Image bad pixel removal method, device and correction system based on normal distribution

Abstract

The present invention discloses a method, device and correction system for removing bad pixels from an image based on normal distribution, comprising: performing RGB channel separation on an image to be processed to obtain R, G, and B channel images; performing bad pixel removal processing on each channel image; the bad pixel removal processing comprises: using the pixel values ​​of each pixel point in the channel image after removing the background pixel points, calculating the pixel average μ and the pixel standard deviation σ of the channel image; based on μ and σ, screening out pixel points suspected to be bad pixels as candidate points to be detected; judging whether the candidate point to be detected is a bad pixel based on the pixel value of the candidate point to be detected and the pixel value of the pixel point within the preset neighborhood of the candidate point to be detected; if the candidate point to be detected is a bad pixel, correcting the pixel values ​​of the candidate point to be detected and the pixel points within the preset neighborhood of the candidate point to be detected. The method, device and correction system provided by the present invention can effectively improve the bad pixel removal efficiency and bad pixel recognition accuracy.

Description

Technical Field

[0001] The present invention relates to the field of LED display screen correction, and in particular to a method, device and correction system for removing bad pixels of an image based on normal distribution. Background Art

[0002] During the LED display screen calibration process, the host computer needs to use a camera to capture the light point image as the calibration image, so as to calculate the correction coefficient through the calibration image.

[0003] However, due to reasons such as wear and tear, some photosensitive devices in the camera imaging module may have problems, resulting in abnormal color pixels in the captured light point image (that is, due to damage to the photosensitive device in the camera, there are bad pixels in the light point image captured by the camera). The bad pixels sometimes appear in the center of the light point image, for example Figure 1 The lamp point image shown in is a lamp point image with a bad point (where P and Q indicate the position of the bad point). If the collected lamp point image is abnormal, the subsequent calculation of the correction coefficient using the lamp point image will cause the calculated correction coefficient to be inaccurate, seriously affecting the correction effect of the LED display.

[0004] Some existing methods for removing bad pixels have many problems and cannot effectively remove bad pixels in the corrected image. For example, the amount of calculation required to remove bad pixels is too high, resulting in low bad pixel removal efficiency. In addition, the accuracy of bad pixel identification is not high, which affects the correction effect of the LED display and the display effect of the LED display. Summary of the Invention

[0005] To address the above issues, the present invention provides a method, device, and correction system for removing bad pixels from an image based on normal distribution. The method includes:

[0006] Separate the RGB channels of the image to be processed to obtain R, G, and B channel images;

[0007] For each channel image, bad pixel removal processing is performed; the bad pixel removal processing includes:

[0008] The pixel value of each pixel in the channel image after removing the background pixels is used to calculate the pixel average μ and pixel standard deviation σ of the channel image;

[0009] According to μ and σ, the pixels suspected to be bad pixels are selected as candidate points to be detected;

[0010] Determine whether the candidate point to be detected is a bad pixel based on the pixel value of the candidate point to be detected and the pixel values ​​of the pixels within the preset neighborhood of the candidate point to be detected;

[0011] If the candidate point to be detected is a bad pixel, the pixel values ​​of the candidate point to be detected and the pixels in the preset neighborhood of the candidate point to be detected are corrected to remove the bad pixel.

[0012] Furthermore, the bad pixel removal process further includes identifying background pixels in the channel image by triangulation, wherein identifying background pixels in the channel image by triangulation includes:

[0013] Count the pixel distribution of the channel image to obtain a pixel distribution graph, where the horizontal axis is the pixel value and the vertical axis is the number of pixels;

[0014] Determine the highest point and the lowest point in the pixel distribution diagram, and connect the highest point and the lowest point to form a straight line;

[0015] The pixel value corresponding to the point with the longest vertical distance from the straight line in the pixel distribution diagram is used as the pixel threshold;

[0016] Pixels with values ​​less than the pixel threshold are determined as background pixels.

[0017] Furthermore, based on μ and σ, the pixels suspected to be bad pixels are screened out as candidate points to be detected, including:

[0018] According to the preset pixel selection order, one pixel is selected from the channel image in sequence as the target pixel, and a screening operation is performed on the target pixel until the screening operation is completed on all the pixels in the channel image, so as to screen out the pixel points suspected to be bad pixels as candidate points to be detected;

[0019] The screening operation includes: for a target pixel point, determining whether the pixel value of the target pixel point in the channel image is greater than μ+σ;

[0020] Determine whether the target pixel point is the highest local pixel value point within a preset neighborhood range in the channel image;

[0021] If the pixel value of the target pixel in the channel image is greater than μ+σ, and the target pixel is the highest point in the local pixel value within the preset neighborhood in the channel image, the target pixel is taken as a candidate point to be detected.

