Systems and methods for maintaining pixel intensity during rounding

Abstract

A method for maintaining pixel intensity implemented by an image sensor processor, comprising: receiving, by an image processor, a numerical value indicating a corresponding pixel intensity; determining, by the image processor, whether a least significant portion of the received numerical value is equal to a predetermined numerical value; and in response to determining that the least significant portion of the received numerical value is equal to the predetermined numerical value, rounding, by the image processor, the received numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence, and if the least significant portion of the received numerical value is not equal to the predetermined value, then rounding the received numerical value to a higher or lower value based on the received numerical value; and fusing the rounded value.

Description

Technical Field

[0001] The present disclosure relates to the field of image processing technology, and in particular, to a system and method for maintaining pixel intensity during rounding, and particularly to a method for maintaining pixel intensity implemented by an image sensor processor and a camera system. Background Art

[0002] Camera modules in commercial products such as standalone digital cameras, mobile devices, automotive components, and medical devices include image sensors having pixel arrays. The pixel array includes a plurality of pixels arranged in an N×N array. Pixel intensity is a key feature for classifying the information stored in a pixel. For a grayscale image, the intensity of a pixel is typically represented as a numerical value, or for a color image, the intensity of a pixel is represented as three numerical values. An N×N pixel group of the pixel array can be fused, for example, by averaging the pixel intensities of each pixel within the N×N pixel array to form a superpixel. Maintaining consistent pixel intensity values during fusion is crucial for maintaining an accurate image for display. Pixel fusion results in a higher signal-to-noise ratio but at the cost of reduced overall resolution. Summary of the Invention

[0003] According to an embodiment, a method for maintaining pixel intensity implemented by an image sensor processor includes: receiving, by an image processor, a numerical value indicating a corresponding pixel intensity; determining, by the image processor, whether a least significant part of the received numerical value is equal to a predetermined numerical value; and in response to determining that the least significant part of the received numerical value is equal to the predetermined numerical value, rounding, by the image processor, the received numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence, and if the least significant part of the received numerical value is not equal to the predetermined value, then rounding the received numerical value to a higher or lower value based on the received numerical value; and fusing, by the image processor, the rounded value.

[0004] In some embodiments, rounding the received numerical value of the corresponding pixel intensity to a higher or lower value according to the bit sequence is performed according to a digital sequence.

[0005] In some embodiments, the digital sequence is provided by a random number generator.

[0006] In some embodiments, the digital sequence is a trigger sequence of alternating values, where approximately half of the numerical values are rounded to lower values and approximately half of the numerical values are rounded to higher values.

[0007] In some embodiments, the predetermined numerical value is the least significant value 0.5.

[0008] According to another embodiment, a camera system includes a module that includes an image sensor processor and a memory storing machine-readable instructions communicatively coupled to the image sensor processor. When executed by the image sensor processor, the machine-readable instructions cause the image sensor processor to: receive, by the image processor, a numerical value indicative of a corresponding pixel intensity; determine, by the image processor, whether a least significant portion of the received numerical value is equal to a predetermined numerical value; in response to determining that the least significant portion of the received numerical value is equal to the predetermined numerical value, round, by the image processor, the received numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence, and if the least significant portion of the received numerical value is not equal to the predetermined value, then round the received numerical value to a higher or lower value based on the received numerical value; and fuse, by the image processor, the rounded values.

[0009] In some embodiments, rounding the received numerical value of the corresponding pixel intensity to a higher or lower value according to the bit sequence is performed according to a digital sequence.

[0010] In some embodiments, the digital sequence is provided by a random number generator.

[0011] In some embodiments, the digital sequence is a trigger sequence of alternating values, where approximately half of the numerical values are rounded to lower values and approximately half of the numerical values are rounded to higher values.

[0012] In some embodiments, the predetermined numerical value is the least significant value 0.5.

[0013] According to another embodiment, a method implemented by an image sensor processor for maintaining pixel intensity includes: receiving, by the image processor, a set of numerical values indicative of corresponding pixel intensities read from an image sensor; fusing, by the image processor, by a method that includes rounding the set of numerical values according to a rounding algorithm, where the method of rounding the set of numerical values according to the rounding algorithm further includes: determining, by the image processor, for each numerical value whether a least significant portion of each numerical value is equal to a predetermined numerical value; and in response to determining that the least significant portion of each numerical value is equal to the predetermined numerical value, rounding, by the image processor, each numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence, and if the least significant portion of the numerical value is not equal to the predetermined value, then rounding the numerical value to a higher or lower value based on the numerical value; averaging the rounded numerical values to produce a fused value; and storing, by the image processor, the fused value.

