CN109035167A - Method, device, device and medium for processing multiple human faces in an image - Google Patents

Abstract

According to the present disclosure, it relates to a method, device, device and medium for processing multiple human faces in an image. According to one embodiment, a method for processing multiple human faces in an image is provided. In the method, multiple human faces in the image are detected; the image is segmented to separate the multiple human faces from the background of the image; for a given human face in the separated multiple human faces , generate a given evaluation vector based on the attributes of the given face; based on the given evaluation vector, perform corresponding blur processing on the given face; and restore the multiple faces after the blur processing to in the image.

Description

Method, device, device and medium for processing multiple human faces in an image

technical field

The present disclosure relates to the field of image processing, and in particular to a method, device, device and medium for performing face blurring processing on an image containing multiple faces.

Background technique

With the advent of the era of big data, massive amounts of data are generated around people every day, including pictures, voice, video, text and so on. With the full popularity of the Internet, ordinary people can easily get a variety of pictures, which will involve issues of personal privacy and portrait rights. Based on this, people often need to blur these pictures and videos when using them.

Contents of the invention

Based on the above problems, according to example implementations of the present disclosure, a solution for processing multiple human faces in an image is provided.

In a first aspect of the present disclosure, a method for processing multiple human faces in an image is provided. Specifically, the method includes: detecting multiple human faces in an image; performing segmentation processing on the image to separate the multiple human faces from the background of the image; for a given human face among the separated multiple human faces, based on a given The attribute of the face generates a given evaluation vector; based on the given evaluation vector, corresponding blurring is performed on the given face; and multiple faces after blurring are restored to the image.

In a second aspect of the present disclosure, an apparatus for processing multiple human faces in an image is provided. Specifically, the device includes: a detection module configured to detect multiple human faces in an image; a segmentation module configured to perform segmentation processing on the image to separate the multiple human faces from the background of the image; an evaluation module configured to target For a given face among the separated multiple faces, a given evaluation vector is generated based on the attributes of the given face; the processing module is configured to perform corresponding blurring processing on the given face based on the given evaluation vector; and The recovery module restores multiple faces after blurring to the image.

In a third aspect of the present disclosure, there is provided an apparatus comprising one or more processors; and storage means for storing one or more programs, when the one or more programs are executed by the one or more processors Executing such that the one or more processors implement the method according to the first aspect of the present disclosure.

In a fourth aspect of the present disclosure, there is provided a computer-readable medium on which is stored a computer program which, when executed by a processor, implements the method according to the first aspect of the present disclosure.

The embodiments of the present disclosure can perform automatic face blurring processing on large batches of pictures and videos. In addition, the embodiment of the present disclosure reduces the influence of the blurring process on the picture or video quality while performing the face blurring process on multiple faces, thereby improving the viewing effect of the picture or video.

It should be understood that what is described in the Summary of the Invention is not intended to define key or critical features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will be readily understood through the following description.

Description of drawings

The above and other features, advantages and aspects of various embodiments of the present disclosure will become more apparent with reference to the following detailed description when taken in conjunction with the accompanying drawings. In the drawings, identical or similar reference numerals denote identical or similar elements, wherein:

FIG. 1 schematically shows a flowchart of a method for processing multiple human faces in an image according to an embodiment of the present disclosure;

FIG. 2A schematically shows a flowchart of a method for detecting multiple human faces in an image according to an embodiment of the present disclosure;

FIG. 2B schematically shows a flowchart of a method for image segmentation processing according to an embodiment of the present disclosure;

Fig. 3 schematically shows a flowchart of a method for processing multiple human faces in an image according to another embodiment of the present disclosure.

Fig. 4 schematically shows a block diagram of an apparatus for processing multiple human faces in an image according to an embodiment of the present disclosure.

Fig. 5A schematically shows a block diagram of a detection module according to one embodiment of the present disclosure.

FIG. 5B schematically shows a schematic diagram of a segmentation module according to one embodiment of the present disclosure.

Fig. 6 schematically shows a block diagram of an apparatus for processing multiple human faces in an image according to another embodiment.

