CN108875759B - An image processing method, device and server - Google Patents

Abstract

The embodiment of the present invention discloses an image processing method, device and server, which are used for reducing the interference caused by the background image in the Mask to the foreground target. In the embodiment of the present invention, in the process of image processing, the bounding box containing the foreground target Mask is divided into a plurality of small blocks, the color system of each small block is determined, and for each color system, the corresponding scaling coefficient of the color system is determined, and the The luminance value of each pixel in a patch belonging to this color family is multiplied by this scaling factor. In the embodiment of the present invention, the Mask can be divided into multiple small blocks, and the small blocks can be classified according to the color system. The small blocks of each type can be processed in different ways, and the blocks of the same type can be uniformly processed in the same way. Distinguish small blocks with large differences in color values between the foreground target and the background image in the Mask, and perform different processing on them, thereby reducing the interference caused by the background image in the Mask to the foreground target.

Description

An image processing method, device and server

technical field

The present application relates to the field of image processing, and in particular, to an image processing method, device and server.

Background technique

In the application of video analysis such as image search, it is a very critical technology to accurately obtain the color value of the whole or part of the target to extract features or perform corresponding processing.

However, the current mask (Mask) used to determine the target has not been able to accurately frame the contour of the target. In addition to the foreground target, the Mask will inevitably contain a part of the background image. In some cases, these background images can be very noisy for determining the color of foreground objects.

However, the current technology cannot effectively remove the background image in the Mask, and the Mask contains more or less background images other than the foreground target. Therefore, in the case where the accurate Mask cannot be obtained, in the image enhancement processing, the background image in the Mask will inevitably cause noise interference to the foreground target.

SUMMARY OF THE INVENTION

The embodiments of the present invention provide an image processing method, device and server, which can reduce the interference caused by the background image in the Mask to the foreground target.

A first aspect of the present application provides an image processing method, which includes the following steps:

Divide the bounding box containing the foreground target Mask into a plurality of small blocks, and each small block contains at least one pixel; then determine the color system of each small block in the small blocks occupied by the Mask; for each color system, A scaling coefficient corresponding to the color system is determined, and the luminance value of each pixel in the small blocks belonging to the color system is multiplied by the scaling coefficient to separately process the small blocks of each color system.

When the color values of the foreground target and the background image are very different, the background image has a great influence on the interference of the foreground target. The image processing method in the embodiment of the present invention is based on the above-mentioned principle, by dividing the bounding box of the Mask containing the foreground target into blocks, that is, dividing the Mask into blocks, and then classifying the divided small blocks according to the color system, so as to distinguish The small blocks with large differences in color values between the foreground target and the background image can be processed in different ways, and the small blocks of the same type can be uniformly processed in the same way, so as to realize the difference between the foreground target and the background image. When the difference in color value is large, different processing is performed on the foreground target and the background image, thereby reducing the interference caused by the background image in the Mask to the foreground target.

Optionally, the scaling coefficient corresponding to each color system may specifically be a preconfigured scaling preset value, and the scaling preset value is an empirical value obtained through experimental data or data model training.

Optionally, the scaling factor corresponding to each color system can be calculated and obtained according to the average brightness value of the small blocks of the color system. The average luminance value (first average luminance value) of the pixels included in the small block; the ratio between the first average luminance value and the expected luminance value is calculated to obtain the scaling factor corresponding to the color system.

Optionally, the method for determining the color system of each small block in the small blocks occupied by the Mask may be: for each small block in the small blocks occupied by the Mask, according to the background brightness value of the small block, the The color system of the small block is determined by at least one of the luminance value of the target portion of the small block, the saturation value of the target portion of the small block, and the hue value of the target portion of the small block. The background brightness value of the small block refers to the average brightness value of multiple pixels in the bounding box except the Mask, and the target part of the small block refers to the part of the small block that belongs to the Mask. part.

Specifically, the method of determining the color system of the small block according to the background brightness value of the small block and the brightness value of the target part of the small block is as follows: first obtain the corresponding relationship between the background brightness value and the illumination coefficient; the illumination coefficient corresponding to the background luminance value of the small block (the first target illumination coefficient), and then multiply the luminance value of the target portion of the small block by the first target illumination coefficient to obtain the first adjusted luminance value; and then adjust the luminance according to the first The luminance value determines the color family of the patch.

Wherein, determining the color system of the small block according to the first adjusted brightness value is specifically: by judging which color system (dark color system, white light color system or colored color system) the first adjusted brightness value is within the corresponding brightness value range Determine the color family it belongs to.

Specifically, the method of determining the color system of the small block according to the background brightness value of the small block and the brightness value of the target part of the small block is as follows: first obtain a preset background brightness preset value; the ratio between the background luminance values of the blocks to obtain the second target illumination coefficient; multiply the luminance value of the target portion of the small block by the second target illumination coefficient to obtain the second adjusted luminance value; determine the Small blocks of color. The method for determining the color system of the small block according to the second adjusted brightness value may be the same as the above-mentioned method for determining the color system of the small block according to the first adjusted brightness value.

Specifically, the method for determining the color system of the small block according to the saturation value of the target part of the small block and the hue value of the target part of the small block is: judging whether the hue value of the target part of the small block is in a certain color system (the first color system) within the hue threshold range; and determine whether the saturation value of the target part of the small block is within the saturation threshold range of the first color system; if the hue value of the target part of the small block is within the first color range Within the hue threshold range of a color system, and the saturation value of the target portion of the small block is within the saturation threshold range of the first color system, it is determined that the color system of the small block belongs to the first color system.

It should be noted that this method can be combined with other methods to judge the color system of small blocks.

Specifically, according to the background luminance value of the small block, the luminance value of the target portion of the small block, the saturation value of the target portion of the small block, and the hue value of the target portion of the small block, determining the color system of the small block includes: applying the The background luminance value of the small block, the luminance value of the target part of the small block, the saturation value of the target part of the small block and the hue value of the target part of the small block are input into the parameter model to obtain the color system of the small block .

Optionally, the method for determining the parameter model may be: acquiring multiple sets of model training parameters, each group of model training parameters includes a color system determination parameter and a color system value corresponding to the color system determination parameter, wherein the color system determination parameter includes a background image. The brightness value of the selected foreground target, the brightness value of the selected foreground target, the saturation value of the selected foreground target, and the hue value of the selected foreground target; input the above-mentioned multiple sets of model training parameters into the SVM to obtain a parameter model.