[0014] In some embodiments, rounding each received numerical value of the corresponding pixel intensity to a higher or lower value according to the bit sequence is performed according to a digital sequence.

[0015] In some embodiments, the digital sequence is provided by a random number generator.

[0016] In some embodiments, the digital sequence is a trigger sequence of alternating values, where approximately half of the numerical values are rounded to lower values and approximately half of the numerical values are rounded to higher values.

[0017] In some embodiments, the predetermined numerical value is the least significant bit 0.5. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1A A diagram showing a method of fusing pixel intensities from a 2×2 pixel array into a single superpixel according to an embodiment.

[0019] Figure 1B A diagram showing fusing pixel intensities from four 2×2 pixel arrays into a single 2×2 pixel array according to an embodiment, and corresponding numerical values indicating the pixel intensities of each pixel.

[0020] Figure 2 A diagram depicting fusing a uniform pixel distribution using a rounding algorithm according to an embodiment.

[0021] Figure 3 A diagram showing the grayscale of various pixel intensities using various methods according to an embodiment.

[0022] Figure 4 A diagram depicting a computing system implementing an embodiment of the present invention according to an embodiment.

[0023] Figure 5 A flowchart illustrating a method of maintaining a weighted average pixel intensity when fusing an N×N pixel array according to an embodiment.

[0024] Figures 6A-6F A diagram showing curves at various grayscale intensities, each curve illustrating the average error as a function of automatic white balance, with the first curve using a conventional method and the second curve using an embodiment of the present invention.

[0025] Figure 7 A diagram showing the rounding of floating-point numerical values using an embodiment of the present invention according to an embodiment.

[0026] Figure 8A 、 8B A diagram illustrating fusing grayscale pixel intensities using a conventional method versus using an embodiment of the present invention according to an embodiment. DETAILED DESCRIPTION

[0027] Figure 1AFIG. 100 shows a method of fusing pixel intensities from a 2×2 pixel array 102 into a single superpixel 104. The resulting single pixel 104 has a numerical value represented as P, which is a weighted average of the individual pixel intensities of the 2×2 pixel array 102 (i.e., upper left (UL), upper right (UR), lower left (BL), and lower right (BR)). For example, the weighted average is represented as

[0028]

[0029] where w xy is the weight of pixel xy and p xy is the intensity of pixel xy. When the numerical value of the pixel intensity is rounded up or down, the weighted average pixel intensity of the 2×2 pixel array 102 may not be consistent with the weighted average pixel intensity of the fused single pixel 104. For example, Figure 1B FIG. 101 shows a method of fusing pixel intensities from four 2×2 pixel arrays 106 of a single 4×4 pixel array 105 into a single 2×2 pixel array 108, and the corresponding numerical values indicating the pixel intensities of each pixel. When averaging each 2×2 pixel array 106, the 2×2 pixel array 108 shows that the weighted average of each 2×2 pixel array 106 is 0.5. When using a conventional method to average, when the pixel intensity is greater than or equal to 0.5, the pixel intensity is rounded up, and when the pixel intensity is less than 0.5, the pixel intensity is rounded down. The conventional method cannot accurately maintain the weighted average pixel intensity throughout the fusion process. For example, using Figure 1B the conventional method in, by rounding the pixel intensity of 0.5, all values below 0.5 are rounded down, while 0.5 and higher values are rounded up, and the resulting rounded value is 1. The 2×2 pixel 110 shows that the resulting rounded value of each pixel is 1. The difference in terms of the resulting weighted average of the 2×2 pixel 110 (where each pixel intensity is 1) is different from the weighted average pixel intensity of 0.5 for each 2×2 pixel 106.

