Method for processing an image and image processor unit - Patents.com

Abstract

The present invention relates to a method for processing image data provided by an image sensor, the image data comprising an array of pixels, the method comprising the steps of: a) determining a respective local brightness difference between a selected pixel and a set of pixels located adjacent to the respective pixel and assigned the same color as the respective pixel; b) identifying the selected pixel as a defective pixel when the local brightness difference for the selected pixel exceeds an upper threshold and / or is less than a lower threshold and when the local brightness difference for the selected pixel exceeds a weighted maximum local brightness difference determined for a set of pixels located adjacent to the selected pixel and / or is less than a weighted minimum local brightness difference determined for a set of pixels located adjacent to the selected pixel; A method is disclosed comprising the steps of:

Description

[Technical field]

[0001] The present invention relates to a method for processing image data provided by an image sensor, the image data comprising an array of pixels.

[0002] The invention further relates to an image processor unit for processing image data provided by an image sensor, the image data comprising an array of pixels.

[0003] The invention further relates to a computer program for carrying out the above method. [Background technology]

[0004] Digital imagers are widely used in everyday items such as smartphones, tablets, notebooks, cameras, cars, and wearables. Many of the imaging systems in those products have autofocus (AF) features to produce sharp images / videos. The traditional approach to autofocus is based on contrast detection. The lens is moved to the position where the contrast of the scene is highest. Contrast autofocus is generally slow, so the technology of phase detection autofocus (PDAF) can be combined with it to improve the speed and sometimes the accuracy of focusing.

[0005] PDAF sensors have so-called phase-detection (PD) pixels that are arranged across the entire sensor area. Phase information derived from the phase-detection pixels can be used to determine the focal distance, which then drives the lens to move to a position with optimal focus, a process that is generally very fast if the phase information is accurate enough. The phase-detection pixels are generally arranged periodically across the entire sensor in the horizontal and vertical directions. After extracting the phase information from those pixels, they must be concealed or corrected. Otherwise, they may appear as a mesh of defects clustered across the image. It is therefore important to have a phase-detection pixel concealment (PPC) module in the image signal processor to maintain high image quality.

[0006] N. El-Yamany's "Robust Defect Pixel Detection and Correction for Bayer Imaging Systems" in IS&T International Symposium on Electronic Imaging 2017, pp. 46-51, discloses a method for identifying singlets and couplets of hot pixels, cold pixels, or a mixture of both types. If those pixels are not corrected early during the image processing pipeline, demosaicing and filtering operations will cause them to appear as diffuse colored clusters that are detrimental to image quality. The method operates on raw data coming from a Bayer sensor. A defective pixel is identified if two conditions are met. The first condition checks whether the pixel is significantly different from its same-colored neighbors in an S×S Bayer window centered on the pixel. The second condition tests whether the local brightness difference in the pixel per color channel is significantly higher for hot pixels or significantly lower for cold pixels than the minimum local brightness difference when a 3×3 Bayer window is centered on the pixel.

[0007] Detected defective pixels are replaced with robust detail-preserving estimates determined through the use of directional filters.

[0008] E. Chang, "Kernel-size selection for defect pixel identification and correction," in Proc. SPIE6502, Digital Photography III, 65020J, 20-02-2007, describes a minimum and maximum limiting filter with varying kernel size for correction of defective pixels.

[0009] "Adaptive defect correction and noise suppression module in the CIS image processing system" by S. Wang, S. Yao, O. Faurie, and Z. Shi in Proc. SPIE Int. Symposium on Photoelectronic Detection and Imaging, vol. 7384, p. 73842V-1-6 describes a spatially adaptive noise suppression algorithm that combines the correction functions of defective pixels for implementation in a CMOS image sensor chip. A center-weighted median filter is applied to correct the defective pixels. Random noises are treated separately according to their background detail level.