In a second aspect, an embodiment of the present invention further provides an image processing apparatus, which specifically implements functions corresponding to the image processing method provided in the first aspect. The functions can be implemented by hardware or by executing corresponding software programs in hardware. The hardware and software include one or more unit modules corresponding to the above functions, and the unit modules may be software and/or hardware.

In a possible design, the image processing device includes:

a small block dividing unit, configured to divide the bounding box into a plurality of small blocks, each small block contains at least one pixel, and the bounding box contains the mask Mask of the foreground target;

a color system judgment unit for determining the color system of each small block in the small blocks occupied by the Mask;

a scaling factor determining unit, configured to determine, for each color system, a scaling factor corresponding to the color system;

A brightness adjustment unit, configured to multiply the brightness value of each pixel in the small block belonging to the color system by the scaling factor.

In a third aspect, an embodiment of the present invention provides a server, which specifically implements functions corresponding to the image processing method provided in the first aspect. The server in this embodiment of the present invention may include:

A processor, a memory, and a network interface; wherein, the memory is used to store program codes, and the processor invokes the program codes in the memory to execute all or part of the steps in the image processing method of the first aspect.

In a fourth aspect, the present application provides a computer-readable storage medium, where instructions are stored in the computer-readable storage medium, when the computer-readable storage medium runs on a computer, the computer executes the method described in the first aspect.

Yet another aspect of the present application provides a computer program product comprising instructions which, when run on a computer, cause the computer to perform the method described in the first aspect above.

It can be seen from the above technical solutions that only when the color values of the foreground target and the background image are very different, the background image has a great influence on the interference of the foreground target, and the foreground target and the background image in the Mask can pass through to a certain extent. Small pieces to distinguish. Therefore, in the image processing process of the embodiment of the present invention, the bounding box is first divided into a plurality of small blocks, and the small blocks occupied by the Mask are classified according to the color system, so that the color values of the foreground target and the background image are greatly different. Small pieces are distinguished. Then, for each color system, determine the scaling coefficient corresponding to the color system, and multiply the luminance value of each pixel in the small blocks belonging to the color system by the scaling coefficient, so as to use the same method for small blocks of the same type Unified processing, and each type of small blocks can be processed in different ways, so as to achieve different processing of pixels with large difference values between the foreground target and the background image in the Mask. Therefore, the interference caused by the background image in the Mask to the foreground target can be reduced.

Description of drawings

1a is a schematic diagram of a bounding box of a target pedestrian in an embodiment of the present invention;

FIG. 1b is a schematic diagram of the Mask of the target pedestrian in FIG. 1a in an embodiment of the present invention;

2 is a schematic diagram of an image processing system in an embodiment of the present invention;

3 is a flowchart of an image processing method in an embodiment of the present invention;

FIG. 4 is a block schematic diagram of the Mask of the target pedestrian in FIG. 1 b according to an embodiment of the present invention;

Fig. 5 is a group of frame images used when judging the color system in the embodiment of the present invention;

6 is a schematic structural diagram of a functional module of an image processing apparatus in an embodiment of the present invention;

FIG. 7 is a schematic diagram of a hardware structure of a server in an embodiment of the present invention.

Detailed ways

The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

In addition to the foreground target, the Mask inevitably contains background noise, and the background noise is the background image except the foreground target in the Mask. Since Mask contains background noise, it is difficult to remove these background noises effectively. In some cases, for example, when the foreground object is black or white, and the brightness value of the background image is quite different from the brightness value of the foreground object, these background noises have great interference in determining the color of the foreground object.

Figure 1a shows the bounding box (Bounding Box) of the target pedestrian (that is, the foreground target) in a frame of image, wherein the bounding box refers to the rectangular box containing the foreground target. Generally speaking, it is the smallest rectangle of the Mask containing the foreground target. frame. Figure 1b is the Mask of the target pedestrian in Figure 1a. As can be seen from Figure 1b, the Mask contains the background between the pedestrian's legs and the shadow under the foot. These background images should not be included in the Mask. These background images will become background noise and interfere with determining the color of the target pedestrian. .

For example, the pixel value of a single pixel near the foot of the target pedestrian is (0, 0, 1). If the ratio of the pixels on the background ground (near the foot) contained in the Mask to the number of pixels near the foot of the pedestrian is 1:3, then the average pixel value of these pixels may be (48, 47, 50), which is far away from the average pixel value (0, 0, 1) of the target pedestrian’s foot, which will lead to big deviation.

Therefore, the purpose of this application is to reduce the interference of the background noise on the image enhancement processing as much as possible when the Mask is not accurate enough and the Mask contains background noise.

When the color values of the foreground target and the background image in Mask are not very different, the impact of background noise on the foreground target will be very limited, and it may not be necessary to distinguish; only when the color values of the foreground target and the background image are very different, the background Noise has a great influence on the interference of foreground targets. Therefore, it is only necessary to distinguish the pixels with a large difference in color value between the foreground object and the background image, and perform noise reduction processing, so as to avoid the interference of the background image to the foreground object.

The image processing method in the embodiment of the present invention is based on the above-mentioned principle. By dividing the pixels in the Mask into blocks, dividing them into multiple small blocks, and classifying the blocks (ie, small blocks) according to the color system, the foreground target is distinguished. The blocks with large differences in color values from the background image, each type of block can be processed in different ways, and the blocks of the same type are processed uniformly in the same way, so as to achieve the color value of the foreground target and the background image. When the difference is large, different processing is performed on the foreground target and the background image, thereby reducing the interference caused by the background image in the Mask to the foreground target.

The image processing method in the embodiment of the present invention may be executed by a network device, and the network device may specifically be various forms of computer devices, including a server, a server cluster, or a terminal device with a communication function.

For convenience of description in the embodiments of the present invention, the method in the embodiments of the present invention is described by taking the network device as a server or a server cluster as an example.

As shown in FIG. 2 , the image processing system for implementing the image processing method in the embodiment of the present invention mainly includes two network elements, a server and a client device.

The server in Figure 2 can be a single server or a distributed deployed server cluster. The client equipment may be various forms of computer equipment or user equipment, such as terminal equipment such as a mobile phone, a tablet computer, and a personal computer (Personal Computer, PC).