[0022] Furthermore, judging whether the candidate point to be detected is a bad pixel according to the pixel value of the candidate point to be detected and the pixel values ​​of the pixels within a preset neighborhood range of the candidate point to be detected includes:

[0023] Calculate the pixel average μ1 and pixel standard deviation σ1 of the pixel points in the preset neighborhood range of the candidate point to be detected in the channel image;

[0024] Get the color of the image to be processed and the color of the channel image, and determine whether the channel image is the main channel image based on the color of the image to be processed and the color of the channel image, where the color of the R channel image is red, the color of the G channel image is green, and the color of the B channel image is blue;

[0025] If the channel image is the main channel image, determine whether the pixel value of the candidate point to be detected in the channel image is greater than μ1+(3*σ1). If so, determine that the candidate point to be detected is a bad pixel.

[0026] Furthermore, judging whether the candidate point to be detected is a bad pixel according to the pixel value of the candidate point to be detected and the pixel values ​​of the pixels within the preset neighborhood of the candidate point to be detected further includes:

[0027] If the channel image is an auxiliary channel image, obtain the pixel value μ2 of the candidate point to be detected in the main channel image

[0028] Calculate the pixel standard deviation σ2 of the pixels in the preset neighborhood of the candidate point to be detected in the main channel image;

[0029] Determine whether the pixel value of the candidate point to be detected in the channel image is greater than μ1+(3*σ1) and greater than μ2-σ2. If so, the candidate point to be detected is determined to be a bad pixel.

[0030] Furthermore, the preset neighborhood range is an 8-neighborhood centered on the candidate point to be detected.

[0031] Furthermore, correcting the pixel values ​​of the candidate point to be detected and the pixel values ​​of the pixel points within the preset neighborhood of the candidate point to be detected includes:

[0032] Obtain pixel values ​​of correction pixels corresponding to the candidate point to be detected in the channel image, where the correction pixels are peripheral pixels within a preset neighborhood of the candidate point to be detected;

[0033] Calculate the pixel average μ3 of the corrected pixel in the channel image;

[0034] The pixel values ​​of the candidate points to be detected in the channel image and the pixel values ​​of the pixels within the preset neighborhood of the candidate points to be detected in the channel image are adjusted to μ3.

[0035] The present invention also provides an image bad pixel removal device based on normal distribution, the device comprising an image separation module and a bad pixel processing module, wherein:

[0036] The image separation module is connected to the bad pixel processing module and is used to separate the RGB channels of the image to be processed to obtain R, G, and B channel images;

[0037] The bad pixel processing module is used to perform bad pixel removal processing on each channel image. The bad pixel processing module includes a candidate point screening unit to be detected, a bad pixel judgment unit, and a bad pixel correction unit. The candidate point screening unit to be detected is connected to the bad pixel judgment unit and is used to calculate the pixel average μ and pixel standard deviation σ of the channel image using the pixel value of each pixel in the channel image after removing the background pixels. Based on μ and σ, the pixel points suspected to be bad pixels are screened as candidate points to be detected.

[0038] a bad pixel judging unit connected to the bad pixel correcting unit, for judging whether the candidate point to be detected is a bad pixel based on the pixel value of the candidate point to be detected and the pixel values ​​of pixels within a preset neighborhood of the candidate point to be detected;

[0039] The bad pixel correction unit is used to correct the pixel values ​​of the candidate point to be detected and the pixel values ​​of the pixel points in the preset neighborhood of the candidate point to be detected to remove the bad pixel when the candidate point to be detected is determined to be a bad pixel.

[0040] Furthermore, the candidate point screening unit to be detected screens out pixels suspected to be bad pixels as candidate points to be detected based on μ and σ, including:

[0041] According to the preset pixel selection order, one pixel is selected from the channel image in sequence as the target pixel, and a screening operation is performed on the target pixel until the screening operation is completed on all the pixels in the channel image, so as to screen out the pixel points suspected to be bad pixels as candidate points to be detected;

[0042] The screening operations include:

[0043] For the target pixel point, determine whether the pixel value of the target pixel point in the channel image is greater than μ+σ;

[0044] Determine whether the target pixel point is the highest local pixel value point within a preset neighborhood range in the channel image;

[0045] If the pixel value of the target pixel in the channel image is greater than μ+σ, and the target pixel is the highest point in the local pixel value within the preset neighborhood in the channel image, the target pixel is taken as a candidate point to be detected.

[0046] A calibration system includes a camera, a host computer, a sending card, and an LED display, wherein:

[0047] The camera is connected to the host computer to capture the calibration image of the LED display and send the calibration image to the host computer;

[0048] The host computer is connected to the sending card and is used to remove bad pixels in the corrected image according to the above-mentioned image bad pixel removal method based on normal distribution, and calculate the correction coefficient of each light point in the LED display screen based on the corrected image after bad pixel removal, and send the correction coefficient to the sending card;

[0049] The sending card is connected to the receiving card of the LED display screen to send the correction coefficient to the receiving card of the LED display screen;

[0050] The LED display screen is used to correct display data according to the correction coefficient of each light point received through the receiving card, and display the corrected display data.