FIG. 7 shows a block diagram of a computing device capable of implementing various embodiments of the present disclosure.

Detailed ways

Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be construed as limited to the embodiments set forth herein; It is for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and implementations of the present disclosure are for exemplary purposes only, and are not intended to limit the protection scope of the present disclosure.

In the description of the embodiments of the present disclosure, the term "comprising" and its similar expressions should be understood as an open inclusion, ie "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same object. Other definitions, both express and implied, may also be included below.

In image processing, the process of blurring sensitive factors such as faces can be called "desensitization" processing. Traditional desensitization methods usually use professional image processing software to manually blur the face, which requires professionals to use complex and cumbersome image processing software to process, and manually select the area to be processed. However, if there are many people in the image, the traditional processing method requires a very large workload and consumes a lot of time.

In addition, when the traditional face blurring process processes images containing multiple faces, a single blurring method is used for all faces in different orientations and situations, and it does not consider whether the blurred face image is in harmony with the background . As a result, after all the faces of an image with multiple faces are blurred, the blurred area is relatively abrupt, which leads to obvious degradation of the viewing effect of the image.

The present disclosure proposes a solution for desensitizing an image containing multiple human faces. This solution can automatically desensitize a large number of pictures or videos, greatly reducing the workload of desensitization work. In addition, the disclosure also improves the image effect of desensitized images, so that after desensitization of multiple people’s images Can maintain a better viewing effect.

Embodiments of the present disclosure mainly relate to desensitizing images containing multiple human faces. However, it should be understood that the present disclosure is not limited to be used only for the blurring of human faces, and any object requiring desensitization as defined by the spirit and scope of the appended claims of the present disclosure should be considered as belonging to the present disclosure protected range.

Fig. 1 schematically shows a flowchart of a method 100 for processing multiple human faces in an image according to an embodiment of the present disclosure. As shown in FIG. 1 , the process of the method 100 may include blocks 101 - 105 . The processing procedure of the method 100 will be described in detail as an example below.

As shown in FIG. 1, at block 101, a plurality of human faces in an image are detected. This involves collecting images that need to be desensitized, and the images here can be pictures or videos. If it is a video file, the video needs to be parsed, and subsequent processing will be performed according to the video frame. After the images that need to be processed are collected, batch face blurring can be started. Hereinafter, more details about the operations performed at block 101 will be described in detail with reference to FIG. 2A .

FIG. 2A shows a flowchart of a method 200A for detecting multiple human faces in an image according to an embodiment of the present disclosure. Firstly, the face is processed on the image, and the face images of all people in this picture or video frame are detected. Specifically, as shown in FIG. 2A, the process of detecting a human face may include blocks 210-230.

At block 210, an image containing multiple faces is first preprocessed, which may include mean subtraction, normalization, and resizing of the image. Specifically, for the effective operation of the algorithm, it is hoped that the image features have a similar value range, so a normalization operation is required, and the normalization operation may include mean subtraction, normalization, and image size adjustment (resize).

At block 220, detection is performed on the preprocessed image to generate face regions. Here, it is necessary to use the existing face detection algorithm to perform face detection on the image to obtain the face area in the picture or video frame. The face detection algorithm may be a detection algorithm based on a neural network, which includes a process of rough detection of a face, detection of key points of a face, and re-screening of a face. The face detection algorithm can be formed in the form of a module, so that an image can be input into the module and a detection result can be output from the module. The final detection result can be the position of all faces in the image, and use (x, y, w, h) to represent, where x, y represent the vertex coordinates of the face area, w and h represent the width and width of the face area high.

At block 230, the face region is expanded outward by a certain number of pixels. After going through block 220, the location areas of all the faces in the image can be obtained. In order to prevent the lack of human faces caused in the subsequent segmentation process, it can be extended outwards based on the above location area, for example, 5-10 pixels outwards on the four borders of the frame area.

After going through block 101, a plurality of images of human faces are obtained, but these images are only rough human face images, and include certain non-human face backgrounds in the peripheral area of the human faces. Since these backgrounds are not needed in the fixed-point processing of the face, the face image needs to be further segmented.

