performed for each pixel by considering the blue green and red green channels as the same channel. The second method is applied to the green pixel, and the first method is applied to the red and blue pixels.

Third, the blue-green and red-green channels are regarded as the same channel, and according to the edge signal, the average value of the predetermined number of first peripheral pixel groups adjacent to the center pixel, for example, pixels 4, 5, and 6 of FIG. The average value of the pixel groups 7, 8 and 9 of FIG. 4 is obtained, for example.

Subsequently, a fifth deviation of the two mean values is obtained, and an average value of the two mean values and a sixth deviation of the center pixel are obtained.

When the fifth deviation is smaller than the first threshold and the sixth deviation is larger than the second threshold, the center pixel may be determined to be a defective pixel, and the correcting step of correcting the defective pixel may be performed (S115).

That is, a third edge line detection step may be performed for each pixel by considering the blue green and red green channels as the same channel. The third method is applied to the green pixel, and the first method is applied to the red and blue pixels.

6 is a diagram illustrating a Bayer pattern when the third edge line detection step is performed by the edge signal in the diagonal direction according to the embodiment.

The third edge line detection step may be performed by the edge signal in the vertical direction, or may be performed by the edge signal in the horizontal direction or the edge signal in the diagonal direction as shown in FIG. 6.

When the defective pixel is detected by the edge comparison method as described above, the defective pixel is corrected (S115).

As described above, the center pixel which is not recognized as a bad pixel through the edge comparison method S100 and S105 detects again whether or not the defective pixel is through the method S110 through pattern comparison.

The method S110 through the pattern comparison has no edges in all the pixels present in the kernel composed of the Bayer pattern, or the center pixel in which edges are detected through the methods S100 and S105 through the edge comparison but are suspected to be bad pixels. When the pixel line where is located is independent of the kernel edge data, the correlation of the center pixel is compared for all cases.

7 to 9 illustrate a bad pixel detection method through pattern comparison according to an embodiment.

In the method using the pattern comparison, first, a peripheral pixel located in the peripheral region of the central pixel of the kernel is selected. In this case, when the center pixel Pc is a red pixel or a blue pixel, peripheral pixels spaced apart from each other by one pixel vertically and horizontally selected from the center pixel are selected as shown in FIG. 7.

In addition, when the central pixel is a green pixel, the pixel may be selected as shown in FIG. 7, the peripheral pixels surrounding the central pixel as shown in FIG. 8, or the peripheral pixels sequentially spaced one pixel from the central pixel as shown in FIG. 9. have.

Second, the deviation of the center pixel with respect to the distribution of the selected peripheral pixel is determined to determine whether the center pixel is defective. This may proceed as follows.

(1) The peripheral pixels are arranged in the order of difference by comparing the selected peripheral pixels with the central pixels, and the peripheral pixels having the largest difference and the peripheral pixels having the smallest difference are selected.

(2) From the neighboring pixels of the maximum / minimum difference, select the neighboring pixels of the intermediate difference except for the predetermined number, respectively, and obtain the average value of the neighboring pixels of the current kernel.

(3) The difference between the brightness value of the peripheral pixel average value and the human discernible brightness value is obtained, and the deviation between the center pixel and the peripheral pixel is calculated. The threshold value of the pattern defective pixel is defined by setting the difference between the brightness value and the minimum range of the deviation.

(4) If the deviation between the selected peripheral pixel and the center pixel is larger than the threshold, the center pixel is detected as a defective pixel, and the center pixel is corrected using a correction value corresponding to the threshold (S115).

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, It will be understood that various modifications and applications other than those described above are possible. For example, each component specifically shown in the embodiment of the present invention can be modified. And differences relating to such modifications and applications will have to be construed as being included in the scope of the invention defined in the appended claims.

2 is a flowchart illustrating a bad pixel detection and correction method according to an embodiment;

3 is a diagram illustrating a Bayer pattern to which an edge line detection step in which red and blue green are regarded as the same color according to an embodiment is applied.

4 is a diagram illustrating a Bayer pattern to which an edge line detection step of dividing red green and blue green according to an embodiment is applied.

6 is a diagram illustrating a Bayer pattern when the third edge line detection step according to the embodiment is performed by the edge signal in the diagonal direction.

Claims

Kernel edge detection step of detecting edges and generating edge signals for a plurality of edge directions among the kernels composed of the Bayer pattern and located at the center pixel and a series of kernel lines according to the edge signals. An edge comparison step of detecting a defective pixel by identifying a linearity and a distribution tendency with the neighboring pixels;

A pattern comparison step of detecting defective pixels by identifying a linearity and a distribution tendency with peripheral pixels located in the peripheral area around the central pixel; And

A correction step of performing correction on the detected defective pixels;

And the Bayer pattern is formed by forming a matrix structure of a few pixels having color information of red, red green, blue, and blue green.

