Blind pixel detection method and system based on maximum value image

Abstract

The application relates to a blind pixel detection method and system based on a maximum value image, which comprises the following steps: obtaining a maximum value image based on a time sequence image; obtaining a suspected blind pixel based on the maximum value image; obtaining neighborhood information of the suspected blind pixel, and obtaining the position of the blind pixel based on the neighborhood information. The application solves the technical problem that the blind pixel detection method in the prior art depends on external equipment and has poor real-time performance.

Description

Technical Field

[0001] This invention relates to the field of infrared remote sensing imaging technology, and more specifically, to a blind pixel detection method and system based on extremum images. Background Technology

[0002] Infrared detectors play a vital role in modern scientific research and industrial applications, and one of the key factors affecting their performance is the detection and compensation of blind pixels. A blind pixel is a pixel in the detector that fails to output a signal correctly due to response failure or anomaly. If left untreated, it will severely impact image quality and the accuracy of subsequent analysis. Currently, blind pixel detection algorithms are mainly divided into two categories: scene-based detection methods and calibration methods.

[0003] The calibration method relies on acquiring single or consecutive frames of uniform radiation images through blackbody imaging, and then determining blind elements based on the response characteristics of the blackbody images. This method typically requires a long interruption during normal system operation, affecting the continuity of practical use. Furthermore, due to the lack of blackbody calibration devices in some satellite equipment, or the long calibration intervals, the calibration method is significantly limited in practical applications.

[0004] In contrast, scene-based blind pixel detection methods do not rely on external blackbody devices, and have therefore become a research hotspot in recent years. In related research, Leng Hanbing et al., in their 2015 paper "Adaptive Blind Pixel Detection and Compensation Based on Fuzzy Median IRFPA" published in Infrared and Laser Engineering, proposed a blind pixel detection and compensation algorithm combining fuzzy median and temporal accumulation. This algorithm extracts potential blind pixels in the scene using a fuzzy median filter and determines the distribution location of blind pixels using cumulative analysis of multiple frames. Gu Guohua et al., in their 2010 paper "Adaptive Blind Pixel Detection and Compensation Algorithm Based on Sliding Window and Multi-Frame Compensation" published in Infrared Technology, accumulated multiple frames, averaged them, and then used the 3σ criterion of a sliding window for blind pixel detection, followed by local median filtering for blind pixel compensation. Zheng Xiao et al., in their 2019 paper "Research on Blind Pixel Detection Algorithm Based on Sliding Window for Infrared Focal Plane Array," developed a blind pixel detection algorithm based on a sliding window and compensated for detected blind pixels using local median filtering.

[0005] Although existing methods have achieved some success, there is still room for improvement in terms of the accuracy and real-time performance of blind pixel detection. Summary of the Invention

[0006] In view of the deficiencies in the prior art, the purpose of this invention is to provide a blind pixel detection method and system based on extremum images.

[0007] According to one aspect of the present invention, a blind pixel detection method based on an extremum image is provided, characterized by comprising the following steps: Step 1: obtaining an extremum image based on a time-series image; Step 2: obtaining suspected blind pixels based on the extremum image; Step 3: obtaining neighborhood information of the suspected blind pixels, and obtaining the position of the blind pixels based on the neighborhood information.

[0008] Preferably, the extremum image includes a maximum value image and a minimum value image.

[0009] Preferably, in step two, obtaining suspected dead pixels includes: obtaining high and low thresholds, and identifying pixels with maximum values ​​less than the low threshold as suspected dead pixels, and identifying pixels with minimum values ​​greater than the high threshold as suspected overheated pixels.

[0010] Preferably, in step three, obtaining the neighborhood information of the suspected blind cell and obtaining the location of the blind cell based on the neighborhood information includes: obtaining the final blind cell detection result by analyzing the neighborhood information of the suspected blind cell.

[0011] Preferably, the maximum and minimum images are defined as follows:

[0012]

[0013] in, For the maximum value image, This is the minimum value image, where T is the total number of frames in the time series image, and t is the number of frames. This represents the grayscale value of the image in frame t, row x, and column y.