[0030] Embodiments of the present invention solve this difference between the pre-fusion pixel intensity and the post-fusion pixel intensity by alternately rounding each pixel intensity of a specific numerical value (e.g., 0.5) between rounding up and rounding down, thereby maintaining the average weighted pixel intensity of the initial N×N pixels. Figure 2FIG. 200 depicts an initial uniform distribution graph 202 of the percentage of each pixel relative to the total pixels and various pixel intensities, including a first pixel 204 with a pixel intensity of 0, a second pixel 206 with a pixel intensity of 0.25, a third pixel 208 with a pixel intensity of 0.5, a fourth pixel 210 with a pixel intensity of 0.75, and a fifth pixel 212 with a pixel intensity of 1. In an embodiment, each pixel 204-212 may represent a larger group of pixels, such as a group of 32 pixels, 64 pixels, etc. When transformation is performed using a conventional method, the resulting graph 214 shows a weighted average pixel intensity of 0.6, which differs from the weighted average pixel intensity of graph 202 by 0.1. This is because the values of pixels below 0.5 are rounded down to 0, while the values of pixels greater than or equal to 0.5 are rounded up to 1. The conventional method causes the third pixel 208 with a pixel intensity of 0.5 to be rounded up to 1.

[0031] However, using an embodiment of the present invention, FIG. 216 shows how the weighted value of the third pixel 208 can be split, with half of the value rounded up to one and the other half rounded down to zero. The weighted average from FIG. 216 resulting from using an embodiment of the present invention results in a value (0.5) equal to the weighted average of the pixel intensities of pixels 202-212 shown in FIG. 202. In an embodiment, the number of pixels equal to 0.5 can be any number greater than zero (e.g., four pixels, nine pixels, sixteen pixels, fifteen pixels, thirty-two pixels, etc.), and any type of number sequence (such as a sequence generated using a random number generator, a trigger, etc.) is used to round the pixel intensity up to 1 or down to 0. For example, there can be four pixels, each with a corresponding pixel intensity of 0.5, where two pixel intensities are rounded down to zero and the other two pixel intensities are rounded up to one. Continuing with this example, the number sequence can include a trigger sequence where the first pixel intensity of 0.5 is rounded down to zero, the second pixel intensity of 0.5 is rounded up to one, the third pixel intensity of 0.5 is rounded down to zero, and the fourth pixel intensity of 0.5 is rounded up to 1. Alternatively, an embodiment of the present invention can employ a random number generator where each pixel intensity of 0.5 is randomly rounded up to 1 or down to 0 such that half of the pixel intensities are rounded up to 1 and the other half are rounded down to 0.

[0032] Figure 3The grayscale levels of various pixel intensities after fusion are shown. Ground truth 302 illustrates the true value of the pixel intensity after fusion, where the pixel intensity of each pixel of the pixel array after fusion is consistent with the pixel array before fusion. Traditional rounding 304 illustrates the pixel intensity of various grayscale values when the N×N pixel array is fused using the conventional method, as discussed above with reference to FIG. 1. New method 306 illustrates the fusion of the N×N pixel array using an embodiment of the present invention. Compared with traditional rounding 304, the new method 306 is closer to the pixel intensity consistent with the ground truth 302.

[0033] Figure 4 A module 402 implementing an embodiment of the present invention is depicted. In an embodiment, the module 402 includes a memory 404, a processor 412, an image sensor 414, and an output 416. In an embodiment, the module 402 may not include each of the components 404, 412, 414, but is electrically connected to each component. The memory 404 may be communicatively coupled to the processor 412, and the memory 404 also includes an image buffer, a rounding algorithm 406 digitally represented as machine-readable instructions, which when executed by the processor 412 causes the processor 412 to implement the rounding algorithm 406, which includes receiving pre-fused pixel data 408, executing the rounding algorithm 406 on the pre-fused data 408, and outputting fused pixel data 410. For example, the pre-fused pixel data 408 includes pixels corresponding to the N×N pixel array (e.g., the region of interest 802, reference 410). Figure 8A , B) corresponding pixel intensity (e.g., the pixel intensity shown in Figure 804, reference Figure 8A , B), and the fused pixel data 410 (e.g., the pixel intensities shown in graph 808) include pixel intensities. In addition, the rounding algorithm 406 rounds half or approximately half of the pixel intensities (decimal value 0.5) from the pre-fused pixel data 408 to zero, while rounding the other half or approximately half of the pixel intensities to one. The rounding algorithm 406 can be any algorithm, including a random number generator, a trigger, etc. In some embodiments, the rounding algorithm 406 can be hardware implemented.