The method of claim 1,

The edge comparison step is preceded by the pattern comparison step, and when the center pixel of the kernel is detected as a bad pixel in the edge comparison step, the pattern comparison step does not proceed, and when the bad pixel is not detected in the edge comparison step, the pattern Defective pixel detection and correction method characterized in that the comparison step is carried out.

The method of claim 1, wherein the kernel edge detection step is

Defective pixel detection and correction method, characterized in that the vertical edge, the horizontal edge, the first diagonal edge of 45 degrees, the second diagonal edge of 135 degrees.

4. The method of claim 3, wherein in the kernel edge detection step

In the case of diagonal edges, a threshold is set for diagonal pixels, and when a vertical edge and a horizontal edge occur at the same time, a diagonal edge of the corresponding edge is detected.

The method of claim 1, wherein the Bayer pattern

Assuming that the center pixel is a red green pixel, the upper and lower pixels adjacent to the center pixel are blue pixels, the left and right pixels adjacent to the center pixel are red pixels, and the four pixels adjacent to the diagonal direction are arranged in blue green pixels. And the red-green pixels are arranged in the vertical and horizontal directions spaced one pixel apart.

The method of claim 1, wherein the detecting edge line is

Defective pixel detection and correction method characterized in that it detects defective pixels according to the edge line by considering red green and blue green as the same color, or detect defective pixels along the edge line by separating red green and blue green.

The method of claim 1, wherein the detecting edge line is

Obtaining a first deviation of the remaining pixels excluding the center pixel from the plurality of pixels aligned on the center pixel line according to the edge signal, and obtaining an average value of the remaining pixels and a second deviation of the center pixel;

And determining the central pixel as a defective pixel when the first deviation is smaller than the first threshold value and the second deviation is larger than the second threshold value.

The method of claim 7, wherein the detecting edge line

And detecting each of the four color channels as independent channels and proceeding with respect to each pixel.

The method of claim 1, wherein the detecting edge line is

Considering the blue-green and red-green channels as the same channel, selecting pixels having a median value among a predetermined number of first peripheral pixel groups adjacent to the center pixel according to the edge signal, and having a median value of the second peripheral pixel group according to the edge signal. Selecting a pixel;

Obtaining a third deviation of the selected two pixels and obtaining a fourth deviation of the average value of the two selected pixels and the center pixel; And

And determining the central pixel as a defective pixel when the third deviation is smaller than the first threshold and the fourth deviation is larger than the second threshold.

The method of claim 1, wherein the detecting edge line is

Considering the blue-green and red-green channels as the same channel, obtaining an average value of a predetermined number of first peripheral pixel groups adjacent to the center pixel according to the edge signal, and obtaining an average value of the second peripheral pixel group;

Obtaining a fifth deviation of the two mean values, and obtaining a sixth deviation of the mean value of the two mean values and the center pixel; And

And determining a center pixel as a defective pixel when the fifth deviation is smaller than the first threshold and the sixth deviation is larger than the second threshold.

11. The method according to any one of claims 7, 9 and 10, wherein the detecting edge line is

And the edge signal in the vertical direction, the edge signal in the horizontal direction, and the edge signal in the diagonal direction, respectively.

The method of claim 1, wherein the pattern comparison step

Selecting a plurality of peripheral pixels located in the peripheral region of the central pixel of the kernel;

And determining whether the center pixel is defective by determining a deviation of the center pixel with respect to the distribution of the selected neighboring pixels.

The method of claim 12, wherein the central pixel is

And a peripheral pixel spaced apart from the center pixel by one pixel in the case of a red pixel or a blue pixel.

The method of claim 12, wherein the central pixel is

In the case of green pixels, peripheral pixels spaced vertically from the center pixel are selected, peripheral pixels surrounding the center pixel are selected, or peripheral pixels spaced one pixel sequentially from the center pixel. Defective pixel detection and correction method characterized in that.

The method of claim 12, wherein determining whether the central pixel is defective

Comparing the selected neighboring pixels with each of the center pixels to list neighboring pixels in a difference order, and selecting neighboring pixels having the largest difference and neighboring pixels having the smallest difference;

Selecting a peripheral pixel group having an intermediate difference except a predetermined number from neighboring pixels of the maximum difference and neighboring pixels of the minimum difference, respectively, to obtain an average value of the peripheral pixels of the current kernel;

Obtaining a difference between the brightness value of the average value of the peripheral pixels and the distinguishable brightness value of the human, and calculating a deviation between the center pixel and the peripheral pixels; And

Determining a threshold value of a pattern defective pixel by setting a difference between the brightness value and the minimum range of the deviation, and detecting the center pixel as the defective pixel when the selected peripheral pixel and the center pixel have a predetermined range greater than the threshold value. Defective pixel detection and correction method, characterized in that.