[0014] Preferably, obtaining the high and low thresholds includes: Based on the grayscale distribution range of the detector and the value distribution range of the extreme value image, high thresholds and low thresholds are obtained; wherein, the high thresholds and low thresholds satisfy the following formula:

[0015] in, Indicates a high threshold; Indicates a high threshold; This represents the 85th percentile of the detector's gray-scale statistical distribution; This represents the 15th percentile of the detector's gray-scale statistical distribution. This represents the 85th percentile of the gray-level distribution of the image with the minimum value; This represents the 15th percentile of the grayscale distribution of the maximum value image; a percentile indicates that a corresponding proportion of data in the statistical sample is less than or equal to this value.

[0016] Preferably, the 15% is an empirical statistical ratio, which can be adjusted within the range of 10% to 20% according to the noise level of different detector models.

[0017] Preferably, the analysis of the neighborhood information of suspected blind cells includes: Dead pixel detection: Obtain the gray values ​​of the 8 neighboring areas of a suspected dead pixel. If there are pixels with gray values ​​higher than the low threshold, the pixel is determined to be a dead pixel. Otherwise, obtain the gray values ​​of the 24 neighboring areas of the suspected dead pixel. If more than 20 pixels have gray values ​​lower than the low threshold, the pixel is determined to be a normal pixel. Otherwise, the pixel is determined to be a dead pixel. Overheated pixel detection: Obtain the gray values ​​of the 8 neighboring pixels of the suspected overheated pixel. If there are pixels with gray values ​​below the high threshold, they are determined to be overheated pixels. Otherwise, extract the gray values ​​of the 24 neighboring pixels of the suspected overheated pixel. If more than 20 pixels have gray values ​​greater than the high threshold, they are determined to be normal pixels. Otherwise, they are determined to be overheated pixels.

[0018] According to another aspect of the present invention, a blind pixel detection system based on an extremum image is provided, characterized in that it includes: module M1: for obtaining an extremum image based on a time-series image; module M2: for obtaining suspected blind pixels based on the extremum image; and module M3: for obtaining neighborhood information of the suspected blind pixels and obtaining the position of the blind pixels based on the neighborhood information.

[0019] Preferably, the extremum image includes a minimum value image and a maximum value image, and the minimum value image and the maximum value image are defined as follows:

[0020]

[0021] in, For the maximum value image, This is the minimum value image, where T is the total number of frames in the time series image, and t is the number of frames. This represents the grayscale value of the image in frame t, row x, and column y.

[0022] Preferably, module M2 further includes a processing module, which is used to: obtain high and low thresholds based on the grayscale distribution range of the detector and the value distribution range of the extreme value image, and identify pixels with extreme values ​​less than the low threshold as suspected dead pixels, and pixels with extreme values ​​greater than the high threshold as suspected overheated pixels; the processing module is also used to: wherein the high threshold and the low threshold satisfy the following formula:

[0023] in, Indicates a high threshold; Indicates a high threshold; This represents the 85th percentile of the detector's gray-scale statistical distribution; This represents the 15th percentile of the detector's gray-scale statistical distribution. This represents the 85th percentile of the gray-level distribution of the image with the minimum value; This represents the 15th percentile of the grayscale distribution of the maximum value image; a percentile indicates that a corresponding proportion of data in the statistical sample is less than or equal to this value.

[0024] Preferably, the 15% is an empirical statistical ratio, which can be adjusted within the range of 10% to 20% according to the noise level of different detector models.

[0025] Preferably, module M3 further includes a dead pixel detection module and an overheated pixel detection module. The dead pixel detection module is used to: acquire the gray values ​​of the 8 neighboring pixels of a suspected dead pixel; if there are pixels with gray values ​​higher than the low threshold, the pixel is determined to be a dead pixel; otherwise, acquire the gray values ​​of the 24 neighboring pixels of the suspected dead pixel; if more than 20 pixels have gray values ​​lower than the low threshold, the pixel is determined to be a normal pixel; otherwise, the pixel is determined to be a dead pixel.

[0026] The overheated pixel detection module is used to: obtain the gray values ​​of the 8 neighboring pixels of a suspected overheated pixel; if there are pixels with gray values ​​below the high threshold, they are determined to be overheated pixels; otherwise, extract the gray values ​​of the 24 neighboring pixels of the suspected overheated pixel; if more than 20 pixels have gray values ​​greater than the high threshold, they are determined to be normal pixels; otherwise, they are determined to be overheated pixels.

[0027] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention utilizes time-series images to generate maximum and minimum value images, and extracts anomalous pixels based on statistical characteristics, thus enabling blind pixel identification without additional hardware support. This feature significantly improves the practicality of the algorithm in space-constrained environments or where equipment cannot be frequently calibrated, making it particularly suitable for applications such as spaceborne infrared imaging systems and long-term orbital missions.