[0010] "Adaptive pixel defect correction" by A. Tanbakuchi, A. van der Sijde, B. Dillen, A. Theuwissen and W. de Haan in Proc. SPIE Sensors and Camera Systems for Scientific, Industrial and Digital Photography Applications IV, Vol. 5017, pp. 360-370, 2003, presents a defect correction algorithm that utilizes raw Bayer image data. When a pixel in an image is found to be defective, the neighboring pixels provide the best information to interpolate the defective pixel. Directional derivatives are used to correlate the closest point to the defective color plane.

[0011] M. Schoberi, J. Seiler, B. Kasper, S. Foessel and A. Kaup, "Sparsity-based detect pixel compensation for arbitrary camera raw images," in IEEE International Conference on Acoustics, Speech and Signal Processing, pp. 1257-1260, 2011, disclose an algorithm for detecting defective pixels and interpolating missing values ​​for the defective pixels. Previously interpolated pixels are reused. Summary of the Invention

[0012] It is an object of the present invention to provide an improved method and image processor unit which provides a robust scheme for detecting defective pixels.

[0013] The object is achieved by a method with the features of claim 1, by an image processor unit with the features of claim 12 and by a computer program according to claim 13. Preferred embodiments are disclosed in the dependent claims.

[0014] a) determining a respective local brightness difference between a selected pixel and a set of pixels located adjacent to the respective pixel and assigned the same color as the respective pixel; b) identifying the selected pixel as a defective pixel when the local brightness difference for the selected pixel exceeds an upper threshold and / or is less than a lower threshold and when the local brightness difference for the selected pixel exceeds a weighted maximum local brightness difference determined for a set of pixels located adjacent to the selected pixel and / or is less than a weighted minimum local brightness difference determined for a set of pixels located adjacent to the respective pixel; It is proposed to carry out the steps:

[0015] Two detection conditions are combined to identify a selected pixel as defective. The conditions are in principle independent of the color filter array used by the image sensor to give the pixel array. Different types of defects can be identified by combinations of conditions that can be easily tuned by given thresholds and weighting coefficients.

[0016] By comparing the local brightness difference for the selected pixel with a given threshold, it is checked whether the selected pixel is significantly different from its same-color neighbors in a small window that includes the selected pixel and a set of surrounding pixels. The selected pixel may be placed in the center of the small window when the method is performed in a pixel array having the pattern of the respective color filter array, i.e. before the demosaicing step.

[0017] In the second condition, the local brightness for the selected pixel is compared against the local brightness differences determined for the neighboring pixels located adjacent to the selected pixel. Preferably, it is checked whether the local brightness difference of the selected pixel is significantly different from the local brightness differences of the neighbors of the same color in a small window in which the selected pixel is centered. The maximum and minimum local brightness difference are not necessarily the absolute maximum or minimum of the set of local brightness differences. This is understood as one selected maximum / minimum, for example the first (absolute) or preferably the second or even the third, fourth, ... maximum / minimum of the set, which corresponds to excluding the absolute maximum / minimum as an outlier from the set of values ​​representing the local brightness difference. Thus, also the local maximum / minimum or adjusted maximum / minimum are included in the meaning of the maximum / minimum local brightness difference according to the invention.

[0018] The second condition makes it possible to distinguish between real defective pixels and small image details.

[0019] The combination of the two conditions gives a robust scheme for identifying selected pixels as defects that can be easily performed on an image data stream comprising pixel data of an array of pixels with low computational complexity and reliable results. The method can be used for a variety of color filter arrays such as standard Bayer, Quad Bayer, RGBW, Nonacell and Hexadeca Bayer color filter array patterns. It is possible to easily identify singlets or couplets of hot defects, cold defects or a mixture of both types of defects. The method is also based on identifying higher order defects.

[0020] For determining the respective local brightness difference for the pixel selected in step a), an average value of a set of pixels located adjacent to the respective pixel may be calculated. The respective local brightness difference for the selected pixel may then be calculated as the difference between the value of the selected pixel and the calculated average value of the set of adjacent pixels.

[0021] The neighboring pixels may be located immediately adjacent to the selected pixel. It is also possible to consider neighboring pixels located near the selected pixel with pixels between the selected pixel and the neighboring pixels in the pixel array.