[0034] The memory 404 also includes an automatic white balance (AWB) routine 409, a color restoration block 410, and a compression block 411. The AWB routine 409 may perform automatic white balance on the fused pixel data 408 by adjusting pixel intensities to present more natural colors at the output 416. The compression block 411 may compress the fused pixel data to remove background noise, optimize memory allocation, etc.

[0035] The processor 412 receives the pre - fused pixel data 408 from the image sensor 414 (e.g., after the image sensor captures an image), and then may store the pre - fused pixel data 408 in the memory 404. The processor 412 may transmit the fused pixel data 410 to the output 416 for display, e.g., within a graphical user interface of a client device such as a smart device, a camera, etc.

[0036] In an embodiment, bit - width reduction may occur within the module 402 (such as the processor 412). For example, the image sensor 414 may include a 10 - bit analog - to - digital converter (ADC), and the converted pixel intensity is scaled by subtracting the full - dark value. The 10 - bit representation of the pixel intensity may be stored within the pre - fused pixel data 408 and then rounded (in accordance with embodiments of the present invention) to reduce it to an 8 - bit value, averaged, and then stored in the fused pixel data.

[0037] Figure 5 is a flowchart illustrating a method 500 for maintaining a weighted - average pixel intensity when fusing an N×N pixel array. The method 500 may be implemented by a processor (such as the processor 412) communicatively coupled to a memory (e.g., the memory 404) representing machine - readable instructions that, when executed by the processor, cause the processor to implement a rounding algorithm (such as the rounding algorithm 406 stored in the memory 404). The method 500 includes receiving (502) a numerical value indicating a corresponding pixel intensity read from an image sensor (such as the image sensor 414). In one example of block 502, the processor receives the pixel intensities of the corresponding pixels of the N×N pixel array captured by the image sensor 414 and stores the received pixel intensities in the memory (e.g., in the pre - fused pixel data 408 of the memory 404). In one example of block 502, the received pixel intensities may be similar to the pixel intensities represented as digits, as shown in FIG. 804( Figure 8A , B).

[0038] The method 500 further includes determining (504) whether the least - significant portion of the received numerical value of the corresponding pixel intensity is equal to a predetermined value (e.g., 0.5, etc.). In one example of block 504, the processor selects the pixel intensities equal to 0.5 and then counts their total. For example, the processor may count the number of pixel intensities having a least - significant value of 0.5.

[0039] If the processor determines that the least significant portion is equal to a predetermined value (Decision: "Yes"), then method 500 includes rounding (506) the received value of the corresponding pixel intensity to a lower or higher value. In one example of block 506, whether to round to a lower or higher value can depend on the bit sequence. In one example of block 506, the rounding can be performed according to a numerical sequence (e.g., a random number generator, a trigger, etc.). In one example of block 506, the processor rounds the least significant portion of the received value that is less than 0.5 to the least significant valid value of zero and rounds the least significant valid value that is greater than 0.5 to one.

[0040] However, if the processor determines that the least significant portion is not equal to the predetermined value (Decision: "No"), then method 500 proceeds to round (508) the received value to a higher or lower value based on the received value. Method 500 also includes fusing (510) the rounded values from blocks 506, 508 for the corresponding pixel intensity. In one example of block 510, the processor may store the fused value in a memory (e.g., the fused pixel data 410 of memory 402). In one example of block 510, the processor may call an output 416 to display the fused pixel data (e.g., the fused pixel data 410) within a graphical user interface of a client device (e.g., a smart phone, a tablet, a laptop, a camera, etc.).

[0041] Figure 6A -F each show two curves graphing the mean error as a function of the auto white balance gain. The first curve 602(1)-(6) uses a conventional method, while the second curve 604(1)-(6) uses an embodiment of the present invention. For example, curves 602(1)–(6) and 604(1)–(6) show the absolute errors of the mean pixel intensities of 71.59, 69.84, 67.92, 66.15, 65.06, and 64.02, respectively. Curves 602(1)-(6) and 604(1)-(6) use real-world data to describe the difference between the conventional method and the rounding algorithm. Curve 602(1)-(6) graphs the relationship between the mean error and the auto white balance (AWB) gain at various intensities using the conventional method as described with reference to FIG. 1. For example, the mean error is the difference between the weighted mean pixel intensities before and after rounding using the conventional method. Each curve 602(1)-(6) shows the mean error after using the conventional method, with an absolute value of 0.1 up to an auto white balance gain of approximately 2000, and then an absolute error of approximately 0.05.