[0028] 2. This invention introduces 8-neighbor and 24-neighbor grayscale value analysis mechanisms, fully considering image spatial context information, effectively reducing false positives of isolated outliers, and improving the robustness and stability of blind pixel recognition. This mechanism can accurately distinguish between real blind pixels and scene grayscale anomaly pixels, thereby improving detection accuracy and system fault tolerance.

[0029] 3. The detection logic employed in this invention is primarily based on extreme value extraction and threshold comparison, eliminating the need for frequent complex calculations such as Fourier transforms or high-dimensional filtering operations, thus significantly reducing the overall computational cost of the algorithm. Its core algorithm is suitable for deployment on embedded platforms or edge computing devices, facilitating efficient applications in scenarios such as onboard real-time processing and onboard discrimination. Attached Figure Description

[0030] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart of the method provided in this embodiment; Figure 2 This is a logic diagram of step three in this embodiment; Figure 3 This is a schematic diagram of the original image with blind cells; Figure 4 This is a schematic diagram of the maximum value image; Figure 5 This is a schematic diagram of the minimum value image; Figure 6 A schematic diagram of the identified blind cells. Detailed Implementation

[0031] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the scope of protection of the present invention.

[0032] For ease of understanding, the terms or concepts involved in the methods provided in this application are explained below: (1) Time series image A series of images from the same sensor (frame 1 to frame T) acquired over time are used to determine the upper and lower bounds of a pixel's value on the time axis.

[0033] (2) Blind pixels (dead pixels) Pixels exhibiting abnormal responses are commonly categorized into two types: dead pixels and overheated pixels. Dead pixels are noticeably darker and do not respond to stimuli; overheated pixels are noticeably brighter and tend to become overexposed.

[0034] (3) Grayscale The numerical intensity of a pixel in an image refers to its brightness or darkness. In this application, the value of pixel (x,y) in the t-th frame is I. t (x,y) refers to the grayscale value of the pixel.

[0035] (4) Gray level distribution range of the detector This refers to the range of brightness levels that the device can recognize, that is, the grayscale value range from the darkest to the brightest. For example, an 8-bit grayscale image typically contains 256 grayscale levels (0-255), while a 10-bit grayscale image can support 1024 grayscale levels. Additionally, the statistical distribution of grayscale values ​​across all pixels of the detector can also be understood as a histogram. The grayscale distribution is visualized through the grayscale histogram, used to reflect the statistical characteristics of an image, such as contrast and brightness. The shape of the histogram can determine the nature of the image; for example, the histogram of a high-brightness image is right-skewed, while the histogram of a low-contrast image is concentrated in the middle grayscale band.

[0036] (5) Graph of extreme values Two images, which are formed by taking the maximum and minimum gray levels of the same pixel in multiple frames (time series images) over a period of time, are usually called the maximum value image and the minimum value image.

[0037] (6) Range of values ​​in the extreme value graph The distribution (which can be a histogram / cumulative distribution function CDF) obtained after statistically analyzing the gray values ​​of all pixels in the "maximum (or minimum) map" reflects the overall statistical form of the maximum (or minimum) response of each pixel within this time window.

[0038] (7) 8-neighborhood / 24-neighborhood The set of neighboring pixels centered on the target pixel. The 8-neighborhood is the closest ring; the 24-neighborhood is a larger ring (excluding the center, there are 24 pixels in total).

[0039] This application applies to the detection of blind elements in infrared detectors and proposes a blind element detection method based on maximum and minimum value images. This method generates a maximum value image and a minimum value image for each pixel position from multiple frames of images over a period of time (i.e., sequential images), uses the obtained high and low thresholds to coarsely screen suspected blind elements, and then uses neighborhood consistency to verify and obtain the final blind element list.

[0040] Figure 1 The flowchart of this method is shown below, and each step is explained in detail below: Step 1: Based on the time series image, obtain the extremum image; where the extremum image includes the maximum value image and the minimum value image. The definitions of the maximum value image and the minimum value image are as follows:

[0041]

[0042] in, For the maximum value image, This is the minimum value image, where T is the total number of frames in the time series image, and t is the number of frames. This represents the grayscale value of the image in frame t, row x, and column y.