[0022] The set of neighboring pixels considered for determining the respective local brightness difference for the selected pixel may be selected only from those neighboring pixels that are assigned the same color as the selected pixel, and the average value of the neighboring pixels of the set of pixels is calculated to determine the local brightness difference as the difference between the value of the selected pixel and the calculated average value of the set of neighboring pixels.

[0023] Thus, each selected pixel may be assigned a respective local brightness difference value.

[0024] The average value may be calculated as the average, alpha trimmed mean or median of a set of pixels surrounding the selected pixel and assigned the same color as the selected pixel.

[0025] When determining the alpha trimmed average value, pixel values ​​are sorted by size on the top and bottom alpha portions of some pixel values ​​that are excluded from the average. For example, a 2 alpha trimmed average filter removes two pixel values ​​for calculating the average value, namely the lowest and highest values ​​(i.e. the luminance level for each color).

[0026] Other methods for calculating the average value, such as a midpoint filter, are also applicable.

[0027] The set of pixels considered in step a) of determining a local brightness difference for a selected pixel may be located in a window of a predetermined size, preferably a 4x4, 5x5, 6x6 or 8x8 window, wherein the set of pixels are assigned to the same color of the selected pixel.

[0028] The selected pixel may be located in the center of this window.

[0029] By use of a window that includes the selected pixel, where the frame of the window surrounds the selected pixel, pixels of the same color within the window are selected for the set of neighboring pixels.

[0030] The first condition may be checked to identify a hot or cold pixel by determining a maximum local brightness difference of a set of pixels located adjacent to the selected pixel in a window of a predetermined size, and identifying the selected pixel as a defective hot pixel when the local brightness difference exceeds a weighted maximum local brightness difference, or determining a minimum local brightness difference of a set of pixels located adjacent to the selected pixel in a window of a predetermined size, and identifying the selected pixel as a defective cold pixel when the local brightness difference is less than a weighted minimum local brightness difference.

[0031] Improved detection results for identifying defective pixels may be achieved by considering the second largest or second smallest local difference instead of the maximum or minimum of the local difference. By considering the second largest local difference value, which is smaller than the maximum local difference value, or in the case of the minimum local difference value, by considering the second smallest value, instead of the minimum, outliers in the local difference values ​​appear to be excluded. This may be compared to considering the median for the mean instead of the mean.

[0032] More preferably, a low-order maximum of the local difference, e.g., the second largest local brightness difference, is determined from the set of positive values ​​of the local brightness difference. Similarly, a low-order minimum of the local difference, e.g., the second smallest local brightness difference, is determined from the set of negative values ​​of the local brightness difference.

[0033] Another option is to determine the maximum value from the absolute values ​​of the local brightness differences of a set of pixels located adjacent to the selected pixel, in which case the consideration of the minimum value for the second condition is ignored.

[0034] The maximum or minimum local difference in the sense of the present invention also includes the second, third, fourth etc. largest or smallest local difference values ​​when excluding outliers. The upper and / or lower thresholds considered for identifying the selected pixel as a defective pixel may be predetermined and stored in memory for application to the method. Alternatively, the upper and lower thresholds may be adapted as required to improve the final image quality. Preferably, the upper and / or lower thresholds may be adapted by the use of a neural network.

[0035] The same applies to the weighting coefficients taken into account for the minimum and maximum local brightness differences, which again can be preset as fixed values ​​or adapted as variables to improve image quality. Preferably, the weighting coefficients taken into account for the minimum and maximum local brightness differences are adaptively learned and improved by the use of a neural network to give a high final image quality. The weighting coefficients can be set as one value for the maximum and minimum local brightness differences. Optionally, each of the maximum and minimum local brightness differences is assigned a respective weighting coefficient.

[0036] Step a) of determining respective local brightness differences for the selected pixels and step b) of identifying the selected pixels as defective pixels may be performed iteratively by selecting each pixel of a pixel array provided by an image camera for the image and repeating the method for each selected pixel. Alternatively, rather than each pixel of the pixel array, various pixels of the pixel array, for example only pixels of a particular selected color within the pixel array, may be selected when repeating the method.