[0051] The image bad pixel removal method, device and correction system based on normal distribution provided by the present invention have at least the following beneficial effects:

[0052] (1) Before bad pixel recognition, the pixel mean μ and pixel standard deviation σ of the channel image are used to screen out suspected bad pixels. Then, the suspected bad pixels are judged one by one to identify whether they are bad pixels. Therefore, there is no need to identify bad pixels for all pixels in the channel image, which greatly reduces the amount of calculation and improves the efficiency of bad pixel recognition in the channel image.

[0053] (2) Before calculating the pixel average μ and pixel standard deviation σ of the channel image, the background pixels in the channel image are first identified, and the pixel average μ and pixel standard deviation σ of the channel image are calculated using the pixel values ​​of each pixel in the channel image after removing the background pixels, thereby reducing or even eliminating the influence of the background pixels on the pixel average and pixel standard deviation of the channel image, improving the accuracy of identifying suspected bad pixels, and thus improving the accuracy of bad pixel identification;

[0054] (3) After a suspected bad point is determined as a candidate point to be detected, whether the candidate point to be detected is a bad point is determined based on the pixel value of the candidate point to be detected and the pixel values ​​of the pixels within the preset neighborhood of the candidate point to be detected. That is, for each candidate point to be detected, the data used to determine the bad point are specific to the candidate point to be detected, thereby further improving the accuracy of bad point recognition;

[0055] (4) After the bad pixel is identified, since the bad pixel also affects the pixel values ​​of the surrounding pixels, in this embodiment, not only the pixel value of the bad pixel is corrected, but also the pixel values ​​of the pixels within the preset neighborhood of the bad pixel are corrected, thereby eliminating the influence of the bad pixel on the corrected image to the greatest extent, which is beneficial to improving the correction effect of the display screen and further improving the display effect of the display screen. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0057] Figure 1 This is a schematic diagram of a lamp image with a bad pixel;

[0058] Figure 2 This is a flowchart of a method for removing bad pixels from an image based on normal distribution in one embodiment of the present invention;

[0059] Figure 3 for Figure 2 Flowchart of the method for step S102;

[0060] Figure 4 Schematic diagram of pixel distribution of a channel image;

[0061] Figure 5 for Figure 3 Flowchart of the screening operation method of step S203;

[0062] Figure 6 Schematic diagram of the candidate point to be detected and the pixels within the 8-neighborhood range of the candidate point to be detected;

[0063] Figure 7 for Figure 3 Method flow chart of step S204;

[0064] Figure 8 for Figure 3 Method flow chart of step S205;

[0065] Figure 9 A schematic diagram of the peripheral pixels of the pixel point within the 8-neighborhood range centered on the candidate point to be detected;

[0066] Figure 10 Schematic diagram of the structure of an image bad pixel removal device based on normal distribution in one embodiment of the present invention;

[0067] Figure 11 Schematic diagram of the correction system structure in one embodiment of the present invention;

[0068] 1-image separation module, 2-bad pixel processing module, 201-background pixel recognition unit, 202-candidate point screening unit to be detected, 203-bad pixel judgment unit, 204-bad pixel correction unit, 3-camera, 4-host computer, 5-sending card, 6-LED display. DETAILED DESCRIPTION

[0069] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0070] In one embodiment of the present invention, Figure 2 As shown, a method for removing bad pixels in an image based on normal distribution is disclosed, wherein the bad pixels in the image referred to in the present invention refer to bad pixels in the captured light point image caused by problems with individual photosensitive devices in the camera imaging module.

[0071] The method comprises the following steps:

[0072] Step S101: performing RGB channel separation on the image to be processed to obtain R, G, and B channel images.

[0073] The image to be processed is a light point image captured by a camera after the LED display screen is lit. Preferably, in this embodiment, the camera is a Canon camera, that is, the image to be processed is captured by a Canon camera.

[0074] When calibrating an LED display, the screen color of the LED display will first be adjusted to red, green, and blue, and the corresponding light point images will be captured by the camera as the images to be processed. The screen color of the LED display can be controlled by the host computer software.

[0075] The image to be processed captured by the camera has three channels: R, G, and B. When the screen color is red, the image to be processed is red, with the R channel as the primary channel and the G and B channels as the secondary channels. Similarly, when the screen color is green, the image to be processed is green, with the G channel as the primary channel and the R and B channels as the secondary channels. When the screen color is blue, the image to be processed is blue, with the B channel as the primary channel and the R and G channels as the secondary channels.

