TW202127857A - White balance adjustment method, image processing device and image processing system - Google Patents

Abstract

A white balance adjustment method, an image processing device and an image processing system are provided. A raw image is received. The raw image is segmented into multiple first image blocks according to brightness information of the raw image. A plurality of local illuminant information respectively corresponding to the first image blocks is generated for each first image block. The raw image is segmented into multiple second image blocks, and mixed illuminant information respectively corresponding to the second image blocks is generated according to at least one of the local illuminant information corresponded to the second image blocks. Pixels values within the second image blocks are respectively adjusted according to the mixed illuminant information of the second image blocks to obtain a white balance image.

Description

White balance adjustment method, image processing device and image processing system

The present invention relates to an image processing method, and more particularly to a white balance adjustment method, an image processing device and an image processing system.

With the development of technology, various smart image capture devices, such as tablet computers, personal digital assistants, and smart phones, have become indispensable tools for modern people. Among them, the high-end smart image capture device is equipped with camera lenses that are comparable to traditional consumer cameras, and can even be replaced. A few high-end models have more pixel and image quality close to digital monocular or provide more advanced The functions and effects of the order.

The photographer may use the image capturing device to shoot at different places or times, so the image capturing device will shoot images under different light source environments. The light source in the shooting environment will directly affect the color of the target in the captured image. In an application scenario, the injured may take images of the wound or the affected area and provide the taken images to the medical unit for diagnosis. If the color information in the image is distorted due to the light source in the shooting environment, it may not be correct based on the captured image. diagnosis. In this regard, the auto white balance (AWB) algorithm used by general image capture devices mainly uses the grayscale content of the scene to adjust the white balance after the image sensor captures the image of the scene. To stably display the color content of the scene. However, in an environment with complex light sources, since different positions on the same subject may be illuminated by different mixed light sources, it is relatively difficult to restore the color information of the subject very accurately. For example, the use of an auxiliary light source (such as a flash) will cause the image after the white balance adjustment to still have local or overall color shift. Therefore, how to avoid image color cast by a better white balance adjustment method is actually one of the directions that those skilled in the art are striving for.

In view of this, the present invention provides a white balance adjustment method, an image processing device, and an image processing system, which can adaptively adjust the white balance of the image according to different image regions in the image, thereby achieving high-quality image output.

The embodiment of the present invention provides a white balance adjustment method. The method includes the following steps. Receive an original image. According to the brightness information of the original image, the original image is divided into a plurality of first image regions. For each first image area, multiple area light source information corresponding to the first image area is generated. The original image is divided into a plurality of second image areas, and a plurality of mixed light source information respectively corresponding to the second image area is generated according to at least one of the area light source information corresponding to the second image area. According to the mixed light source information of the second image area, a plurality of pixel values in the second image area are respectively corrected to obtain a white balance image.

An embodiment of the present invention provides an image processing device, which includes a storage device and a processor. The storage device stores multiple modules. The processor is coupled to the storage device and is configured to execute the module to perform the following steps. Receive an original image. According to the brightness information of the original image, the original image is divided into a plurality of first image regions. For each first image area, multiple area light source information corresponding to the first image area is generated. The original image is divided into a plurality of second image areas, and a plurality of mixed light source information respectively corresponding to the second image area is generated according to at least one of the area light source information corresponding to the second image area. According to the mixed light source information of the second image area, a plurality of pixel values in the second image area are respectively corrected to obtain a white balance image.

An embodiment of the present invention provides an image processing system, which includes an image sensor, a storage device, and a processor. The storage device stores multiple modules. The processor is coupled to the image sensor and the storage device, and is configured to execute the module to perform the following steps. Receive an original image. According to the brightness information of the original image, the original image is divided into a plurality of first image regions. For each first image area, multiple area light source information corresponding to the first image area is generated. The original image is divided into a plurality of second image areas, and a plurality of mixed light source information respectively corresponding to the second image area is generated according to at least one of the area light source information corresponding to the second image area. According to the mixed light source information of the second image area, a plurality of pixel values in the second image area are respectively corrected to obtain a white balance image.

Based on the above, in the embodiment of the present invention, the original image can be divided into a plurality of image areas to perform white balance correction based on different mixed light source information, and the mixed light source information is also generated by estimating different image areas. In this way, the embodiment of the present invention can effectively correct the color shift phenomenon in the original image that is reflected in a variety of ambient light sources, so as to perform more accurate image white balance correction.