[0042] Graph 604(1)–(6) illustrates the relationship of the average error value at the corresponding intensity using an embodiment of the present invention with respect to the AWB gain, showing a significantly lower average error compared to the conventional method. Graphs 604(1)-(6) show that the embodiment of the present invention significantly compresses the average value error compared to the conventional method shown in Graphs 602(1)-(6). For example, for most of Graphs 602(1)-(6), the absolute value of the average value error is approximately zero.

[0043] Figure 7 Table 700 illustrates exemplary data of being expressed as various floating-point numbers (such as the pixel data 407 before fusion), and the corresponding output of a counter that counts the number of values with the least significant bit being 0.5; the output of the rounding algorithm; and the final output value (e.g., for a value with a fractional part of 0.5, either 0 or 1). For example, the floating-point values are 65.25, 65.75, 66.5, 64.5, 65.00, 66.00, 65.00, 67.75, 67.5, 67.25, 64.5, 65.00, 65.25, 65.70, 64.5, 65.75, 64.25, 63.5, which can form the pixel data 407 before fusion. The rounding algorithm (e.g., rounding algorithm 406) can start counting the number of floating-point numbers with the fractional part equal to 0.5 to the right of the decimal point, including 66.5, 64.5, 67.5, 64.5, 64.5, and 63.5. Table 700 also shows the corresponding rounded values of each floating-point number with a fractional value of 0.5. For example, the rounded values of the output are shown as "0" or "1" in the output s[n] row. The rounded values are determined according to the rounding algorithm, implementing a digital sequence (i.e., s[n], as Figure 7 shown), such as a random number generator, a flip-flop, etc. The last output row shows the rounded values of the floating-point numbers in the final form, such as 66, 65, 67, 65, 64, and 64.

[0044] Embodiments of the present invention are not limited to floating-point numbers, but include scaled integer rounding truncation as discussed in reference Figure 4 and floating-point numbers with mantissa and exponent, etc. After truncating the following set of values {50.55, 40.65, 70.55} to the first N decimal places (such as N = 1) to obtain the following set of truncated values {50.5, 40.6, 70.5}, embodiments of the present invention can be applied. Applying embodiments of the present invention to the truncated set results in the set {50, 41, 71}, whose weighted average is 54; while applying the conventional method to the set of truncated values results in the set {51, 41, 71}, whose weighted average is 54.3.

[0045] Figure 8A illustrates from various gray intensities such asFigure 3 FIG. 800 of the region of interest (ROI) 802 (shown in). FIG. 800 shows rounding all fractional values of pixel intensity of 0.5 to one using a conventional method. Curve graph 804 shows the fused pixel intensity of ROI 802 and the percentage of the fused pixel intensity. The pixel intensity is represented as scaled digital numbers (DN), with each DN having 10 bits (i.e., 0 DN–1023 DN). The weighted average pixel intensity is 130.16. Curve graph 806 shows the fused pixel intensity and the percentage of the fused intensity after the conventional method is implemented when rounding to 10 bits during truncation. Before and after fusion, the data is 10 bits. ROI 802 has 2N×2N pixels, and each 2×2 pixel is fused into a superpixel. Thus, the intermediate fusion result is 12 bits and is then rounded to 10 bits during truncation. The weighted average pixel intensity is 130.28. The percentage of pixel intensity values rounded up is 49.015% (including 24.546% of pixels with a fractional value of 0.5). The percentage of unchanged pixel intensity is 25.951%. The percentage of pixel intensity rounded down is 25.034%. The difference between the weighted average pixel intensities before and after the conventional method is implemented is 0.12.