The client device communicates with the server through the network, and the client device sends an image processing request to the server. The embodiments of the present invention are described by taking the image processing request as an example of processing an image in a video. In practical applications, an image processing request may not necessarily be processing an image in a video, but may also be a single or multiple images stored in a memory. image is processed.

When the server performs video image processing, the video content management program executes the image processing request of the client device, and the image processing request may be image enhancement processing, lighting processing, or the like. After that, the server returns the image processing result of the video content management module to the client device.

As shown in Figure 2, the video content management program in the server includes a management module and a video analysis module.

Among them, the management module is responsible for performing functions such as scheduling requests, resource management, and data retrieval.

The video analysis module is used to perform the function of video analysis, which includes an image preprocessing module. The image preprocessing module includes a background noise removal module and a color system judgment module, and the background noise removal module and the color system judgment module are specifically used to execute the method steps in the method embodiments of the present application.

The image preprocessing module preprocesses the image read from the storage module by the video analysis module and returns the preprocessing result to the core analysis module in the video analysis module. After the preprocessing result is further processed by the core analysis module of the video analysis module, The processing result is obtained, and finally the server returns the processing result to the client device.

It should be noted that the image processing system shown in FIG. 2 is only an example, and in practical applications, the network element composition and deployment manner of the image processing system may be determined according to specific application scenarios. In addition, the division of functional modules in the video content management program in FIG. 2 is only an example, which is only for explaining the process of video image processing.

The image processing method in the embodiment of the present invention will be described in detail below with reference to FIG. 3 .

301. Divide the bounding box of the Mask containing the foreground target into a plurality of small blocks, and each small block contains at least one pixel;

Due to the illumination, the wrinkles and occlusion of the foreground target, the pixel values of the pixels at different positions of the same color of the foreground target will be quite different. Judging and processing the color system based on the pixel points will lead to great deviation, so that the image Obvious distortion. Therefore, it is more appropriate to judge the color system based on a small block composed of multiple pixels as the basic unit.

Therefore, in this embodiment of the present invention, the bounding box for framing the foreground object is divided into n×m small blocks, wherein the bounding box can be divided into n columns in the horizontal direction and m rows in the vertical direction to divide into small blocks of n × m. Each small block contains at least one pixel.

In this way, the Mask in the bounding box is divided into multiple small blocks. Some small blocks in the mask all belong to the foreground target; some small blocks all belong to the background image; some small blocks belong to the foreground target and some belong to the background image.

Taking the bounding box containing the target pedestrian shown in FIG. 1a as an example, FIG. 4 is a cutout of the bounding box of the lower body of the target pedestrian in FIG. 1a, and the bounding box of the lower body is divided into a plurality of small blocks. 401 shows the bounding box, and 402 shows the outline of the mask of the lower body. In the mask of the lower body, the one marked H is every small piece that belongs to the foreground; the one marked G is every small piece that belongs to the background; the one marked S1 is half or more of each small piece belonging to the background; marked as More than half of S2 belongs to each patch in the foreground; the rest of the unlabeled patches outside the Mask in the bounding box are all partial patches that belong to the background. Usually, the number of small blocks marked S1 and S2 in the Mask will be much less than the number of small blocks marked H.

After dividing the bounding box into small blocks, the Mask is divided into small blocks. It should be noted that no marking is required in the actual steps. Whether the marked block in Figure 4 belongs to the foreground or the background is just to be able to better understand the foreground part (that is, the foreground target) and the background part (that is, the background image) in the Mask. ) can be distinguished to some extent by small blocks.

It should be noted that, because the purpose of this application is to minimize the interference caused by the background noise in the Mask to the foreground target. Therefore, the essence of dividing the bounding box into small blocks is to divide the Mask into small blocks. Because the bounding box is a regular rectangle, and the mask is an irregular shape, it is more appropriate to divide the smallest rectangular box containing the mask into small blocks than to divide the mask into small blocks.

However, in the process of processing the small blocks in the bounding box (that is, in steps 302 and 303 ), only the small blocks involved in the Mask can actually be processed, and the small blocks not involved in the Mask in the bounding box can be processed. If the processing in steps 302 and 303 is not performed, the small blocks not involved in the Mask can refer to the small blocks that are not marked in the background image except for the Mask in FIG. 4 .

The small blocks involved in the Mask here are the small blocks occupied by the Mask, that is, all the small blocks in the bounding box belong to the Mask, and a part of the small blocks belong to the Mask.

302. Determine the color system of each small block in the small blocks occupied by the Mask;

After dividing the bounding box into multiple small blocks, for each small block occupied by Mask in the bounding box, determine the color system to which it belongs. The color system includes one of the dark color system, the colored color system, and the white light color system.

Among them, the dark color system refers to colors such as black, black gray, navy blue, and navy black and similar colors. One of their main characteristics is that the hue value (Hue, H) of the image remains unchanged or changes very little, or in the light When the intensity is not excessive, the value of the luminance (Value, V) component is small, and the value range of V is also small, for example, the value of V ranges from 0.02 to 0.18.

Colored colors refer to colors such as red, orange, yellow, green, cyan, blue and purple and similar colors. One of their main characteristics is that the V component of the image remains unchanged or changes very little. There can be a wide range of values, and the value of V can be small or large, such as from 0.2 to 0.8.

White and bright colors refer to white or similar colors. One of their main characteristics is that when the H value of the image remains unchanged or changes very little, the value range of the V component is small (relative to the colored color system), But the value of V is larger, for example, the value of V is from 0.7 to 0.9.

Optionally, the method for determining the color system to which each small block belongs may be: according to the background luminance value of the small block, the luminance value of the target portion of the small block, the saturation value of the target portion of the small block, and the target portion of the small block. at least one of the hue values to determine the color family of the patch.

The target part of the small block refers to the part of the small block belonging to the Mask of the foreground target, that is, the pixels belonging to the Mask. If for a small block, the whole small block is located in the Mask, then the target part of the small block is all the pixels in the whole small block; The target part of is the part of the pixels located in the Mask. The brightness value of the target part of the small block refers to the average brightness of the pixels in the small block belonging to the Mask; the saturation value of the target part of the small block refers to the average saturation value of the pixels belonging to the Mask in the small block; The hue value of the target part refers to the average hue of the pixels in the patch that belong to the Mask.