[0037] After identifying pixels as defective, the image data comprising the pixel array may be further processed by correcting the pixels identified as defective.

[0038] The image processor unit is arranged to process the image data by performing the method steps described above. The image processor unit may be a hardware unit such as an FPGA or a suitably programmed microprocessor or a data signal processor.

[0039] The method may be readily performed by use of a computer program comprising instructions which, when the program is executed by a processing unit, cause the processing unit to perform the steps of the aforementioned method.

[0040] The invention is illustrated by way of example in the accompanying drawings, in which: [Brief description of the drawings]

[0041] [Figure 1] 1 is a block diagram of a method for processing image data. [Diagram 2] FIG. 2 illustrates an example pixel array for a Bayer color filter array pattern with a selected pixel and a window of adjacent pixels. [Diagram 3] 4 is a block diagram of an example method for correcting a pixel identified as defective. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0042] FIG. 1 shows a block diagram of a method for processing image data provided by an image sensor and the functional units of an image processor unit, in which the image data is input to an array of pixels PA, which is used as input for further steps. IN Equipped with.

[0043] Step a) is a step of forming a pixel array PA IN determining a respective local brightness difference between the selected pixel of and a set of pixels located adjacent to the respective pixel and assigned the same color as the respective pixel;

[0044] The local brightness difference LBD(x,y) for a selected pixel location x,y in the pixel array may be determined as the difference between the value of the selected pixel P(x,y) and the robust estimate p_est(x,y) of the selected pixel P(x,y) as follows:

[0045]

number

[0046] The robust estimate p_est(x,y) of the selected pixel P(x,y) may be determined by using pixel values ​​of a set of pixels located adjacent to the selected pixel P(x,y). For example, the average or median value of all pixels of the same color as the selected pixel, located within a predetermined window, said window including the selected pixel P(x,y), may be calculated for the robust estimate p_est(x,y). Preferably, the selected pixel P(x,y) is located in the center of the window.

[0047] To detect or identify defective couplets, the robust estimate p_est(x,y) must be robust to the presence of couplets on the raw window centered on the selected pixel P(x,y). One possibility is to compute p_est(x,y) as the alpha-trimmed average (with α=2) of neighbors of the same color in the raw window.

[0048] This can be better understood by the example array shown in Figure 1 for the Bayer color filter array pattern. For example, a selected pixel P(x,y) is G(2,3). A small window 3x3 around this selected pixel G(2,3) is used to select pixels of the same value for the set of neighboring pixels. In this example case, pixels G(1,2), G(3,2), G(1,4) and G(3,4) are selected for the set of neighboring pixels.

[0049] The average of this set of neighboring pixels G(1,2), G(3,2), G(1,4) and G(3,4) is calculated and the resulting average is used as the robust estimate p_est(x,y).

[0050] In the case of a 2 alpha trimmed average, for example, the highest and lowest values ​​of a set of four pixels are ignored and the average is calculated from the remaining median pixel values ​​of a set of (e.g., four) adjacent pixels (e.g., using the two remaining four pixels).

[0051] Similarly, a window of a particular size is centered on the selected blue B(x,y) pixel for the selected red R(x,y) pixel. The size of the window used for the selected pixel of each color may be selected differently depending on the respective color. The size of the window may be selected, for example, such that it includes at least three adjacent pixels of the same color that are assigned for the set of neighboring pixels for the selected pixel of a particular color. The size of the window may be selected depending on each different color filter array pattern and / or the location of the particular selected pixel P(x,y). The objective is to determine a robust estimate of the selected pixel value P(x,y) by using a sufficient number of neighboring pixels of the same color.

[0052] A local demosaicing process may additionally be performed, which is one option to protect a sufficient number of neighboring pixels, so that a small window around the selected pixel P(x,y) may be chosen with a sufficiently large number of pixel values ​​of the same color as the selected pixel color.