[0046] Figure 8B FIG. 801 illustrates ROI 802 from various gray intensities. FIG. 801 shows using an embodiment of the present invention to round all fractional values of pixel intensity of 0.5 to one and approximately the other half to zero. Curve graph 804 shows the fused pixel intensity of ROI 802 and the percentage of the fused pixel intensity, where the weighted average pixel intensity is 130.16. Curve graph 808 shows the fused pixel intensity and the percentage of the fused intensity after an embodiment of the present invention (such as rounding algorithm 406) is implemented. After the embodiment of the present invention is implemented, the weighted average pixel intensity of curve graph 808 is 130.16, which is equal to the weighted average of the pixel intensities of curve graph 804 before rounding. The percentage of pixel intensity values rounded up is 36.747% (including 12.268% of pixels with a fractional value of 0.5). The percentage of unchanged pixel intensity is 25.951%. The percentage of pixel intensity rounded down is 37.302% (including 12.268% of pixels with a fractional value of 0.5). Thus, the percentage of pixel intensity with a fractional value of 0.5 rounded to one is approximately equal to the percentage of pixel intensity rounded down to zero. The difference between the weighted average pixel intensities before the conventional rounding method is implemented is 0.

[0047] Changes may be made to the above methods and systems without departing from the scope of the disclosure. It should thus be noted that the content contained in the above description or shown in the drawings should be construed as illustrative and not restrictive. The following claims are intended to cover all general and specific features described herein, as well as all statements of the scope of the methods and systems, which, as to language, may be said to fall therebetween.

[0048] Combination of features

[0049] Specifically consider the following embodiments, and any combination of such embodiments that are compatible with each other:

[0050] (A) A method implemented by an image sensor processor for maintaining pixel intensity, including: receiving, by the image processor, a numerical value indicating a corresponding pixel intensity read from an image sensor; determining, by the image processor, whether the least significant part of the received numerical value is equal to a predetermined numerical value; and in response to determining that the least significant part of the received numerical value is equal to the predetermined numerical value, rounding, by the image processor, the received numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence, and if the least significant part of the received numerical value is not equal to the predetermined value, then rounding the received numerical value to a higher or lower value based on the received numerical value; and storing, by the image processor, the rounded value.

[0051] (B) In the embodiment represented by (A), wherein rounding the received numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence is performed according to a digital sequence.

[0052] (C) In the embodiment represented by (A) or (B), wherein the digital sequence is provided by a random number generator.

[0053] (D) In any one of the embodiments represented by (A)-(C), wherein the digital sequence is a trigger sequence of alternating values, wherein approximately half of the numerical values are rounded to lower values and approximately half of the numerical values are rounded to higher values.

[0054] (E) In any one of the embodiments represented by (A)-(D), wherein the predetermined numerical value is the least significant value 0.5.

[0055] (F) A camera system, comprising a module that includes an image sensor processor and a memory storing machine-readable instructions communicatively coupled to the image sensor processor. When executed by the image sensor processor, these machine-readable instructions cause the image sensor processor to: receive, by the image processor, a numerical value indicating a corresponding pixel intensity; determine, by the image processor, whether the least significant part of the received numerical value is equal to a predetermined numerical value; in response to determining that the least significant part of the received numerical value is equal to the predetermined numerical value, round, by the image processor, the received numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence, and if the least significant part of the received numerical value is not equal to the predetermined value, then round the received numerical value to a higher or lower value based on the received numerical value; and store, by the image processor, the rounded value.

[0056] (G) In the embodiment represented by (F), wherein rounding the received numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence is performed according to a numerical order.

[0057] (H) In the embodiment represented by (F) or (G), wherein the numerical sequence is provided by a random number generator.

[0058] (I) In any one of the embodiments represented by (F)-(H), wherein the numerical sequence is a trigger sequence of alternating values, wherein approximately half of the numerical values are rounded to lower values and approximately half of the numerical values are rounded to higher values.

[0059] (J) In any one of the embodiments represented by (F)-(I), wherein the predetermined numerical value is the least significant value 0.5.

[0060] (K) A method implemented by an image sensor processor for maintaining pixel intensity, comprising: receiving, by the image processor, a set of numerical values indicating corresponding pixel intensities read from an image sensor; fusing, by the image processor, by a method including rounding the set of numerical values according to a rounding algorithm, and the method of rounding the set of numerical values according to the rounding algorithm further includes: determining, by the image processor, for each numerical value whether the least significant part of each numerical value is equal to a predetermined numerical value; and in response to determining that the least significant part of each numerical value is equal to the predetermined numerical value, rounding, by the image processor, each numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence, and if the least significant part of the numerical value is not equal to the predetermined value, then rounding the numerical value to a higher or lower value based on the numerical value; and averaging the rounded numerical values to produce a fused value; and storing, by the image processor, the fused value.