The background brightness value of the small block is the average brightness value of multiple pixels in the bounding box except for the Mask of the foreground target. The average luminance value of some pixels, where the partial pixels may be some pixels located in the adjacent area of the current small block.

The specific manner of determining the color system of the small block according to at least one of the background luminance value of the small block, the luminance value of the target part of the small block, the saturation value of the target part of the small block and the hue value of the target part of the small block includes: : The color system of the small block can be determined according to the background brightness value of the small block and the brightness value of the target part of the small block; the color system of the small block can also be directly determined according to the brightness value of the target part of the small block; The saturation value of the foreground target and the hue value of the foreground target of the patch determine the color system of the patch.

Optionally, the method for determining the color system to which each small block belongs can also be, by combining multiple sets of parameters (each set of parameters includes a color system value and H, S and/or V values for determining the color system value) The parameter model is obtained by inputting the machine learning model for training, and the color value of a small block can be obtained by inputting the H, S and/or V values of a small block into the parameter model.

303. For each color system, determine a scaling coefficient corresponding to the color system, and multiply the luminance value of each pixel in the small block belonging to the color system by the scaling coefficient.

After determining the color system to which each small block occupied by the Mask belongs, classify all the small blocks occupied by the Mask according to the color system, and determine all the small blocks belonging to the same color system.

For each patch belonging to the same color family, "unify" it. "Unified processing" is to suppress noise, avoid causing large distortion of the image and obvious blocking effect, and reduce the apparent difference of foreground objects, such as reducing the apparent difference caused by different lighting, and the pixel points in the bounding box. The brightness value is adjusted to the brightness value under normal lighting environment.

The unified processing method is: for each color system, the luminance value of each pixel in the small block belonging to the color system is scaled by the same multiple, that is, multiplied by a scaling factor.

Optionally, the scaling factor may be a preconfigured scaling preset value, and the scaling preset value may be an empirical value obtained according to experimental data or data model training. For example, the value range of the scaling factor corresponding to the dark color system can be in the range of [0.5-5], the value range of the zoom factor corresponding to the colored color system can be in the range [0.625-2], and the value range of the zoom factor corresponding to the white and light color system The value range can be in the interval [0.98-1.25].

计算第一平均亮度值与亮度预期值之间的比值

以得到在边界框中该色系对应的缩放系数。举例如下：Optionally, the scaling factor may be calculated and obtained according to the luminance value of the small block of the color system in the bounding box. The step of calculating the scaling factor corresponding to a certain color system is as follows: determining the expected brightness value V ₀ corresponding to the color system, and calculating the average brightness value of the pixels (ie, pixel points) included in the small blocks belonging to the color system (called the first average value). brightness value)

Calculate the ratio between the first average luminance value and the expected luminance value

to get the scaling factor corresponding to the color system in the bounding box. An example is as follows:

根据各色系的亮度值的范围可知，这是黑暗色系的小块，并且其所处于光照较暗的环境，黑暗色系的亮度预期值V₀＝0.16，计算

然后该类小块中的每一个像素点的亮度值都乘以4。For example, calculating for a certain class of small blocks yields

According to the range of brightness values of each color system, it is known that this is a small block of dark color system, and it is located in a dark environment, the expected brightness value of dark color system V ₀ =0.16, calculate

Then the brightness value of each pixel in this type of small block is multiplied by 4.

After multiplying the small blocks in each color system by the corresponding scaling factor of the color system, the actual effect obtained is to enhance the image small blocks in a dark environment to make them match the dark color in a normal light environment. The brightness value of the system image is closer.

In this way, the independent processing of the pixels of the dark color system, the white light color system and the colored color system is realized. Because the foreground part and the background part in the Mask can be distinguished by small blocks to a certain extent, the embodiment of the present invention divides the image in the bounding box into blocks. After the image is divided into blocks, the blocks are classified according to the color system, so that the pixels with large difference in color value between the foreground target and the background image can be distinguished. Then, the same type of blocks are processed uniformly in the same way, and each type of blocks can be processed in different ways, so as to achieve different processing of pixels with a large difference between the foreground target and the background image, which can reduce the Mask. The interference caused by the background image to the foreground target.

The following describes in detail how to determine the color system of each small block.

1. Judging color system by fuzzy distinction method:

In some specific embodiments, the color system of the small block may be determined according to the background luminance value of the small block and the luminance value of the target portion of the small block. The specific judgment method is:

Obtain the corresponding relationship between the preset background brightness value and the illumination coefficient, which can be obtained through a large amount of experimental data; and then determine the target illumination corresponding to the background brightness value of the small block according to the corresponding relationship between the background brightness value and the illumination coefficient coefficient (referred to as the first target illumination coefficient); the luminance value of the target portion of the small block is multiplied by the first target illumination coefficient to obtain the first adjusted luminance value, so as to adjust the luminance value of the pixels belonging to the Mask in the small block to The brightness value under normal lighting; and then determine the color system of the small block according to the first adjusted brightness value.

Wherein, determining the color system of the small block according to the first adjusted brightness value is specifically, by judging whether the first adjusted brightness value is in the V value range corresponding to the dark color system, or in the V value range corresponding to the colored color system, or in the white light system The corresponding V value range of the color system is used to determine the color system to which it belongs.

When the first adjusted brightness value is in the V value range corresponding to the dark color system (can be set to 0.02-0.18), the color system to which the small block belongs is the dark color system; when the first adjusted brightness value V corresponds to the colored color system When the V value range (can be set to 0.2-0.8), the color system of the small block is a colored color system; when the first adjusted brightness value V is in the V value range corresponding to the white and bright color system (can be set to 0.7- 0.9), the color of the small block is the dark color.

Further, if the first adjusted brightness value is at a critical value between two color systems (for example, between 0.18-0.2), and the color system of the current small block cannot be determined, one or more adjacent small blocks can be combined The color system and its adjusted brightness value are used to determine the color system of the current small block.

In some specific embodiments, if all the pixels included in a small block are located in the Mask of the foreground target, the color system of the small block may be determined according to the luminance value of the target portion of the small block. The specific judging method is: calculating the average brightness value of the pixels in the small block (called the second average brightness value), and determining the color system of the small block according to the second average brightness value.