[0076] In this step, the image to be processed is separated into RGB channels to obtain R, G, and B channel images. Subsequently, bad pixel removal processing is performed on each channel image.

[0077] Step S102: performing bad pixel removal processing on each channel image.

[0078] Further, such as Figure 3 As shown in FIG, for a channel image, bad pixel removal processing includes the following steps:

[0079] Step S201: Identify background pixels in the channel image.

[0080] Among them, the background pixels in the channel image can be identified by Otsu's method or triangulation method. Specifically, the pixel distribution of the channel image is counted to obtain a pixel distribution diagram (such as Figure 4 As shown, the horizontal axis represents the pixel value, and the vertical axis represents the number of pixels), a pixel threshold is determined by Otsu's method or triangulation method, and pixels with pixel values ​​less than the pixel threshold are determined as background pixels.

[0081] Preferably, in this step, triangulation is used to identify background pixels in the channel image, and identifying background pixels in the channel image by triangulation includes:

[0082] Count the pixel distribution of the channel image and get the pixel distribution diagram, such as Figure 4 As shown in the figure, the horizontal axis of the pixel distribution diagram is the pixel value, and the vertical axis is the number of pixels;

[0083] Determine the highest point and the lowest point in the pixel distribution diagram, and connect the highest point and the lowest point to form a straight line;

[0084] The pixel value corresponding to the point with the longest vertical distance from the straight line in the pixel distribution diagram is used as the pixel threshold;

[0085] Pixels with values ​​less than the pixel threshold are determined as background pixels.

[0086] Step S202: Calculating the pixel average μ and pixel standard deviation σ of the channel image using the pixel value of each pixel in the channel image after removing the background pixels;

[0087] Step S203: Filter out pixels suspected to be bad pixels as candidate points to be detected based on μ and σ.

[0088] In this step, the pixel average μ and pixel standard deviation σ of the channel image are used to screen out pixels suspected to be bad pixels from all the pixels in the channel image as pixels to be detected.

[0089] Step S204: determining whether the candidate point to be detected is a bad pixel based on the pixel value of the candidate point to be detected and the pixel values ​​of the pixels within a preset neighborhood of the candidate point to be detected;

[0090] Step S205: If the candidate point to be detected is a bad pixel, the pixel values ​​of the candidate point to be detected and the pixels in the preset neighborhood of the candidate point to be detected are corrected to remove the bad pixel.

[0091] It should be understood that steps S201 to S205 are performed on one channel image. When each of the R, G, and B channel images is processed using steps S201 to S205, bad pixel removal of the image to be processed can be completed.

[0092] The image bad pixel removal method based on normal distribution provided in this embodiment first uses the pixel average value μ and pixel standard deviation σ of the channel image to screen out suspected bad pixels before bad pixel identification, and then judges the suspected bad pixels one by one to identify whether they are bad pixels. Therefore, there is no need to identify bad pixels for all pixels in the channel image, which greatly reduces the amount of calculation and improves the efficiency of bad pixel identification in the channel image. At the same time, before calculating the pixel average value μ and pixel standard deviation σ of the channel image, the background pixels in the channel image are first identified, and the pixel average value μ and pixel standard deviation σ of the channel image are calculated using the pixel values ​​of each pixel in the channel image after removing the background pixels, thereby reducing or even eliminating the influence of the background pixels on the pixel average value and pixel standard deviation of the channel image, thereby improving the identification of suspected bad pixels. The accuracy of bad pixel recognition is improved; in addition, after determining the suspected bad pixel and taking it as the candidate point to be detected, the candidate point to be detected is judged to be a bad pixel according to the pixel value of the candidate point to be detected and the pixel value of the pixel points within the preset neighborhood of the candidate point to be detected. That is, for each candidate point to be detected, the data used to judge the bad pixel are specific to the candidate point to be detected, thereby further improving the accuracy of bad pixel recognition; finally, after the bad pixel is identified, since the bad pixel also affects the pixel value of the surrounding pixels, in this embodiment, not only the pixel value of the bad pixel is corrected, but also the pixel value of the pixel points within the preset neighborhood of the bad pixel is corrected, thereby eliminating the influence of the bad pixel on the corrected image to the greatest extent, which is beneficial to improving the correction effect of the display screen, and thus improving the display effect of the display screen.

[0093] In another embodiment of the present invention, step S203 is: according to a preset pixel selection order, a pixel point is selected from the channel image in turn as a target pixel point, and a screening operation is performed on the target pixel point until the screening operation is completed on all pixel points in the channel image, so as to screen out pixel points suspected to be bad pixels as candidate points to be detected.

[0094] The pixel selection order is preset by technicians, and the present invention does not impose any limitation on this.