[0053] In step b) of FIG. 1, when the local brightness difference LBD(x,y) for the selected pixel P(x,y) exceeds an upper threshold S_hot and / or is smaller than a lower threshold S_cold, the selected pixel Py(x,y) is identified as a defective pixel. This can be expressed by the following equation:

[0054] For hot pixels: LBD(x,y)>S_hot For cold pixels: LBD(x,y) <S_cold Alternatively, one common threshold S may be set for the upper and lower thresholds, when comparing the absolute value of the local brightness difference with this common upper / lower threshold S, which may be expressed by the following formula:

[0055]

number

[0056] The condition that the local brightness difference is greater than or less than the threshold S or the upper and lower thresholds, S_hot and S_cold, includes the condition that the local brightness difference LBD(x,y) is also equal to the thresholds S, S_hot, S_cold.

[0057] By using the weights and a common threshold, it is checked within one step whether the local brightness difference LBD(x,y) exceeds an upper threshold S and is lower than a lower threshold S.

[0058] Furthermore, in step b), the second condition is also applied, i.e. the selected pixel P(x,y) is identified as defective in case the local brightness difference LBD(x,y) at the selected pixel P(x,y) is significantly different from the local brightness difference LBD(i,j) of the neighboring pixels, which makes it possible to distinguish between real defective pixels and small image details.

[0059] Thus, a local brightness difference is calculated not only for the selected pixel P(x,y) but also for the pixels P(i,j) located adjacently, i.e. as direct or indirect neighbors of the selected pixel P(x,y). For calculating the robust estimate p_est(x,y), the pixels of the set of adjacent pixels and their respective local brightness differences LBD(x,y) are considered to determine a weighted maximum local brightness difference LBD(i,j) and a weighted minimum local brightness difference LBD(i,j). This can be expressed by the following mathematical formula:

[0060] For hot pixels: LBD(x,y)>F_hot×max(LBD(i,j)∀(i,j)∈Ω) For cold pixels: LBD(x,y) <F_hot×min(LBD(i,j)∀(i,j)∈Ω) Ω is the set of same-color pixels in the pixel array within a window centered on the selected pixel P(x,y). LBD(x,y) is the local brightness difference determined for the selected pixel P(x,y). F_hot and F_cold are weighting coefficients that may have the same or different values. Location i, j relate to a particular pixel location x,y in the pixel array window that is related to the neighboring pixels of the same color as the selected pixel P(x,y).

[0061] To be robust against couplets, Ω must include a robustness strategy, for example by excluding pixels of the same color with the maximum / minimum LBD values ​​in a window. This can be achieved by considering the second largest and second smallest values ​​of a set of local brightness differences LBD(i,j) for a set of adjacent pixels. This can be expressed by the following equation:

[0062] For hot pixels: LBD(x,y)>F_hot×2nd max(LBD_positive(i,j)∀(i,j)∈Ω) For cold pixels: LBD(x,y) <F_hot×2nd min(LBD_negative∀(i,j)(i,j)∈Ω) LBD_positive(i,j) denotes the positive local brightness difference in set Ω at each pixel location i,j in the pixel matrix. LBD_negative(i,j) denotes the negative local brightness difference in set Ω at each pixel location i,j in the pixel matrix.

[0063] The upper and lower thresholds S, S_hot and S_cold and the weighting factors F_hot and F_cold can be physically stored in memory or can be changed. They are input to step c) which includes a neural network for adapting the thresholds S and the weighting factors F.

[0064] Using small weighting factors F, F_hot and F_cold will result in more false positives. Using large values ​​for the weighting factors F, F_hot and F_cold will result in fewer false positives.

[0065] Preferably, S may be selected from the group [0, 2^NBits-1], with a weighting factor F≧1, where NBits is the bit depth of the raw data.

[0066] In step c), the thresholds S, S_hot, S_cold and the weighting factors F, F_hot, F_cold may be tuned together to achieve the best trade between defect detection and detail preservation, i.e., image quality. The adaptation may be performed by using a learning-based framework, i.e., a neural network, to achieve an improved defect detection rate close to 100% and a reduced false positive rate, preferably close to 0%.

[0067] The neural network may be trained with the goal of achieving optimized image quality.