In order to make the above-mentioned features and advantages of the present invention more comprehensible, the following specific embodiments are described in detail in conjunction with the accompanying drawings.

Part of the embodiments of the present invention will be described in detail in conjunction with the accompanying drawings. The reference symbols in the following description will be regarded as the same or similar elements when the same symbol appears in different drawings. These embodiments are only a part of the present invention, and do not disclose all the possible implementation modes of the present invention. To be more precise, these embodiments are just examples of methods, devices, and systems within the scope of the patent application of the present invention.

FIG. 1 is a block diagram of an image processing system according to an embodiment of the present invention, but this is only for convenience of description, and is not intended to limit the present invention. First, Figure 1 first introduces all the components and configuration relationships of the image processing system. Detailed functions will be disclosed in conjunction with Figure 2.

1, the image processing system 100 includes an image sensor 110 and an image processing device 120, where the image processing device 120 includes a storage device 122 and a processor 124. In this embodiment, the image processing system 100 may be an image capturing device that integrates the image sensor 110 and the image processing device 120 into a single device (all-in-one), such as a digital camera with a lens or a single-lens camera. , Digital cameras, smart phones, tablets, etc. In another embodiment, the image processing device 120 may be an electronic device with image processing functions such as a personal computer, a notebook computer, a smart phone, a tablet computer, etc., and may be wired or wireless via a communication interface (not shown) The image captured by the image sensor 110 is received.

In this embodiment, the image sensor 110 is an imaging lens including a lens and a photosensitive element. The photosensitive element is used to sense the intensity of light entering the lens to generate an image. The photosensitive element may be, for example, a charge coupled device (CCD), a complementary metal-oxide semiconductor (CMOS) element, or other elements. The image captured by the lens will be imaged on the sensing element and converted into a digital signal for output to the processor 124.

The storage device 122 is used to store images, program codes and other data, and it can be, for example, any type of fixed or removable random access memory (RAM), read-only memory (ROM). ), flash memory, hard disk or other similar devices, integrated circuits and combinations thereof.

The processor 124 is used to control the actions between the components of the image processing system 100, and it can be, for example, a central processing unit (CPU), a graphics processing unit (GPU), or other programmable ones. General purpose or special purpose microprocessor (microprocessor), digital signal processor (digital signal processor, DSP), image signal processor (image signal processor, ISP), programmable controller, special application integrated circuit (application specific integrated circuits (ASIC), programmable logic device (PLD) or other similar devices or a combination of these devices.

The following is an example to illustrate the detailed steps of the method for adjusting the white balance of the image processing system 100. In the following embodiments, the image processing system 100 will be implemented as an image capture device for illustration, and the processor 124 can be implemented by an image signal processor, and its role is to process the output signal of the front-end image sensor 110 , In order to restore the details of the scene under different conditions.

FIG. 2 is a flowchart of a method for adjusting white balance according to an embodiment of the present invention. The method of this embodiment is applicable to the image processing system 100 in FIG.

Please refer to FIG. 1 and FIG. 2 at the same time. First, in step S201, the processor 124 of the image processing device 120 receives an original image. In this embodiment, since the processor 124 is implemented as an image signal processor, the first image can be one of the image sequences captured by the processor 124 that it receives from the image sensor 110 in real time. Images. However, in another embodiment, the first image may be one of the images in the image sequence captured by the external image sensor 110 obtained by the processor 124 directly or indirectly through other devices.

In step S202, the processor 124 divides the original image into a plurality of first image regions according to the brightness information of the original image. In detail, the original image is composed of a plurality of pixels arranged in an array, and the pixel value of each pixel includes a plurality of pixel components corresponding to a plurality of color space components. Generally speaking, the pixel components generated by the image sensor 110 are the R component corresponding to the red channel, the G component corresponding to the green channel, and the B component corresponding to the blue channel. The processor 124 converts the RGB components of each pixel into the brightness value of each pixel. The brightness value is, for example, the brightness component in the YCbCr color space (ie Y component), the brightness component in the YUV color space (ie Y component), or The brightness component (that is, the V component) in the HSV color space is not limited by the present invention.