[0095] like Figure 5 As shown, for the target pixel, the screening operation includes the following steps:

[0096] Step S2031: determining whether the pixel value of the target pixel in the channel image is greater than μ+σ;

[0097] Step S2032: determining whether the target pixel point in the channel image is the highest local pixel value point within a preset neighborhood range;

[0098] Among them, Figure 6 As shown, the preset neighborhood range can be an 8-neighborhood range centered on the target pixel point P, that is, the pixel points within the preset neighborhood range of the target pixel point P are Figure 6 The 8 pixels at the position marked as A in .

[0099] The pixel value of the target pixel is the highest local pixel value, that is, the pixel value of the target pixel is greater than the pixel values ​​of the eight pixels in the eight-neighborhood range centered on the target pixel.

[0100] Since bad pixels are usually located in the middle of the light point image and rarely appear at the edge, if a pixel is located at the very edge of the light point image and there are no 8-neighborhood pixels centered on it, it can be directly considered that the pixel is not a suspected bad pixel.

[0101] Step S2033: If the pixel value of the target pixel point in the channel image is greater than μ+σ, and the target pixel point is the highest point in the local pixel value within the preset neighborhood range in the channel image, the target pixel point is used as a candidate point to be detected.

[0102] In another embodiment of the present invention, Figure 7 As shown, step S204 includes the following steps:

[0103] Step S2041: Calculate the pixel average μ1 and pixel standard deviation σ1 of the pixel points within the preset neighborhood range of the candidate point to be detected in the channel image.

[0104] Similarly, if the preset neighborhood range is an 8-neighborhood centered on the candidate point to be detected, that is, the pixel values ​​of the 8 pixels corresponding to the 8-neighborhood in the channel image are used to calculate the pixel average μ1 and the pixel standard deviation σ1.

[0105] Step S2042: Determine whether the channel image is a main channel image. If so, execute step S2043; otherwise, execute steps S2044-S2045.

[0106] Specifically, this step requires obtaining the color of the image to be processed and the color of the channel image, and determining whether the channel image is the main channel image based on the color of the image to be processed and the color of the channel image. Among them, the color of the R channel image is red, the color of the G channel image is green, and the color of the B channel image is blue.

[0107] If the color of the image to be processed is red, then the R channel image corresponding to the image to be processed is the main channel image, and the same applies to the other channels. Therefore, the color of the image to be processed can be used to determine whether the channel image being processed is the main channel image.

[0108] Step S2043: Determine whether the pixel value of the candidate point to be detected in the channel image is greater than μ1+(3*σ1). If so, determine that the candidate point to be detected is a bad pixel.

[0109] Step S2044: Obtain the pixel value μ2 of the candidate point to be detected in the main channel image, and calculate the pixel standard deviation σ2 of the pixel points within the preset neighborhood range of the candidate point to be detected in the main channel image.

[0110] Similarly, if the preset neighborhood range is an 8-neighborhood centered on the candidate point to be detected, that is, the pixel standard deviation σ2 is calculated using the pixel values ​​of the 8 pixels corresponding to the 8-neighborhood in the main channel image.

[0111] Specifically, in one implementation of the present invention, bad pixel removal processing is performed on the R, G, and B channel images simultaneously. Therefore, the main channel image used in this step is the original main channel image without correcting the bad pixel pixel values.

[0112] Step S2045: Determine whether the pixel value of the candidate point to be detected in the channel image is greater than μ1+(3*σ1) and greater than μ2-σ2. If so, determine that the candidate point to be detected is a bad pixel.

[0113] When the currently processed image is an auxiliary channel image, the above step S2044 is performed, and it is determined whether the pixel value of the candidate point to be detected in the channel image is greater than μ2-σ2, in order to determine whether the auxiliary channel value is close to the main channel value.

[0114] In another embodiment of the present invention, Figure 8 As shown, step S205 includes the following steps:

[0115] Step S2051: Obtain pixel values ​​of correction pixels corresponding to the candidate point to be detected in the channel image, where the correction pixels are peripheral pixels within a preset neighborhood of the candidate point to be detected.

[0116] like Figure 9 As shown, the preset neighborhood range is: the 8-neighborhood range with the candidate point P to be detected as the center is taken as an example, then the peripheral range pixels of the pixel point A in the 8-neighborhood range refer to the 16 pixels adjacent to the pixel point in the 8-neighborhood range and located in the peripheral range of the pixel point in the 8-neighborhood range, that is, Figure 9 The 16 pixels marked as B in .

[0117] Step S2052: Calculate the pixel average μ3 of the correction pixel in the channel image;

[0118] Step S2053: adjusting the pixel value of the candidate point to be detected in the channel image and the pixel value of the pixel points within the preset neighborhood range of the candidate point to be detected in the channel image to μ3.