Returning to FIG. 1 , at block 102 , the image is segmented to separate the plurality of human faces from the background of the image. Specifically, at this block, the face image obtained in block 101 will be accurately segmented to finally obtain a pixel-level face image. In the following, more details regarding the operations performed at block 102 will be described with reference to FIG. 2B .

Fig. 2B shows a flowchart of a method 200B for segmenting an image according to an embodiment of the present disclosure. As shown in FIG. 2B, method 200B includes a number of operations performed at blocks 240-260.

At block 240, the face region expanded in block 230 is pre-processed, including mean subtraction, normalization, and image resizing.

At block 250, the faces in the preprocessed face region are precisely segmented to generate a pixel-level face image.

According to an embodiment of the present disclosure, at block 250, a face segmentation algorithm may be used to perform precise face segmentation. The algorithm is built on the basis of a fully convolutional neural network, and applies technologies such as atrous convolution, multi-scale FeatureMap collection, and fully connected CRF to ensure the accuracy of the range of face fixed-point processing to the greatest extent. The face segmentation algorithm can be formed in the form of modules, so that images can be input into the segmentation module and results can be output from the module.

At block 260, the pixels of the segmented face are recorded. Specifically, in order to facilitate subsequent restoration processing, the pixels of the segmented face are recorded in the file.

Returning to FIG. 1 , at block 103 , for a given face among the multiple separated faces, a given evaluation vector is generated based on attributes of the given face.

Specifically, after the segmentation processing at block 102, an accurately segmented human face image is obtained. Then, an evaluation needs to be performed on a given face among multiple faces and an evaluation vector is generated. The given face may be a single face or a combination of several faces.

According to an embodiment of the present disclosure, there are multiple evaluation dimensions of the evaluation vector, such as sharpness, resolution, face angle, image brightness, image contrast, etc., wherein the face angle includes roll angle (roll), pitch angle (pitch ) and yaw angle (yaw). In addition, any other face attributes may be set as desired. The aforementioned attributes can be expressed as a 1×7 vector, such as [1,1,0.8,0.9,0.8,1,1], where each dimension corresponds to one of the aforementioned attributes.

According to an embodiment of the present disclosure, the definition, resolution, face angle, image brightness, and image contrast can be calculated in the following manner:

Clarity: The resolution and video bit rate can be used to weight and calculate the clarity of the video, and normalize it to [0,1];

Resolution: It can be obtained directly through the image, and then normalize all resolutions to [0,1];

Face angle: use open-source algorithms to directly calculate the angle values of the three angles (roll, pitch, yaw) of the face, and then convert them into radian values and normalize them;

Image brightness: Randomly take 100 points in the image to calculate the mean value and normalize it;

Image contrast: can be based on the formula Obtained, where δ(i, j)=|ij|, which represents the gray level difference between adjacent pixels, P _δ (i, j) is the distribution probability of the pixel whose gray level difference between adjacent pixels is δ.

As shown in FIG. 1 , at block 104 , based on a given evaluation vector, a corresponding blurring process is performed on a given human face. Specifically, different levels of blurring are performed on each face image in the plurality of faces according to the evaluation vector.

According to an embodiment of the present disclosure, each dimension in the evaluation vector can be involved in the calculation of the blurring process, so as to obtain a weighted coefficient, and the face image is processed according to the weighted coefficient. For example, if a face image with a large head-up angle, poor definition and low image brightness is compared with an image with a relatively straight angle, good definition and normal image brightness, the former will get different weighting coefficients from the latter, and then This allows the former to require only light image processing, while the latter requires a higher level of processing.

According to an embodiment of the present disclosure, the blurring processing includes, for example, Gaussian filtering, contrast enhancement, image shaking, coding and other image processing means.

As shown in FIG. 1 , at block 105 , the blurred faces are restored to the image. Specifically, after the face image is processed at block 104, each face will be restored to the original image.

According to an embodiment of the present disclosure, during the aforementioned segmentation processing (see block 260), the segmented face pixels are recorded, and the processed face image can be restored to its original in the image.