In one embodiment, the processor 124 may divide the pixels in the original image into a plurality of first image regions according to whether the brightness values of the pixels in the original image are greater than a threshold value. In detail, the processor 124 may determine whether the brightness values of these pixels are greater than a critical value one by one, and divide the connected pixels with the brightness values in the same brightness interval into a first image area. In other words, multiple connected pixels with brightness values greater than the threshold will be divided into the same first image area. In the same way, a plurality of pixels that are connected to each other and whose brightness value is not greater than the critical value will also be divided into the same first image area. In other words, the first image areas in the original image can be divided into two categories, one category is the first image area with a brightness value greater than the threshold, and the other category is the first image area with the brightness value not greater than the threshold. Moreover, the pixels in the same first image area have connectivity. For example, the original image may include two first image regions with a brightness value greater than a threshold value and three first image regions with a brightness value not greater than the threshold value.

It should be noted that, in one embodiment, the threshold for dividing the first image area can be determined based on the overall brightness information of the original image. For example, the processor 124 may first find the maximum brightness value and the minimum brightness value in the original image, and take the average value of the maximum brightness value and the minimum brightness value as the critical value for dividing the first image area. In addition, in one embodiment, the number of thresholds used to divide the first image area may be more than one.

For example, FIG. 3 is a schematic diagram of dividing an original image into a plurality of first image regions according to an embodiment of the invention. Please refer to FIG. 3. For the convenience of description, FIG. 3 takes the original image Img-raw including 8*6 pixels as an example for description. However, those skilled in the art can implement the same steps and methods on original images of different image sizes. In FIG. 3, the numerical value at each pixel position represents the brightness value of each pixel in. The processor 124 can obtain the maximum brightness value of "200" and the minimum brightness value of "30" according to the brightness value of each pixel in the original image Img-raw, and calculate the brightness threshold value of "115". Then, the processor 124 judges one by one whether the brightness value of each pixel is greater than the brightness threshold value '115'. For example, the brightness value '30' of the pixel P11 is not greater than the brightness threshold value '115', but the brightness value of the pixel P26 is greater than the brightness threshold value '115'. Based on this, the processor 124 can divide the original image Img-raw into two first image regions Z1 and Z2. In the example of FIG. 3, the first image area Z1 includes 22 pixels (for example, the pixel P11), and the first image area Z2 includes 26 pixels (for example, the pixel P26).

Returning to the flow of FIG. 2, in step S203, the processor 124 generates a plurality of area light source information corresponding to the first image area for each first image area. That is, the processor 124 independently estimates the corresponding area light source information for each first image area. In one embodiment, the area light source information may include color temperature values. In one embodiment, the area light source information may include the ratio of RGB components corresponding to the RGB channels (for example, the light intensity ratio of the RGB three-color light). For example, the area light source information of a certain first image area may be (0.5, 0.4, 0.2) corresponding to the RGB channels. In one embodiment, the processor 124 may generate the area light source information of the first image area according to the pixel information in at least one sub-area of each first image area. There may be more than one sub-areas, and the size can be set according to actual applications.

In step S204, the processor 124 divides the original image into a plurality of second image regions, and generates a plurality of mixed light source information corresponding to the second image region according to at least one of the regional light source information corresponding to the second image region. In detail, assuming that the original image includes (M*P)*(N*Q) pixels, the processor 124 divides the original image into M*N second image regions, and the second image regions each include P*Q pixels. Pixels. The size and number of the second image area can be set according to actual requirements, and the present invention is not limited thereto. It can be seen that a single second image area may include more than one pixel in the first image area. If a single second image area includes pixels in different first image areas, the second image area can correspond to more than two pieces of area light source information. If a single second image area includes pixels in a single first image area, the second image area will correspond to a piece of area light source information. For each second image area, the processor 124 may generate mixed light source information according to one or more pieces of area light source information corresponding to each of the second image areas. When a certain second image area corresponds to multiple pieces of area light source information (that is, different ambient light sources are involved), the processor 124 may perform statistical calculations or weighting processing on the multiple pieces of area light source information to generate mixed light source information for the second image area . When a certain second image area only corresponds to a piece of area light source information (that is, a single ambient light source is involved), the processor 124 can directly use the area light source information as the mixed light source information of the second image area.