[0061] (L) In the embodiment represented by (K), wherein rounding each received numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence is performed according to a numerical order.

[0062] (M) In the embodiment represented by (K) or (L), the digital sequence is provided by a random number generator.

[0063] (N) In any of the embodiments represented by (K)-(M), the digital sequence is a trigger sequence of alternating values, where approximately half of the numerical values are rounded down and approximately half of the numerical values are rounded up.

[0064] (O) In any of the embodiments represented by (K)-(N), the predetermined numerical value is the least significant bit 0.5.

Claims

1. A method for maintaining pixel intensity implemented by an image sensor processor, comprising: Receiving, by an image processor, a numerical value indicating a corresponding pixel intensity read from an image sensor; Determining, by the image processor, whether a least significant portion of the received numerical value is equal to a predetermined numerical value; And In response to determining that the least significant portion of the received numerical value is equal to the predetermined numerical value, rounding, by the image processor, the received numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence, and if the least significant portion of the received numerical value is not equal to the predetermined numerical value, then rounding the received numerical value to a higher or lower value based on the received numerical value; And Storing, by the image processor, the rounded value.

2. The method according to claim 1, wherein rounding the received numerical value of the corresponding pixel intensity to a higher or lower value according to the bit sequence is performed according to a digital sequence.

3. The method according to claim 2, wherein the digital sequence is provided by a random number generator.

4. The method according to claim 2, wherein the digital sequence is a flip-flop sequence of alternating values, wherein half of the numerical values are rounded to lower values and the other half of the numerical values are rounded to higher values.

5. The method according to claim 1, wherein the predetermined numerical value is a least significant value of 0.

5.

6. A camera system, comprising a module, the module including an image sensor processor, a memory storing machine-readable instructions communicatively coupled to the image sensor processor, the machine-readable instructions, when executed by the image sensor processor, causing the image sensor processor to: Receiving, by an image processor, a numerical value indicating a corresponding pixel intensity read from an image sensor; Determining, by the image processor, whether a least significant portion of the received numerical value is equal to a predetermined numerical value; And In response to determining that the least significant portion of the received numerical value is equal to the predetermined numerical value, rounding, by the image processor, the received numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence, and if the least significant portion of the received numerical value is not equal to the predetermined numerical value, then rounding the received numerical value to a higher or lower value based on the received numerical value; And Storing, by the image processor, the rounded value.

7. The camera system according to claim 6, wherein rounding the received numerical value of the corresponding pixel intensity to a higher or lower value according to the bit sequence is performed according to a digital sequence.

8. The camera system according to claim 7, wherein the digital sequence is provided by a random number generator.

9. The camera system according to claim 7, wherein the digital sequence is a flip-flop sequence of alternating values, wherein half of the numerical values are rounded to lower values and the other half of the numerical values are rounded to higher values.

10. The camera system according to claim 6, wherein the predetermined numerical value is a least significant value of 0.

5.

11. A method for maintaining pixel intensity implemented by an image sensor processor, comprising: A set of numerical values indicating corresponding pixel intensities read from an image sensor is received by an image processor; Fusion is performed by the image processor by a method including rounding the set of numerical values according to a rounding algorithm, and the method of rounding the set of numerical values according to the rounding algorithm further includes: Determining, by the image processor for each numerical value, whether the least significant part of each numerical value is equal to a predetermined numerical value; and In response to determining that the least significant part of each numerical value is equal to the predetermined numerical value, rounding each numerical value of the corresponding pixel intensity to a higher or lower value according to a bit sequence, and if the least significant part of the numerical value is not equal to the predetermined numerical value, then rounding the numerical value to a higher or lower value based on the numerical value; and Averaging the rounded numerical values to produce a fused value; and Storing the fused value by the image processor.

12. The method according to claim 11, wherein rounding each received numerical value of the corresponding pixel intensity to a higher or lower value according to the bit sequence is performed according to a digital sequence.

13. The method according to claim 12, wherein the digital sequence is provided by a random number generator.

14. The method according to claim 12, wherein the digital sequence is a trigger sequence of alternating values, wherein half of the numerical values are rounded to lower values and the other half of the numerical values are rounded to higher values.

15. The method according to claim 11, wherein the predetermined numerical value is the least significant value 0.5.

Images