[0068] The result of step b) is the identity of the defective pixel D(x,y), which may be given by a simple list of x,y positions that identifies the pixel in this location as defective. Additional information about such defects (hot pixels, cold pixels) may optionally be included.

[0069] In step d), the step of correcting defective pixels is performed by the incoming pixel array PA IN and the pixel or pixel location D(x,y) identified as defective.

[0070] Correction of defective pixels is known in the art. A preferred method for correcting defective pixels is to OUT may be selected to achieve a corrected pixel array at

[0071] Preferably, the correction of defective pixels seeks to preserve image detail and therefore uses directional correction where possible and relies on limiting the information used in the correction to a small neighborhood.

[0072] FIG. 3 shows an exemplary preferred embodiment for the correction of defective pixels using a nearest neighbor correction step e) to correct the pixel through interpolation of nearest same-color neighbors.

[0073] In separate paths, steps f), g), h) and i) are performed to perform a directional correction h) and an omnidirectional correction i), all based on the first step f) of omnidirectional pre-correction. Prior to performing the directional correction step h), a directional filtering is performed in step g).

[0074] The defect correction mechanism determines which of the three operations is utilized for correction depending on the color filter array pattern and the color channel arrangement in the color filter array data. In particular, when a given color channel in the color filter array pattern exhibits symmetry and thus facilitates directivity estimation and correction, the default correction scheme may follow a three-step procedure with an omnidirectional pre-correction step f), a directional filtering step g), and a directional correction step h), or alternatively, an omnidirectional correction step i).

[0075] In step f) "omnidirectional pre-correction", a robust estimate is calculated for all identified defective pixels in the raw image data centered on the pixel being corrected, and all non-defective pixels are left untouched. The purpose of this pre-correction (pre-processing) in step f) is to provide an intermediate clean estimate of the raw image data around the pixel being corrected, thus facilitating robust directional identification at that pixel location. The robust estimate p_est(x,y) defined as above may be used in this step.

[0076] The directional filtering step g) serves to identify feature directions at the pixel location being corrected. Thus, directional filtering is performed at the pixel location by convolving the pixel with its same-color neighbors with a set of directional filters. Based on the maximum absolute response of the filter set, the feature direction is identified.

[0077] The number and size of the coefficients depends on several factors:

[0078] 1. Several directions that are intended to be identified and supported in the correction scheme 2. The size of the raw support around which the pixel is centered. 3. Color Filter Array Channel Placement 4. The type of filter, e.g. gradient-based or other Once the feature direction in the pixel being corrected has been identified, the pixel may be corrected by interpolating clean, same-color pixels along the direction identified in step h) of the directional correction. To reduce or eliminate the possibility of correction artifacts, the directional correction is generally limited to a small neighborhood of the pixel.

[0079] If the feature direction is not identified, for example, in various flat regions, a correction of the omnidirectionality can be traced by replacing the defective pixel with its robust estimate, for example, p_est(x,y). The step is identified as step i) in FIG.

[0080] When a given color channel in the filter array pattern, such as the R, G, or B channels in an RGBW color filter array pattern or the R and G and G channels in a Quad Bayer color filter array pattern, does not exhibit symmetry, the defect correction scheme may only rely on step e) of nearest neighbor correction.

[0081] In step e), a robust estimate is used to correct the defective pixel. One possibility for such an estimate is, for example, the alpha-trimmed average of pixels of the same color in the neighborhood of the pixel being corrected. Another possibility is some weighted average of their same-color neighbors.

[0082] Of course, directional or content-aware correction will give better image quality than omnidirectional correction, so when interpolating the nearest neighbors, the interpolation strategy preferably tries to identify the direction as much as possible or takes into account the content in the raw image data in the pixel P(x,y) being corrected.

[0083] Step d) of FIG. 1, which corrects pixels identified as defective, can be used with various color filter array patterns such as standard Bayer, Quad Bayer, RGBW and Hexadeca Bayer color filter array patterns. The definitions of robust pre-correction, directional filters, directional / omnidirectional correction (interpolation) and nearest neighbor correction can take on different flavors. They can be changed from a fixed global definition to a local definition depending on the given complexity and desired image quality.