In step S205, the processor 124 respectively corrects a plurality of pixel values in the second image area according to the mixed light source information of the second image area to obtain a white balance image. In one embodiment, the processor 124 may directly adjust a plurality of pixel values in the second image area according to the mixed light source information to obtain a white balance image. In one embodiment, the processor 124 may determine white balance gain parameters (such as R channel gain parameters, G channel gain parameters, and B channel gain parameters) according to the mixed light source information, and adjust the white balance gain parameters in the second image area according to the white balance gain parameters. Multiple pixel values to obtain a white balance image. More specifically, the RGB components of the pixels in a certain second image area can be adjusted based on the mixed light source information to achieve the purpose of white balance adjustment. Therefore, in this embodiment, since the original image is adjusted for the regionalized white balance in consideration of different light sources, the white balance image generated and output by the processor 124 can achieve a more accurate white balance. Performance.

For a clearer description, FIG. 4 is a flowchart of a white balance adjustment method according to an embodiment of the present invention. Hereinafter, FIG. 4 will be used to describe an implementation detail of the above process. The method of this embodiment is applicable to the image processing system 100 in FIG.

1 and 4 at the same time, in step S401, the processor 124 receives an original image. In step S402, the processor 124 divides the original image into a plurality of first image regions according to the brightness information of the original image. Then, in this embodiment, the processor 124 will generate the area light source information of the first image area according to at least one sub-area in each first image area. It should be noted that the above method of taking the sub-regions will be different depending on whether the shape of the first image region is rectangular.

In step S403, the processor 124 determines whether the first image area is rectangular. In this embodiment, if one of the first image areas is rectangular (Yes in step S403), in step S404, the processor 124 divides one of the first image areas into a plurality of second sub-regions. The area further generates a plurality of corresponding reference light source information according to the second sub-areas, and generates one of the area light source information of one of the first image areas by counting the reference light source information. Here, the number of these second sub-regions is more than one. On the other hand, if one of the first image areas is not a rectangle (No in step S403), in step S405, the processor 124 extracts the first sub-area in the sub-area from one of the first image areas, and according to The first sub-region generates one of the regional light source information of one of the first image regions. Here, the number of the first sub-region is one.

In this embodiment, the step S404 of generating an area light source for the rectangular first image area can be implemented as sub-steps S4041 to S4043. In step S4041, the processor 124 divides the first image area into a plurality of second sub-areas. The size of these second sub-regions can be w*h, which can be set according to actual applications. In other words, the processor 124 may divide the rectangular first image area into a plurality of second sub-areas.

In step S4042, the processor 124 uses a neural network model to respectively predict a plurality of reference light source information according to the second sub-region. In detail, the processor 124 inputs these second sub-regions one by one to a trained neural network model to predict the respective reference light source information of the second sub-regions respectively. For example, a rectangular first image area can be divided into n second sub-areas, and the processor will use the trained neural network model to respectively predict n pieces of reference light source information. The neural network model can be a machine learning model constructed in advance for machine learning based on training data, and it can be stored in the storage device 122. In other words, the model parameters of the neural network model (such as the number of neural network layers and the weight of each neural network layer, etc.) can be determined through pre-training and stored in the storage device 122. This neural network model is, for example, a recurrent neural network (RNN) model. In the neural network model training phase, multiple training image data of size w*h are assigned real light source information, and a neural network model is constructed based on the pixel information of the training image data and the corresponding real light source information. However, in other embodiments, the processor 124 may also use other light source estimation algorithms to estimate the reference light source information based on the pixel information in the second sub-regions.

In step S4043, the processor 124 generates area light source information by counting the reference light source information of the second sub-area. In one embodiment, the processor 124 may calculate the average value of the reference light source information to generate the area light source information. For example, the processor 124 may average the n R light intensity ratios of the n second sub-regions to obtain the R light intensity ratio in the regional light source information. By analogy, the processor 124 can obtain the area light source information (that is, the RGB light intensity ratio) through averaging.