Wherein, the method of determining the color system of the small block according to the second average brightness value is the same as the method of determining the color system of the small block according to the first adjusted brightness value: when the second average brightness value V is at the V corresponding to the dark color system When the value range (can be set to 0.02-0.18), the color system of the small block is the dark color system; when the second average brightness value V is in the V value range corresponding to the colored color system (can be set to 0.2-0.8) , the color system of the small block is a colored color system; when the second average brightness value V is in the V value range corresponding to the white and bright color system (can be set to 0.7-0.9), the color system of the small block is dark color system.

The following is an example of specific data:

Based on the observation of the data of multiple samples to be processed (for example: 5000 samples), the corresponding relationship between the background brightness value and the illumination coefficient is obtained. As shown in Table 1, the background brightness is divided into dark and dark according to the size of the V value. , normal, bright and bright five ranges, each range has a corresponding background brightness V value and illumination coefficient.

Table 1

Background V value Judgment result of background light and shade Lighting coefficient (value in this example) V<=0.3 dark 2.0 0.3<V<=0.44 darker 1.4 0.44<V<=0.54 normal 1.1 0.54<V<=0.65 brighter 0.9 V>0.65 Bright 0.8

The lighting coefficient is used to adjust the V value of the foreground target with different lighting, so that it is closer to the V value under normal lighting conditions. For example, if the V value of the background brightness of the current small block is 0.4, according to the background V value range and the background brightness judgment result in Table 1, the background brightness is judged to be dark, and the corresponding illumination coefficient is 1.4. If the brightness of the target part of the small block is 0.16, it is adjusted to 1.4*0.16=0.224, and 0.224 belongs to the interval range of the brightness value of the colored color system from 0.2 to 0.8. Therefore, based on the adjusted V value of 0.224, it is judged that the small The color family to which the block belongs is the tinted color family.

In the following, a set of specific image data is used to describe the determination of the color system by the fuzzy discrimination method in the embodiment of the present invention.

A set of specific frame images is shown in Figure 5. In this set of image frames, each frame image is a bounding box containing a target pedestrian (the foreground target is a pedestrian). In order to more intuitively describe how to judge the color system, the present invention The embodiment is described by taking as an example that the bounding box is divided into n×m=1×2=2 small blocks and divided into two small blocks, the upper body and the lower body.

It should be noted that, in practical applications, the bounding box should be divided into more small blocks, so that the foreground part and the background part in the Mask can be better distinguished by small blocks.

The image names of a group of frame images (images 1 to 10) shown in Figure 5 _are 0.001png, 0.002png, 0.003png... value (V _lower ) and the background V value (V _background ), the values are as follows:

Among them, in the example shown in FIG. 5, the V value of the upper body (V _up ) is the brightness value of the target part of the small block corresponding to the upper body, that is, the average brightness value of the pixels in the Mask of the upper body; the V value of the lower body (V _lower ) is the brightness value of the target part of the small block corresponding to the lower body, that is, the average brightness value of the pixels in the Mask of the lower body; in this example, it is considered that the illumination where the upper body and the lower body are located is relatively average, therefore, the V _background is the smaller of the upper body. The background brightness value of the block is also the background brightness value of the small block of the lower body, and the V _background value may be the average value of the brightness of the pixels other than the Mask of the upper body and the lower body.

According to the corresponding relationship between the background V value and the illumination coefficient in Table 1, determine the illumination coefficient corresponding to the V _background , multiply the illumination coefficient by the corresponding V _upper body to obtain the adjusted brightness value of the upper body, and multiply the illumination coefficient by the corresponding V _lower Get the brightness value of the adjusted lower body. The adjusted V and V _of each frame image _are shown in Table 2 and Table 3:

Table 2

table 3

Comparing the pre-adjusted brightness value with 0.18, both are less than 0.18, so it can be judged that the lower body of these 10 frames of images belongs to the dark color system; the upper body of the images 1, 4, 8, 9 and 10 is colored. The upper body brightness of images 2, 3, 5, 6 and 7 is located in the critical value of dark and colored colors: the interval range of 0.18-0.2. Therefore, it is necessary to further judge images 2, 3, 5, 6 and 6 through other methods. The color of the upper body of 7.

Further, when the color system of the upper body cannot be determined by the adjusted brightness value, the judgment can be made in combination with the adjacent small blocks, that is, the corresponding lower body. Because the lighting environment of the upper body and the lower body are generally similar, the brightness value of the lower body is judged to be dark, and the adjusted brightness value is still small, while the adjusted brightness value of the upper body is greater than 0.18, so it can be judged that pictures 2 and 3 , 5, 6 and 7 are not dark and therefore belong to the tinted family.

Further, if it is still impossible to determine the color system of the lower body in combination with the lower body, in order to prevent misjudgment, it can be classified into two categories, that is, it is judged that it is both a dark color system and a colored color system, and then processed according to the two color systems to avoid Misjudgment results in processing errors.

2. Judging the color system by the specific combination of HS

In some specific embodiments, the color system to which the small block belongs can be directly determined according to a specific combination of the HS of the target part of the small block, that is, according to the saturation value of the target part of the small block and the saturation value of the target part of the small block Hue value, which determines the color family of the patch. The method is specifically: judging whether the hue value of the target part of the small block is within the hue threshold range of a certain color system (the first color system); and judging whether the saturation value of the target part of the small block is within the color system If the hue value of the target part of the small block is within the hue threshold value range of the color system, and the saturation value of the target part of the small block is within the saturation threshold value range of the color system, then It is determined that the color system of the small block belongs to the color system, that is, it belongs to the first color system. Among them, the first color system is one of the dark color system, the colored color system and the white light color system. The saturation threshold value range and the hue threshold value range of a certain color system are based on experimental observation, and obtained specific saturation range and hue range corresponding to the color system.

For example: based on experimental observations, in the colored color system, red H and S have a specific combination range, when H is in the [0.9,1] interval, and correspondingly S is also in the [0.9,1] interval, although the value of V Completely different, it can be determined that the color of the small block is red and belongs to the colored color system.