[0084] The method for processing image data according to the present invention can correct singlets and / or couplets of hot defects, cold defects or a mixture of both types. For example, with appropriate adjustment of the size of the raw image data, higher order defects can be handled.

Claims

1. 1. A method for processing image data provided by an image sensor, the image data comprising an array of pixels; a) determining a respective local brightness difference between a selected pixel and a set of pixels located adjacent to the respective pixel and assigned the same color as the respective pixel; b) identifying the selected pixel as a defective pixel when the local brightness difference for the selected pixel exceeds an upper threshold and / or is less than a lower threshold and when the local brightness difference for the selected pixel exceeds a weighted maximum local brightness difference determined for the set of pixels located adjacent to the selected pixel and / or is less than a weighted minimum local brightness difference determined for the set of pixels located adjacent to the selected pixel; The method of claim 1, characterized by the steps of:

2. 2. The method of claim 1, characterized in that in step a) an average value for the set of pixels located adjacent to the selected pixel is calculated, and determining the respective local brightness difference for the selected pixel by calculating the difference between the value of the selected pixel and the calculated average value of the set of adjacent pixels.

3. The method of claim 2 , wherein the average value is calculated as the average, alpha trimmed mean or median value of a set of pixels surrounding the selected pixel and assigned to the same color as the selected pixel.

4. 4. The method according to claim 1, characterized in that the set of pixels considered in step a) of determining the local brightness difference for the selected pixel is located in a window of a predetermined size, preferably a 4x4, 5x5, 6x6 or 8x8 window, and in which the set of pixels are assigned to the same color of the selected pixel.

5. The method of claim 4 , wherein the selected pixel for which the local brightness difference is determined is located in the center of the window.

6. determining a maximum local brightness difference of a set of pixels located adjacent to said selected pixel in a window of a predetermined size; - identifying the selected pixel as a defective hot pixel when the local brightness difference exceeds the weighted maximum local brightness difference; 6. The method according to claim 1 , characterized in that

7. determining a minimum local brightness difference of a set of pixels located adjacent to said selected pixel in a window of a predetermined size; - identifying said selected pixel as a defective cold pixel when said local brightness difference is less than said weighted minimum local brightness difference; 7. The method according to claim 1, characterized in that

8. - determining a second largest local brightness difference for a set of pixels located adjacent to the selected pixel in a window of a predetermined size and / or determining a second smallest local brightness difference for the set of pixels located adjacent to the selected pixel in a window of a predetermined size; identifying the selected pixel as a defective hot pixel when the local brightness difference exceeds the weighted second largest local brightness difference and / or identifying the selected pixel as a defective cold pixel when the local brightness difference is smaller than the weighted second smallest local brightness difference; 6. The method according to claim 1 , characterized in that

9. 9. The method of claim 8, characterized in that it determines the second largest local brightness difference for a set of pixels located adjacent to the selected pixel in a window of a predetermined size from the local brightness difference values ​​in the set having positive values, and / or determines the second smallest local brightness difference for the set of pixels located adjacent to the selected pixel in a window of a predetermined size from the local brightness difference values ​​in the set having positive values.

10. adapting said upper and / or lower thresholds taken into account for identifying defective pixels and / or selected pixels as weighting factors for said minimum and maximum local brightness differences by using a neural network; 10. The method according to claim 1, characterized in that

11. 11. The method according to claim 1, wherein the step a) of determining the respective local brightness difference for selected pixels and the step b) of identifying the selected pixels as defective pixels are performed iteratively by selecting each pixel of a pixel array provided by an image camera for an image and repeating the method for each selected pixel.

12. 12. The method according to claim 1, characterized by correcting pixels identified as defective.

13. 13. An image processor unit for processing image data provided by an image sensor, said image data comprising an array of pixels, characterised in that said image processing unit is configured to perform the steps of the method according to one of claims 1 to 12.

14. A computer program comprising instructions which, when said program is executed by a processing unit, cause said processing unit to perform the steps of the method according to one of claims 1 to 12.

Images