[0119] like Figure 10 As shown, the present invention also provides an image bad pixel removal device based on normal distribution, the device includes an image separation module 1 and a bad pixel processing module 2, wherein:

[0120] The image separation module 1 is connected to the bad pixel processing module 2 and is used to separate the RGB channels of the image to be processed to obtain R, G, and B channel images;

[0121] The bad pixel processing module 2 is used to perform bad pixel removal processing on each channel image.

[0122] The bad pixel processing module 2 includes a candidate pixel screening unit 202, a bad pixel judging unit 203 and a bad pixel correcting unit 204;

[0123] The candidate point screening unit 202 is connected to the bad pixel judgment unit 203 and is used to calculate the pixel average μ and pixel standard deviation σ of the channel image using the pixel values ​​of each pixel in the channel image after removing the background pixels; based on μ and σ, the pixels suspected to be bad pixels are screened as candidate points to be detected;

[0124] A bad pixel judging unit 203 is connected to the bad pixel correcting unit 204 and is used to judge whether the candidate point to be detected is a bad pixel based on the pixel value of the candidate point to be detected and the pixel values ​​of the pixels within the preset neighborhood range of the candidate point to be detected;

[0125] The bad pixel correction unit 204 is configured to correct the pixel values ​​of the candidate point to be detected and the pixel values ​​of the pixel points in the preset neighborhood of the candidate point to be detected, so as to remove the bad pixel when the candidate point to be detected is determined to be a bad pixel.

[0126] In another embodiment of the present invention, the bad pixel processing module 2 also includes a background pixel identification unit 201, which is connected to the candidate point screening unit 202 to be detected, and is used to identify background pixel points in the channel image, specifically, to identify background pixel points in the channel image through Otsu's method or triangulation method.

[0127] In another embodiment of the present invention, the candidate point screening unit 202 screens out pixel points suspected to be bad pixels as candidate points to be detected based on μ and σ, including:

[0128] According to the preset pixel selection order, one pixel is selected from the channel image in sequence as the target pixel, and a screening operation is performed on the target pixel until the screening operation is completed on all the pixels in the channel image, so as to screen out the pixel points suspected to be bad pixels as candidate points to be detected;

[0129] The screening operations include:

[0130] For the target pixel point, determine whether the pixel value of the target pixel point in the channel image is greater than μ+σ;

[0131] Determine whether the target pixel point is the highest local pixel value point within a preset neighborhood range in the channel image;

[0132] If the pixel value of the target pixel in the channel image is greater than μ+σ, and the target pixel is the highest point in the local pixel value within the preset neighborhood in the channel image, the target pixel is taken as a candidate point to be detected.

[0133] In another embodiment of the present invention, the bad pixel determination unit 203 determines whether the candidate point to be detected is a bad pixel according to the pixel value of the candidate point to be detected and the pixel values ​​of the pixels within the preset neighborhood of the candidate point to be detected, including:

[0134] Calculate the pixel average μ1 and pixel standard deviation σ1 of the pixel points in the preset neighborhood range of the candidate point to be detected in the channel image;

[0135] Get the color of the image to be processed and the color of the channel image, and determine whether the channel image is the main channel image based on the color of the image to be processed and the color of the channel image, where the color of the R channel image is red, the color of the G channel image is green, and the color of the B channel image is blue;

[0136] If the channel image is the main channel image, determine whether the pixel value of the candidate point to be detected in the channel image is greater than μ1+(3*σ1). If so, determine that the candidate point to be detected is a bad pixel.

[0137] Furthermore, if the channel image is an auxiliary channel image, obtain the pixel value μ2 of the candidate point to be detected in the main channel image, calculate the pixel standard deviation σ2 of the pixel points within the preset neighborhood range of the candidate point to be detected in the main channel image, and judge whether the pixel value of the candidate point to be detected in the channel image is greater than μ1+(3*σ1) and greater than μ2-σ2. If so, determine that the candidate point to be detected is a bad pixel.

[0138] In one embodiment of the present invention, Figure 11 As shown, a calibration system is also provided, which includes a camera 3, a host computer 4, a sending card 5 and an LED display 6, wherein:

[0139] The camera 3 is connected to the host computer 4 and is used to capture the calibrated image of the LED display screen 6 and send the calibrated image to the host computer 4;

[0140] The host computer 4 is connected to the sending card 5 and is used to remove bad pixels in the corrected image according to the above-mentioned image bad pixel removal method based on normal distribution, and calculate the correction coefficient of each light point in the LED display screen 6 based on the corrected image after removing the bad pixels, and send the correction coefficient to the sending card 5;

[0141] The sending card 5 is connected to the receiving card of the LED display screen 6 and is used to send the correction coefficient to the receiving card of the LED display screen 6;

[0142] The LED display screen 6 is used to correct the display data according to the correction coefficient of each light point received through the receiving card, and display the corrected display data.