It should be noted that the method of judging the color system by a specific combination of HS can be used alone or in combination with other color system judging methods. For example: the following HSV data is for the upper body as shown in Figure 5 as a small block, the calculated HSV data is:

Picture name H _on S _on V _on V _background

0011.png 0.982 0.936 0.095 0.292

The color system of 0.011 in the picture cannot be determined by the fuzzy discrimination method in the embodiment of the present invention, because according to the V _background and the corresponding relationship between the background brightness value and the illumination coefficient shown in Table 1, the illumination coefficient is determined to be 2.0, and the target of the upper body is determined to be 2.0. The adjusted V value obtained by _multiplying the partial brightness value V by the illumination coefficient is 0.19, which is just in the critical value range (0.18-0.2) of the dark color system and the colored color system, and it is difficult to determine the color system. In combination with the specific combination of HS, the H value and S value of the frame image are both in the [0.9, 1] interval, so it can be determined that the color of the small block is red, which belongs to the colored color system.

3. Judging color system by comprehensive evaluation method

In some specific embodiments, a comprehensive evaluation method may also be used to determine the color system of the small blocks. The specific process of the comprehensive evaluation method includes: obtaining a preset background brightness preset value, wherein the preset background brightness preset value can be the background brightness average value calculated according to the background brightness values of a large number of pictures; and then calculating the background brightness The ratio between the brightness preset value and the background brightness value of the current small block to obtain the second target illumination coefficient; and then for each small block, the foreground target brightness value of the small block is multiplied by the obtained second target illumination coefficient to obtain The second adjusts the brightness value to adjust the brightness value of the small block to the brightness value under normal lighting, and then determines the color system of the small block according to the second adjusted brightness value. The background brightness value of the current small block is the average brightness value of multiple pixels in the bounding box except for Mask, and the multiple pixels in the bounding box except for Mask can be all pixels in the bounding box except for Mask, or the boundary Part of the pixels other than the Mask in the box, where the part of the pixels may be a part of the pixels located in the adjacent area of the current small block.

The method of determining the color system of the small block according to the second adjusted brightness value is the same as the method of determining the color system of the small block according to the first adjusted brightness value: that is, when the second adjusted brightness value V corresponds to the dark color system When the V value range (can be set to 0.02-0.18), the color system of the small block is the dark color system; when the second adjusted brightness value V is in the V value range corresponding to the colored color system (can be set to 0.2-0.8 ), the color system of the small block is a colored color system; when the second adjusted brightness value V is in the V value range corresponding to the white and bright color system (can be set to 0.7-0.9), the color system of the small block is Dark shades.

The following describes how to judge the color system by the comprehensive evaluation method in the embodiment of the present invention with a set of specific image data.

First, acquire millions of pictures in the existing database, and calculate the average value of the background brightness of these pictures. If the calculation result is 0.491, set it as the preset background brightness preset value.

Then calculate the ratio of the background brightness of each small block to 0.491, and the calculation result is called the illumination coefficient of the small block. Multiply the illumination coefficient by the brightness value of the target part of the small block to obtain the adjusted brightness value of the foreground target of the small block, and then compare with the set thresholds of each color system to determine the color system to which the small block belongs.

4. Judging color system by machine learning method

In some specific embodiments, a parametric model can also be established by a machine learning method, and the color system of the small block is obtained by inputting the relevant parameters of the small block into the parametric model. The specific implementation method is:

Determine the parameter model, for each small block, input the background brightness value of the small block and the parameters of the target part of the small block (H, S and V values of the target part of the small block) into the parameter model, and obtain the small block's color system.

The specific method for determining the parameter model is: obtaining multiple sets of model training parameters, each group of model training parameters includes a color system determination parameter and a color system value corresponding to the color system determination parameter, and the color system determination parameter includes the brightness value of the background image , the brightness value of the selected foreground target, the saturation value of the selected foreground target, and the hue value of the selected foreground target; inputting the multiple sets of model training parameters into the machine learning model to obtain the parameter model.

Specifically, the machine learning model for model training can use the SVM model, and can use the SVM model that comes with Matlab for training. First, label the data, for example, set the label values of -1, 0, and 1 for dark color, colored color, and white light color, respectively. Then, the H, S, and V values of the selected foreground objects and the V value of the background image, and the label value of the annotation result are input as parameters for training. The result of the training is a parametric model, and the parametric model obtained by training is used for the small blocks that have not been manually labeled, and the output result is a label value. Based on this label value (the label values of the dark color system, the colored color system and the white light color system are -1, 0 and 1 respectively), the color system of the small block can be determined.

It should be noted that the color system judgment methods in the embodiments of the present invention can be used alone, and when one of the methods cannot be used to judge the color system of a small block, it can be judged in combination with other methods.

The above is an introduction to the method embodiments in the embodiments of the present invention, and the following describes the image processing apparatuses in the embodiments of the present invention from the perspective of functional modules.

As shown in Figure 6, the image processing device includes the following functional units:

A small block dividing unit 601, for dividing the bounding box of the Mask containing the foreground target into a plurality of small blocks, and each small block contains at least one pixel;

Color system judgment unit 602, for determining the color system of each small block in the small block occupied by Mask;

A scaling factor determining unit 603, configured to determine, for each color system, a scaling factor corresponding to the color system;

A brightness adjustment unit 604, configured to multiply the brightness value of each pixel in the small block belonging to the color system by a scaling factor.

Optionally, in some specific embodiments, the scaling coefficient determination unit 603 is specifically configured to determine the expected brightness value corresponding to the color system; calculate the first average brightness value of the pixels included in the small block belonging to the color system; calculate the first average brightness value. The ratio between the average luminance value and the expected luminance value to obtain the scaling factor corresponding to the color system in the bounding box.

Optionally, in some specific embodiments, the color system judgment unit 602 is specifically configured to, for each small block in the plurality of small blocks occupied by the Mask, according to the background brightness value of the small block, the target of the small block. at least one of the luminance value of the portion, the saturation value of the target portion of the small block, and the hue value of the target portion of the small block, to determine the color system of the small block; wherein the background luminance value of the small block includes the bounding box The average luminance value of multiple pixels other than Mask, and the target part of the small block includes the part belonging to the Mask in the small block.