[0043] Figure 3 This is a schematic diagram of the original image with blind cells; Figure 4 This is a schematic diagram of the maximum value image; Figure 5 This is a schematic diagram of the minimum value image.

[0044] Referring to the attached figures above, it can be understood that obtaining the extreme value image involves taking the maximum and minimum grayscale values ​​of the same pixel in T frames and stitching them together into two images. Normal pixels fluctuate with scene and noise levels, so recording the brightest or darkest values ​​in T frames is more reasonable. Furthermore, bad pixels are often "black" or "bright" over time, causing their maximum values ​​to be significantly lower. The extreme value image effectively captures this extremely stable anomaly, displaying it without the need for blackbody calibration.

[0045] Step 2: Based on the extreme value images, obtain high and low thresholds, and identify pixels with extreme values ​​less than the low threshold as suspected dead pixels, and pixels with minimum values ​​greater than the high threshold as suspected overheated pixels, thus obtaining suspected blind pixels; the high and low thresholds are obtained based on the detector's grayscale distribution range and the value distribution range of the extreme value images. The high and low thresholds must satisfy the following formula:

[0046] in, Indicates a high threshold; Indicates a high threshold; This represents the 85th percentile of the detector's gray-scale statistical distribution; This represents the 15th percentile of the detector's gray-scale statistical distribution. This represents the 85th percentile of the gray-level distribution of the image with the minimum value; This represents the 15th percentile of the grayscale distribution of the maximum value image; a percentile indicates that a corresponding proportion of data in the statistical sample is less than or equal to this value.

[0047] It should be noted that the 15% mentioned is an empirical statistical ratio, which can be adjusted within the range of 10% to 20% depending on the noise level of different detector models.

[0048] Based on the above scheme, it can be understood that the high and low thresholds are determined according to the grayscale distribution and extreme value distribution of the detector. Then, by comparing with the maximum and minimum value images of each pixel, pixels with maximum values ​​less than the low threshold are suspected dead pixels, and pixels with minimum values ​​greater than the high threshold are suspected overheated pixels, thus completing the coarse screening.

[0049] Figure 2 The diagram illustrates the logic for further screening of suspected blind cells in step three. Step three is explained in detail below: Step 3: Obtain the neighborhood information of suspected blind pixels. Based on this neighborhood information, analyze the neighborhood information of the suspected blind pixels to obtain the final blind pixel detection result, such as... Figure 6 As shown, the analysis of the aforementioned neighborhood information further needs to be divided into dead pixel detection and overheated pixel detection, specifically: Dead pixel detection: Obtain the gray values ​​of the 8 neighboring areas of a suspected dead pixel. If there are pixels with gray values ​​higher than the low threshold, the pixel is determined to be a dead pixel. Otherwise, obtain the gray values ​​of the 24 neighboring areas of the suspected dead pixel. If more than 20 pixels have gray values ​​lower than the low threshold, the pixel is determined to be a normal pixel. Otherwise, the pixel is determined to be a dead pixel. Overheated pixel detection: Obtain the gray values ​​of the 8 neighboring pixels of the suspected overheated pixel. If there are pixels with gray values ​​below the high threshold, they are determined to be overheated pixels. Otherwise, extract the gray values ​​of the 24 neighboring pixels of the suspected overheated pixel. If more than 20 pixels have gray values ​​greater than the high threshold, they are determined to be normal pixels. Otherwise, they are determined to be overheated pixels.

[0050] Based on the above scheme, it can be understood that the core idea of ​​the review in this field is to determine whether the pixel is consistent with its neighboring units. If they are inconsistent, it can be judged as a bad pixel. If they are consistent, it cannot be directly judged as a blind pixel (bad pixel) and further judgment is required, that is, to expand the comparison range. If it is consistent with the vast majority, it is judged as a normal pixel; otherwise, it is judged as a blind pixel.

[0051] The present invention also provides a blind pixel detection system based on extremum images. The blind pixel detection system based on extremum images can be implemented by executing the process steps of the blind pixel detection method based on extremum images. That is, those skilled in the art can understand the blind pixel detection method based on extremum images as a preferred embodiment of the blind pixel detection system based on extremum images.

[0052] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, enabling the system and its various devices, modules, and units to function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered both software modules implementing the method and structures within the hardware component.

[0053] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A blind pixel detection method based on extremum images, characterized in that, Includes the following steps: Step 1: Obtain the maximum / minimum value image based on the time series image; Step 2: Based on the extreme value image, obtain suspected blind cells; Step 3: Obtain neighborhood information of suspected blind cells, and obtain the location of blind cells based on the neighborhood information.