In addition, in this embodiment, the step S405 of generating an area light source for the non-rectangular first image area can be implemented as sub-steps S4051 to S4052. In step S4051, the processor 124 extracts a first sub-region from the first image region. For example, FIG. 5 is a schematic diagram of taking a first sub-region from a first image block according to an embodiment of the invention. Please refer to FIG. 5, following the original image Img-raw of FIG. 3, the original image Img-raw can be divided into first image regions Z1 and Z2 based on brightness information. The processor 124 determines that the first image area Z1 is not a rectangular area, and extracts the 2*3 first sub-area B1 from the first image area Z1. Similarly, the processor 124 determines that the first image area Z2 is not a rectangular area, and extracts the 2*3 first sub-area B2 from the first image area Z2. However, FIG. 5 is only an example, and the processor 124 may also obtain the first sub-areas from other partial areas of the captured first image areas Z1 and Z2, respectively. FIG. 5 is not used to limit the present invention.

In step S4052, the processor 124 uses a neural network model to predict regional light source information based on the first sub-region. For example, the processor 124 may input the first sub-area B1 shown in FIG. 5 to the neural network model to predict the area light source information of the first image area Z1. The operation of step S4052 is similar to the aforementioned step S4042 of using the neural network model to predict the reference light source information. Based on the above, in this embodiment, the area light source information of the first image area is generated based on inputting at least one sub-area into the neural network model for prediction, and at least one sub-area includes w*h pixels, w and h is an integer greater than one, respectively.

Step S403, step S404, and step S405 will be repeatedly executed, so that the processor 124 generates the area light source information of all the first image regions in the original image. After that, in step S406, the processor 124 divides the original image into a plurality of second image regions. Then, the processor 124 will sequentially generate corresponding mixed light source information for the second image area. In this embodiment, the processor 124 determines whether one of the second image areas corresponds to more than two pieces of area light source information. If one of the second image areas corresponds to more than two pieces of area light source information, the processor 124 performs a weighting operation on at least two of the area light source information corresponding to one of the second image areas according to the ratio of the number of pixels to generate a second image One of the mixed light source information of one of the areas. On the contrary, if one of the second image areas does not correspond to more than two pieces of area light source information, the processor 124 generates one of the second image areas according to one of the area light source information corresponding to one of the second image areas One of the mixed light source information.

As shown in FIG. 4, in step S407, the processor 124 determines whether the second image area corresponds to more than two pieces of area light source information. If the second image area corresponds to more than two pieces of area light source information (Yes in step S407), in step S408, the processor 124 performs a weighting operation on the two or more pieces of area light source information corresponding to the second image area to generate a mixed light source News. The aforementioned pixel number ratio is the number of pixels in one of the second image areas corresponding to the first area light source information in the area light source information and one of the second image areas corresponding to the second area light source information in the area light source information The ratio between the number of pixels. If the second image area does not correspond to more than two pieces of area light source information (No in step S407), in step S409, the processor 124 generates mixed light source information according to a piece of area light source information corresponding to the second image area.

For example, FIG. 6A is a schematic diagram of calculating mixed light source information of multiple second image regions according to an embodiment of the present invention. Referring to FIG. 6A, the original image Img-raw will be divided into four second image regions G11, G12, G21, and G22. Following the original image Img-raw of FIGS. 3 and 5, the original image Img-raw can be divided into two first image regions Z1 and Z2 based on brightness information, so that each pixel position has corresponding regional light source information. Referring to FIGS. 5 and 6A at the same time, the processor 124 can predict the area light source information SP1 according to the first sub-area B1, and predict the area light source information SP2 according to the first sub-area B2. Therefore, the pixel P11 in the second image area G11 corresponds to the area light source information SP1 because it belongs to the first image area Z1. The pixel P34 in the second image area G11 corresponds to the area light source information SP2 because it belongs to the first image area Z2.

For the second image area G11, the number of pixels corresponding to the area light source information SP1 is 11 and the number of pixels corresponding to the area light source information SP2 is 1. Therefore, the processor 124 will calculate the mixed light source information of the second image area G11 at a ratio of 11 to 1 pixel number. In detail, the processor 124 multiplies the area light source information SP1 by a weighting factor of 11/12 and multiplies the area light source information SP2 by a weighting factor of 1/12, and adds the above two multiplication results to generate the second image area. G11 mixed light source information.

Similarly, for the second image area G12, the number of pixels corresponding to the area light source information SP1 is 8 and the number of pixels corresponding to the area light source information SP2 is 4. Therefore, the processor 124 calculates the mixed light source information of the second image area G12 at a ratio of 2 to 1 pixel number. In detail, the processor 124 multiplies the area light source information SP1 by a weighting factor of 2/3 and multiplies the area light source information SP2 by a weighting factor of 1/3, and sums the two multiplication results to generate the second image area. G12 mixed light source information. For the second image area G21, the processor 124 calculates the mixed light source information of the second image area G21 at a ratio of 1:3. It should be noted that since the second image area G22 does not correspond to more than two pieces of area light source information, the processor 124 directly uses the area light source information SP2 as the mixed light source information of the second image area G22.