[0143] The present invention provides an image bad pixel removal method, device and correction system based on normal distribution. Before bad pixel identification, the pixel average value μ and pixel standard deviation σ of the channel image are first used to screen out suspected bad pixels, and then the suspected bad pixels are judged one by one to identify whether they are bad pixels. Therefore, there is no need to identify bad pixels for all pixels in the channel image, which greatly reduces the amount of calculation and improves the efficiency of bad pixel identification in the channel image. At the same time, before calculating the pixel average value μ and pixel standard deviation σ of the channel image, the background pixels in the channel image are first identified, and the pixel average value μ and pixel standard deviation σ of the channel image are calculated using the pixel values ​​of each pixel in the channel image after removing the background pixels, thereby reducing or even eliminating the influence of the background pixels on the pixel average value and pixel standard deviation of the channel image, and improving the efficiency of bad pixel identification in the channel image. The accuracy of bad point recognition is improved, thereby improving the accuracy of bad point recognition; in addition, after determining the suspected bad point and taking it as the candidate point to be detected, the candidate point to be detected is judged whether it is a bad point according to the pixel value of the candidate point to be detected and the pixel value of the pixel points within the preset neighborhood range of the candidate point to be detected. That is, for each candidate point to be detected, the data used to judge the bad point are specific to the candidate point to be detected, thereby further improving the accuracy of bad point recognition; finally, after the bad point is identified, since the bad point also affects the pixel values ​​of the surrounding pixels, in this embodiment, not only the pixel value of the bad point is corrected, but also the pixel values ​​of the pixel points within the preset neighborhood range of the bad point are corrected, thereby eliminating the influence of the bad point on the corrected image to the greatest extent, which is beneficial to improving the correction effect of the display screen, and thus improving the display effect of the display screen.

[0144] The terms and expressions used in this specification are for illustrative purposes only and are not intended to be limiting. Those skilled in the art will appreciate that various changes may be made to the details of the embodiments described above without departing from the basic principles of the disclosed embodiments. Therefore, the scope of the present invention is determined solely by the claims. In the claims, unless otherwise indicated, all terms are to be interpreted in their broadest reasonable sense.

Claims

1. A method for removing bad pixels from an image based on normal distribution, characterized in that: The method comprises: Separate the RGB channels of the image to be processed to obtain R, G, and B channel images; For each channel image, bad pixel removal processing is performed; the bad pixel removal processing includes: Calculate the pixel average μ and pixel standard deviation σ of the channel image by using the pixel value of each pixel in the channel image after removing the background pixels; According to μ and σ, the pixels suspected to be bad pixels are selected as candidate points to be detected; According to the pixel value of the candidate point to be detected and the pixel values ​​of the pixels within the preset neighborhood of the candidate point to be detected, determining whether the candidate point to be detected is a bad pixel; If the candidate point to be detected is a bad pixel, correcting the pixel values ​​of the candidate point to be detected and the pixel values ​​of the pixel points in the preset neighborhood of the candidate point to be detected to remove the bad pixel; The method of selecting pixels suspected to be bad pixels as candidate points to be detected according to μ and σ includes: According to a preset pixel selection order, one pixel is selected from the channel image in turn as a target pixel, and a screening operation is performed on the target pixel until the screening operation is completed on all the pixels in the channel image, so as to screen out pixel points suspected to be bad pixels as candidate points to be detected; Among them, the screening operation includes: for the target pixel point, judging whether the pixel value of the target pixel point in the channel image is greater than μ+σ; judging whether the target pixel point in the channel image is the highest local pixel value point within a preset neighborhood range; if the pixel value of the target pixel point in the channel image is greater than μ+σ, and the target pixel point in the channel image is the highest local pixel value point within a preset neighborhood range, then the target pixel point is used as a candidate point to be detected.

2. The image bad pixel removal method based on normal distribution according to claim 1, characterized in that: The bad pixel removal process also includes identifying background pixels in the channel image by triangulation; The identifying of background pixels in the channel image by triangulation includes: Counting the pixel distribution of the channel image to obtain a pixel distribution diagram, wherein the abscissa of the pixel distribution diagram is the pixel value and the ordinate is the number of pixel points; Determine the highest point and the lowest point in the pixel distribution diagram, and connect the highest point and the lowest point to form a straight line; The pixel value corresponding to the point in the pixel distribution diagram that is the farthest from the straight line in vertical direction is used as the pixel threshold; Pixels with pixel values ​​less than the pixel threshold are determined as background pixels.