2. The detection method according to claim 1, characterized in that, The extreme value image includes a maximum value image and a minimum value image.

3. The detection method according to claim 2, characterized in that, The definitions of the maximum and minimum value images are as follows: in, For the maximum value image, This is the minimum value image, where T is the total number of frames in the time series image, and t is the number of frames. This represents the grayscale value of the image in frame t, row x, and column y.

4. The detection method according to claim 2, characterized in that, In step two, obtaining suspected dead pixels includes: obtaining high and low thresholds, and identifying pixels with maximum values ​​less than the low threshold as suspected dead pixels, and identifying pixels with minimum values ​​greater than the high threshold as suspected overheated pixels.

5. The detection method according to claim 4, characterized in that, The acquisition of high and low thresholds includes: Based on the grayscale distribution range of the detector and the value distribution range of the extreme value image, high thresholds and low thresholds are obtained; wherein, the high thresholds and low thresholds satisfy the following formula: in, Indicates a high threshold; Indicates a high threshold; This represents the 85th percentile of the detector's gray-scale statistical distribution; This represents the 15th percentile of the detector's gray-scale statistical distribution. This represents the 85th percentile of the gray-level distribution of the image with the minimum value; This represents the 15th percentile of the grayscale distribution of the maximum value image; a percentile indicates that a corresponding proportion of data in the statistical sample is less than or equal to this value.

6. The detection method according to claim 1, characterized in that, In step three, obtaining the neighborhood information of the suspected blind cell and obtaining the location of the blind cell based on the neighborhood information includes: obtaining the final blind cell detection result by analyzing the neighborhood information of the suspected blind cell.

7. The detection method according to any one of claims 6, characterized in that, The analysis of neighborhood information of suspected blind cells includes: Dead pixel detection: Obtain the gray values ​​of the 8 neighboring areas of a suspected dead pixel. If there are pixels with gray values ​​higher than the low threshold, the pixel is determined to be a dead pixel. Otherwise, obtain the gray values ​​of the 24 neighboring areas of the suspected dead pixel. If more than 20 pixels have gray values ​​lower than the low threshold, the pixel is determined to be a normal pixel. Otherwise, the pixel is determined to be a dead pixel. Overheated pixel detection: Obtain the gray values ​​of the 8 neighboring pixels of the suspected overheated pixel. If there are pixels with gray values ​​below the high threshold, they are determined to be overheated pixels. Otherwise, extract the gray values ​​of the 24 neighboring pixels of the suspected overheated pixel. If more than 20 pixels have gray values ​​greater than the high threshold, they are determined to be normal pixels. Otherwise, they are determined to be overheated pixels.

8. A blind pixel detection system based on extremum images, characterized in that, include: Module M1: Used to obtain the maximum / minimum value image based on the time series image; Module M2: Used to obtain suspected blind cells based on the maximum / minimum value image; Module M3: Used to obtain neighborhood information of suspected blind cells, and to obtain the location of the blind cells based on the neighborhood information.

9. The detection system according to claim 8, characterized in that, The extremum image includes a minimum value image and a maximum value image, which are defined as follows: in, For the maximum value image, This is the minimum value image, where T is the total number of frames in the time series image, and t is the number of frames. This represents the grayscale value of the image in frame t, row x, and column y.

10. The detection system according to claim 9, characterized in that, The module M2 further includes a processing module, which is used to: obtain high and low thresholds based on the gray level distribution range of the detector and the value distribution range of the extreme value image, and identify pixels with extreme values ​​less than the low threshold as suspected dead pixels, and pixels with extreme values ​​greater than the high threshold as suspected overheated pixels; the processing module is also used to: wherein the high threshold and the low threshold satisfy the following formula: in, Indicates a high threshold; Indicates a high threshold; This represents the 85th percentile of the detector's gray-scale statistical distribution; This represents the 15th percentile of the detector's gray-scale statistical distribution. This represents the 85th percentile of the gray-level distribution of the image with the minimum value; This represents the 15th percentile of the grayscale distribution of the maximum value image; a percentile indicates that a corresponding proportion of data in the statistical sample is less than or equal to this value. Preferably, the 15% is an empirical statistical ratio, which can be adjusted within the range of 10% to 20% according to the noise level of different detector models.

Images