Step S407, step S408, and step S409 may be repeatedly executed, so that the processor 124 generates mixed light source information of all the second image regions in the original image. Next, in step S410, the processor 124 respectively corrects a plurality of pixel values in the second image area according to the mixed light source information of the second image area to obtain a white balance image.

For example, FIGS. 6B and 6C are schematic diagrams of adjusting the pixel value of the original image according to an embodiment of the invention. 6A to 6C, it is assumed that the processor 124 can generate 4 pieces of mixed light source information MP11, MP12, MP21, and MP22 for the second image regions G11, G12, G21, and G22, respectively. The processor 124 adjusts the pixel value in the second image area G11 according to the mixed light source information MP11. Such, the RGB components of the pixels P11 _{(r 11, g 11, b} 11) will be adjusted according to the source information MP11 mixing balance of the RGB components of the image Img-B _{(r '11, g' 11} , b '11) , Assuming that the mixed light source information MP11 is the light intensity ratio of the RGB three-color light (R _MP11, G _MP11, B _MP11 ), the formulas (1)~(3) are calculated as follows. _{r 'g 11 = r 11 -r} 11 * R MP11 formula _{(1)' 11 = g 11} -g 11 * G MP11 formula _{(3) b '11 = b} 11 -b 11 * B MP11 formula (2)

However, equations (1) to (3) are only an example. In other embodiments, the processor 124 may generate white balance corrected pixel values according to the mixed light source information MP11 according to other algorithms.

By analogy, the processor 124 will adjust the pixel values in the second image regions G12, G21, and G22 according to the mixed light source information MP12, MP21, and MP22, respectively. Like, RGB components of the pixel _{(r 16, g 16, b} 16) P16 would be adjusted according to the source information MP12 mixing balance of the RGB components of the image Img-B _{(r '16, g' 16} , b '16) , the RGB components of the pixels P48 _{(r 48, g 48, b} 48) the hybrid light source information based on the adjusted RGB components MP22 _{(r '48, g' 48} , b '48) the white balance of the image Img-B.

In summary, in the embodiment of the present invention, the original image can be divided into a plurality of image areas to perform white balance correction based on different mixed light source information, and the mixed light source information is also generated by estimating different image areas . In this way, the embodiment of the present invention can effectively correct the color shift phenomenon in the original image that is reflected in a variety of ambient light sources, so as to perform more accurate image white balance correction. The white balance adjustment method and the image processing device and system proposed by the present invention can adaptively adjust the white balance of the image for scenes with multiple light sources or complex light sources, thereby achieving high-quality image output.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention. Anyone with ordinary knowledge in the relevant technical field can make some changes and modifications without departing from the spirit and scope of the present invention. The protection scope of the present invention shall be subject to those defined by the attached patent application scope.

100: image processing system 110: Image sensor 120: Image processing device 122: storage device 124: processor Img-raw: raw image Img-raw: White balance image P11, P16, P26, P34, P48: pixels Z1, Z2: the first image area B1, B2: the first sub-area G11, G12, G21, G22: second image area S201～S205, S401～S410: steps

FIG. 1 is a block diagram of an image processing system according to an embodiment of the invention. FIG. 2 is a flowchart of a white balance adjustment method according to an embodiment of the invention. FIG. 3 is a schematic diagram of dividing an original image into a plurality of first image regions according to an embodiment of the invention. FIG. 4 is a flowchart of a white balance adjustment method according to an embodiment of the invention. FIG. 5 is a schematic diagram of taking a sub-region from a first image block according to an embodiment of the invention. FIG. 6A is a schematic diagram of calculating mixed light source information of a plurality of second image regions according to an embodiment of the present invention. 6B and 6C are schematic diagrams of adjusting the pixel value of the original image according to an embodiment of the invention.