3. The image bad pixel removal method based on normal distribution according to claim 1, characterized in that: The determining whether the candidate point to be detected is a bad pixel according to the pixel value of the candidate point to be detected and the pixel values ​​of the pixels within the preset neighborhood of the candidate point to be detected comprises: Calculate the pixel average μ1 and pixel standard deviation σ1 of the pixel points within the preset neighborhood range of the candidate point to be detected in the channel image; Obtain the color of the image to be processed and the color of the channel image, and determine whether the channel image is a main channel image according to the color of the image to be processed and the color of the channel image, wherein the color of the R channel image is red, the color of the G channel image is green, and the color of the B channel image is blue; If the channel image is the main channel image, determine whether the pixel value of the candidate point to be detected in the channel image is greater than μ1+(3*σ1); if so, determine that the candidate point to be detected is a bad pixel.

4. The method for removing bad pixels from an image based on normal distribution according to claim 3, characterized in that: The determining whether the candidate point to be detected is a bad pixel according to the pixel value of the candidate point to be detected and the pixel values ​​of the pixels within the preset neighborhood of the candidate point to be detected further includes: If the channel image is an auxiliary channel image, obtain the pixel value μ2 of the candidate point to be detected in the main channel image; Calculate the pixel standard deviation σ2 of the pixel points within the preset neighborhood range of the candidate point to be detected in the main channel image; Determine whether the pixel value of the candidate point to be detected in the channel image is greater than μ1+(3*σ1) and greater than μ2-σ2. If so, determine that the candidate point to be detected is a bad pixel.

5. The image bad pixel removal method based on normal distribution according to claim 1, characterized in that: The preset neighborhood range is an 8-neighborhood centered on the candidate point to be detected.

6. The image bad pixel removal method based on normal distribution according to claim 1, characterized in that: The correcting of the pixel values ​​of the candidate point to be detected and the pixel values ​​of the pixel points in the preset neighborhood of the candidate point to be detected comprises: Obtaining pixel values ​​of correction pixels corresponding to the candidate point to be detected in the channel image, wherein the correction pixels are peripheral pixels of pixels within a preset neighborhood of the candidate point to be detected; Calculate the pixel average μ3 of the correction pixel point in the channel image; The pixel value of the candidate point to be detected in the channel image and the pixel value of the pixel point within the preset neighborhood range of the candidate point to be detected in the channel image are adjusted to μ3.

7. An image bad pixel removal device based on normal distribution, characterized in that: The device comprises an image separation module and a bad pixel processing module, wherein: The image separation module is connected to the bad pixel processing module and is used to separate the RGB channels of the image to be processed to obtain R, G, and B channel images; The bad pixel processing module is used to perform bad pixel removal processing on each channel image, and the bad pixel processing module includes a candidate point screening unit to be detected, a bad pixel judgment unit and a bad pixel correction unit; The candidate point screening unit to be detected is connected to the bad point judgment unit, and is used to calculate the pixel average μ and the pixel standard deviation σ of the channel image by using the pixel value of each pixel in the channel image after removing the background pixels; and screen out the pixel points suspected to be bad points as the candidate points to be detected according to μ and σ; The bad pixel judgment unit is connected to the bad pixel correction unit and is used to judge whether the candidate point to be detected is a bad pixel according to the pixel value of the candidate point to be detected and the pixel values ​​of the pixels within the preset neighborhood range of the candidate point to be detected; The bad pixel correction unit is used to correct the pixel values ​​of the candidate point to be detected and the pixel points in the preset neighborhood of the candidate point to be detected to remove the bad pixel when the candidate point to be detected is determined to be a bad pixel; The candidate point screening unit to be detected screens out pixel points suspected to be bad points as candidate points to be detected according to μ and σ, including: selecting a pixel point from the channel image in sequence as a target pixel point according to a preset pixel point selection order, performing a screening operation on the target pixel point, until the screening operation is completed on all pixel points in the channel image, so as to screen out pixel points suspected to be bad points as candidate points to be detected; Among them, the screening operation includes: for the target pixel point, judging whether the pixel value of the target pixel point in the channel image is greater than μ+σ; judging whether the target pixel point in the channel image is the highest local pixel value point within a preset neighborhood range; if the pixel value of the target pixel point in the channel image is greater than μ+σ, and the target pixel point in the channel image is the highest local pixel value point within a preset neighborhood range, then the target pixel point is used as a candidate point to be detected.

8. A calibration system, characterized in that: The system includes a camera, a host computer, a sending card and an LED display screen, wherein: The camera is connected to the host computer and is used to capture the calibrated image of the LED display screen and send the calibrated image to the host computer; The host computer is connected to the sending card, and is used to remove bad pixels in the corrected image according to the image bad pixel removal method based on normal distribution according to any one of claims 1 to 6, and calculate the correction coefficient of each light point in the LED display screen according to the corrected image after removing the bad pixels, and send the correction coefficient to the sending card; The sending card is connected to the receiving card of the LED display screen, and is used to send the correction coefficient to the receiving card of the LED display screen; The LED display screen is used to correct the display data according to the correction coefficient of each light point received through the receiving card, and display the corrected display data.

Images