Optionally, in some specific embodiments, the color system judgment unit 602 is specifically configured to obtain the corresponding relationship between the background brightness value and the illumination coefficient; determine the background brightness of the current small block according to the corresponding relationship between the background brightness value and the illumination coefficient. The first target illumination coefficient corresponding to the value; the luminance value of the target portion of the small block is multiplied by the first target illumination coefficient to obtain the first adjusted luminance value; the color system of the small block is determined according to the first adjusted luminance value.

Optionally, in some specific embodiments, the color system determination unit 602 is specifically configured to obtain a preset background brightness preset value; calculate the ratio between the background brightness preset value and the background brightness value of the current small block to obtaining a second target illumination coefficient; multiplying the luminance value of the target portion of the small block by the second target illumination coefficient to obtain a second adjusted luminance value; and determining the color system of the small block according to the second adjusted luminance value.

Optionally, in some specific embodiments, the color system determination unit 602 is specifically configured to determine whether the hue value of the target part of the small block is within the hue threshold range of the first color system; determine the saturation of the target part of the small block. Whether the hue value is within the saturation threshold range of the first color system; if the hue value of the target part of the small block is within the hue threshold range of the first color system, and the saturation value of the target part of the small block is within the first color system Within the saturation threshold range of , it is determined that the color system of the small block belongs to the first color system.

Optionally, in some specific embodiments, the color system judgment unit 602 is specifically configured to compare the background luminance value of the small block, the luminance value of the target part of the small block, the saturation value of the target part of the small block and the small block. The hue value of the target part is input into the parametric model to get the hue of the patch.

Optionally, the device further includes: a model training unit (not shown in the figure) for acquiring multiple groups of model training parameters, each group of model training parameters including a color system determination parameter and a color system value corresponding to the color system determination parameter, The color system determination parameters include the brightness value of the background image, the brightness value of the selected foreground object, the saturation value of the selected foreground object, and the hue value of the selected foreground object; input multiple sets of model training parameters into the support vector machine SVM to obtain the parameter model .

For the information interaction between the units in the above image processing apparatus, reference may be made to the above method embodiments (the embodiments shown in FIG. 1 to FIG. 5 ), which will not be described in detail in this application.

In practical applications, the image processing apparatus shown in FIG. 6 may be implemented by an application-specific integrated circuit (ASIC) or a programmable logic device (PLD). The above-mentioned PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL) or any combination thereof.

In practical applications, the image processing apparatus shown in FIG. 6 may be a server. 7 is a schematic diagram of a hardware structure of a server according to an embodiment of the present invention. The server may vary greatly due to different configurations or performance, and may include one or more processors (central processing units, CPU) 722 and memory 730 and one or more wired or wireless network interfaces 750. The memory 730 stores one or more application programs 742, data 744 and one or more operating systems 741. The application program 742 stored in the memory 730 may include one or more modules, and each module may include a series of Command operation.

In this embodiment of the present invention, the application program 742 may include the video content management program shown in FIG. 2 , and the processor 720 runs the video content management program stored in the memory 730 to implement the above method embodiments (FIG. 1 to FIG. 5 ). all or part of the steps in the examples shown).

Data 744 stored in memory 730 may include video, images, and possibly video or image summaries, indexes, and other information.

In addition, the server may also include one or more power supplies 726 , and one or more input and output interfaces 758 .

It should be noted that the server shown in FIG. 7 may be a distributed deployment server cluster, and the application programs, data 744 and the like on the storage 730 may be distributed and stored in each server of the server cluster.

The server shown in FIG. 7 corresponds to the server in the image processing system architecture shown in FIG. 2. The client device interacts with the server. The server executes the request of the client device through the video content management module and returns the corresponding result to the client device. .

It should be noted that, in the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented in whole or in part in the form of a computer program product.

The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer may be a general purpose computer, special purpose computer, computer network, or other programmable device. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be downloaded from a website site, computer, server, or data center Transmission to another website site, computer, server, or data center is by wire (eg, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (eg, infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be stored by a computer or a data storage device such as a server, a data center, etc. that includes one or more available media integrated. The usable media may be magnetic media (eg, floppy disks, hard disks, magnetic tapes), optical media (eg, DVD), or semiconductor media (eg, Solid State Disk (SSD)), and the like.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above may refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

The terms "first", "second", "third", "fourth", etc. (if any) in the description and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to Describe a particular order or sequence. It is to be understood that data so used may be interchanged under appropriate circumstances so that the embodiments described herein can be practiced in sequences other than those illustrated or described herein. Furthermore, the terms "comprising" and "having" and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those expressly listed Rather, those steps or units may include other steps or units not expressly listed or inherent to these processes, methods, products or devices.

In the several embodiments provided in this application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.

The integrated unit, if implemented in the form of a software functional unit and sold or used as an independent product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of software products in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, and the computer software products are stored in a storage medium , including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or optical disk and other media that can store program codes .

As mentioned above, the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand: The technical solutions described in the embodiments are modified, or some technical features thereof are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions in the embodiments of the present application.

Claims (15)