S201~S205: steps

Claims (10)

A white balance adjustment method, the method includes: Receive an original image; According to the brightness information of the original image, the original image is divided into a plurality of first image regions: For each of the first image areas, generating a plurality of area light source information corresponding to the first image areas; Dividing the original image into a plurality of second image areas, and generating a plurality of mixed light source information corresponding to the second image areas according to at least one of the area light source information corresponding to the second image areas: and According to the mixed light source information of the second image areas, the pixel values in the second image areas are respectively corrected to obtain a white balance image. According to the white balance adjustment method described in item 1 of the scope of patent application, the step of dividing the original image into the first image regions according to the brightness information of the original image includes: Dividing the pixels in the original image into the first image regions according to whether the brightness values of the pixels in the original image are greater than a critical value, Wherein, for each of the first image areas, the step of generating the area light source information corresponding to the first image areas respectively includes: The area light source information of the first image areas is generated according to the at least one sub-area in each of the first image areas. According to the white balance adjustment method described in item 2 of the scope of patent application, the step of generating the area light source information of the first image areas according to the at least one sub-area in each of the first image areas includes: Respectively determine whether the first image areas are rectangular; and If one of the first image regions is not a rectangle, a first subregion of the at least one subregion is taken out from one of the first image regions, and the first subregions are generated according to the first subregion. One of the area light source information of one of the image areas, wherein the number of the first sub-area is one. According to the white balance adjustment method described in claim 3, the step of generating the area light source information of the first image areas according to the at least one sub-area in each of the first image areas further includes: If one of the first image areas is rectangular, divide one of the first image areas into a plurality of second sub-areas in the at least one sub-areas, and generate a plurality of second sub-areas respectively according to the second sub-areas Reference light source information, and generate one of the area light source information of one of the first image areas by counting the reference light source information, wherein the number of the second sub-areas is more than one. In the white balance adjustment method described in claim 2, wherein the area light source information of the first image areas is generated based on inputting the at least one sub-area into a neural network model for prediction, and the At least one sub-region includes w*h pixels, and w and h are integers greater than one, respectively. According to the white balance adjustment method described in claim 1, wherein the original image is divided into the second image regions, and at least one of the regional light source information corresponding to the second image regions is generated respectively The steps of corresponding to the plurality of mixed light source information of the second image areas include: Determining whether one of the second image areas corresponds to more than two pieces of light source information of the areas; and If one of the second image areas corresponds to more than two pieces of the area light source information, at least two of the area light source information corresponding to one of the second image areas are weighted according to a ratio of the number of pixels One of the mixed light source information of one of the second image regions is generated by operation. For the white balance adjustment method described in item 6 of the scope of patent application, the ratio of the number of pixels is the number of pixels corresponding to the first area light source information in the area light source information and the number of pixels in one of the second image areas The ratio between the number of pixels corresponding to the second area light source information in the area light source information in one of the second image areas. According to the white balance adjustment method described in item 6 of the scope of patent application, the original image is divided into the second image areas, and at least one of the area light source information corresponding to the second image areas is generated respectively The step of corresponding to the plurality of mixed light source information of the second image areas further includes: If one of the second image areas does not correspond to more than two pieces of the area light source information, generate the second images according to one of the area light source information corresponding to one of the second image areas One of the mixed light source information in one of the areas. An image processing device, including: A storage device storing multiple modules; and A processor, coupled to the storage device, is configured to execute the modules to: Receive an original image; According to the brightness information of the original image, the original image is divided into a plurality of first image regions: For each of the first image areas, generating a plurality of area light source information corresponding to the first image areas; Dividing the original image into a plurality of second image areas, and generating a plurality of mixed light source information corresponding to the second image areas according to at least one of the area light source information corresponding to the second image areas: and According to the mixed light source information of the second image areas, the pixel values in the second image areas are respectively corrected to obtain a white balance image. An image processing system, including: An image sensor for capturing images; A storage device storing multiple modules; and A processor, coupled to the image sensor and the storage device, is configured to execute the modules to: Receive an original image; According to the brightness information of the original image, the original image is divided into a plurality of first image regions: For each of the first image areas, generating a plurality of area light source information corresponding to the first image areas; Dividing the original image into a plurality of second image areas, and generating a plurality of mixed light source information corresponding to the second image areas according to at least one of the area light source information corresponding to the second image areas: and According to the mixed light source information of the second image areas, the pixel values in the second image areas are respectively corrected to obtain a white balance image

Images