1. an image processing method, is characterized in that, described method comprises: Divide the bounding box into a plurality of small blocks, each small block contains at least one pixel, and the bounding box contains the mask Mask of the foreground target; determining the color system of each of the small blocks occupied by the Mask; For each color system, determine a scaling factor corresponding to the color system, and multiply the luminance value of each pixel in the small block belonging to the color system by the scaling factor; The determining the color system of each small block in the small blocks occupied by the Mask includes: For each small block in the small blocks occupied by the Mask, according to the background luminance value of the small block, the luminance value of the target part of the small block, the saturation value of the target part of the small block and the at least one of the hue values of the target portion of the small block, determining the color system of the small block; The background luminance value of the small block includes the average luminance value of a plurality of pixels other than the Mask in the bounding box, and the target portion of the small block includes the portion of the small block that belongs to the Mask. 2. The method according to claim 1, wherein the determining the scaling factor corresponding to the color system comprises: determining the expected brightness value corresponding to the color system; calculating the first average luminance value of the pixels included in the small blocks belonging to the color system; A ratio between the first average luminance value and the expected luminance value is calculated to obtain a scaling factor corresponding to the color system. 3. The method according to claim 1, wherein the determining the color system of the small block according to the background luminance value of the small block and the luminance value of the target portion of the small block comprises: Obtain the corresponding relationship between the background brightness value and the illumination coefficient; determining the first target illumination coefficient corresponding to the background luminance value of the small block according to the correspondence between the background luminance value and the illumination coefficient; multiplying the brightness value of the target portion of the small block by the first target illumination coefficient to obtain a first adjusted brightness value; The color system of the small block is determined according to the first adjusted luminance value. 4. The method according to claim 1, wherein the determining the color system of the small block according to the background luminance value of the small block and the luminance value of the target part of the small block comprises: Get the preset background brightness preset value; calculating the ratio between the background brightness preset value and the background brightness value of the small block to obtain a second target illumination coefficient; multiplying the brightness value of the target portion of the small block by the second target illumination coefficient to obtain a second adjusted brightness value; The color system of the small block is determined according to the second adjusted luminance value. 5. The method according to any one of claims 1 to 4, wherein the determining the Small block shades include: Judging whether the hue value of the target portion of the small block is within the hue threshold range of the first color system; judging whether the saturation value of the target portion of the small block is within the saturation threshold range of the first color system; If the hue value of the target portion of the small block is within the hue threshold range of the first color system, and the saturation value of the target portion of the small block is within the saturation threshold range of the first color system, Then it is determined that the color system of the small block belongs to the first color system. 6. The method according to any one of claims 1 to 4, wherein the background luminance value of the small block, the luminance value of the target portion of the small block, the luminance value of the target portion of the small block The saturation value and the hue value of the target portion of the small block, determining the color system of the small block includes: Input the background brightness value of the small block, the brightness value of the target part of the small block, the saturation value of the target part of the small block and the hue value of the target part of the small block into the parameter model to obtain The color of the small pieces. 7. The method according to claim 6, wherein the method further comprises: Obtain multiple sets of model training parameters, each set of model training parameters includes a color system determination parameter and a color system value corresponding to the color system determination parameter, and the color system determination parameter includes the brightness value of the background image and the brightness value of the selected foreground target. , the saturation value of the selected foreground target and the hue value of the selected foreground target; Inputting the multiple sets of model training parameters into a support vector machine (SVM) to obtain the parameter model. 8. An image processing device, wherein the device comprises: a small block dividing unit, configured to divide the bounding box into a plurality of small blocks, each small block contains at least one pixel, and the bounding box contains the mask Mask of the foreground target; a color system judgment unit for determining the color system of each small block in the small blocks occupied by the Mask; a scaling factor determining unit, configured to determine, for each color system, a scaling factor corresponding to the color system; a brightness adjustment unit, configured to multiply the brightness value of each pixel in the small block belonging to the color system by the scaling factor; The color system judgment unit is specifically configured to, for each small block in the plurality of small blocks occupied by the Mask, according to the background brightness value of the small block, the brightness value of the target part of the small block, the at least one of the saturation value of the target portion of the small block and the hue value of the target portion of the small block, to determine the color system of the small block; wherein the background luminance value of the small block includes the bounding box The average luminance value of a plurality of pixels other than the Mask, and the target part of the small block includes the part of the small block that belongs to the Mask. 9. The device of claim 8, wherein The scaling coefficient determination unit is specifically configured to determine the expected brightness value corresponding to the color system; calculate the first average brightness value of the pixels included in the small blocks belonging to the color system; calculate the first average brightness value and the The ratio between the expected values of luminance is obtained to obtain the scaling factor corresponding to the color system in the bounding box. 10. The device of claim 8, wherein The color system judging unit is specifically configured to obtain the corresponding relationship between the background brightness value and the lighting coefficient; according to the corresponding relationship between the background brightness value and the lighting coefficient, determine the first target lighting coefficient corresponding to the background brightness value of the small block ; multiply the brightness value of the target portion of the small block by the first target illumination coefficient to obtain a first adjusted brightness value; and determine the color system of the small block according to the first adjusted brightness value. 11. The apparatus of claim 8, wherein The color system judgment unit is specifically configured to obtain a preset background brightness preset value; calculate the ratio between the background brightness preset value and the background brightness value of the small block to obtain the second target illumination coefficient; The luminance value of the target portion of the small block is multiplied by the second target illumination coefficient to obtain a second adjusted luminance value; the color system of the small block is determined according to the second adjusted luminance value. 12. The device according to any one of claims 8 to 11, wherein The color system judgment unit is specifically configured to judge whether the hue value of the target part of the small block is within the hue threshold value range of the first color system; judge whether the saturation value of the target part of the small block is within the first color system. If the hue value of the target part of the small block is within the hue threshold range of the first color system, and the saturation value of the target part of the small block is within the range of the saturation threshold of the first color system Within the saturation threshold range of a color system, it is determined that the color system of the small block belongs to the first color system. 13. The device according to any one of claims 8 to 11, wherein The color system judgment unit is specifically configured to compare the background luminance value of the small block, the luminance value of the target portion of the small block, the saturation value of the target portion of the small block, and the target portion of the small block. The hue value of is input into the parametric model to get the hue of the patch. 14. The apparatus of claim 13, wherein the apparatus further comprises: The model training unit is used to obtain multiple groups of model training parameters, each group of model training parameters includes a color system determination parameter and a color system value corresponding to the color system determination parameter, and the color system determination parameter includes the brightness value of the background image, the selected Determine the brightness value of the foreground target, the saturation value of the selected foreground target and the hue value of the selected foreground target; input the multiple sets of model training parameters into the support vector machine SVM to obtain the parameter model. 15. A server, comprising: processors, memory and network interfaces; The memory is used to store program code, and the processor invokes the program code in the memory to perform the following operations: Divide the bounding box into a plurality of small blocks, each small block contains at least one pixel, and the bounding box contains the mask Mask of the foreground target; determine the color system of each small block in the small blocks occupied by the Mask ; For each color system, determine the corresponding scaling factor of the color system, and multiply the brightness value of each pixel in the small block belonging to the color system by the scaling factor; The determining the color system of each small block in the small blocks occupied by the Mask includes: For each small block in the small blocks occupied by the Mask, according to the background luminance value of the small block, the luminance value of the target part of the small block, the saturation value of the target part of the small block and the at least one of the hue values of the target portion of the small block, determining the color system of the small block; The background luminance value of the small block includes the average luminance value of a plurality of pixels in the bounding box except the Mask, and the target portion of the small block includes the portion of the small block that belongs to the Mask.

Images