If the initial input is a picture, the processing process ends here; if it is a video, it is also necessary to perform the same processing on each video frame along with the analysis of the video, and finally synthesize the processed video frames into a video.

Fig. 3 shows a flowchart of a method 300 for processing multiple human faces in an image according to another embodiment of the present disclosure. As shown in FIG. 3 , the method 300 includes blocks 101 - 103 , 105 , 304 , and further includes a block 306 . The difference between the method 300 shown in FIG. 3 and the method 100 shown in FIG. 1 is that a block 306 is added between the block 102 and the block 103, and the processing method at the block 304 is carried out on the basis of the block 104 in FIG. 1 Adjust accordingly.

As shown in Figure 3, at block 306, determine the deviation rate of the image, which indicates the difference between the multiple faces and the background of the image; and at block 304, in addition to the given evaluation vector, it is also necessary to The corresponding blurring process is performed on a given face.

According to an embodiment of the present disclosure, the deviation rate is obtained from the original image, and the deviation rate can indicate the difference between multiple human faces and the background of the image, so as to control the intensity of subsequent image processing. For example, if the deviation rate is 1, that is, no deviation restriction is performed, and if the deviation rate is 0.8, it means that the strength of the face blurring process should be reduced to 0.8 times of the original. By setting the deviation rate, it can ensure that the processed face image will not conflict with the background to the greatest extent. For example, if the background light in the original image to be processed is obviously too dark compared with the background light under normal conditions, the blur effect can be adjusted through the deviation rate, so that the blurred face image as a whole also presents a darker effect, In order to match the dark background, the processed face image is not too abrupt compared with the image background.

According to an embodiment of the present disclosure, the deviation rate can be determined by: normalizing the pixel values of the image; selecting a plurality of sampling points from a plurality of faces and the background of the image based on a predetermined probability distribution; The sampling points determine the deviation rate. For example, first normalize the pixel values of all pixels in the picture to [-1,1], then randomly collect N points from the image with a probability of p using a Gaussian distribution with a mean of 0 and a variance of 1 and obtain each The pixel value of a point, and then obtain the weighted mean by weighted average, the value of the weighted mean is still in [-1,1], and then perform the absolute value operation on the weighted mean, and finally subtract it from 1. Absolute value to calculate the deviation rate, where p and N are preset parameters.

The above-mentioned embodiments of the present disclosure can perform automatic face blurring processing on large batches of pictures and videos through automatic face recognition, segmentation and cropping, blurring, and restoration. In addition, the embodiment of the present disclosure also evaluates each face and establishes an evaluation vector to distinguish each face, and sets the deviation rate according to the image situation to reduce the conflict between the processed face and the background, so that when multiple faces While the face is being blurred, the impact on the picture or video is minimized, and the viewing effect of the picture or video is improved.

Fig. 4 shows a block diagram of an apparatus 400 for processing multiple human faces in an image according to an embodiment of the present disclosure. The apparatus 400 includes modules 401-405. Each module of the device 400 will be described exemplarily below.

As shown in FIG. 4 , module 401 is a detection module configured to detect multiple human faces in an image. Hereinafter, more details of the module 401 will be described in detail with reference to FIG. 5A .

Fig. 5A shows a schematic block diagram of a detection module according to an embodiment of the present disclosure. As shown in FIG. 5A , the detection module 500A (ie, the detection module 401 in FIG. 4 ) may include submodules 510-530.

The sub-module 510 is a first pre-processing module, which is configured to pre-process the image containing multiple faces, and the pre-processing may include subtracting the mean value, normalizing and re-adjusting the size of the image.

The sub-module 520 is a detection sub-module configured to detect the preprocessed image to generate a human face area.

The sub-module 530 is an expansion sub-module, which is configured to expand the face area outward by a certain number of pixels.

Multiple face images are obtained through the detection module 401 . Since these images contain certain non-face backgrounds, it is necessary to further segment the face images.

Returning to Fig. 4, module 402 is a segmentation module. The segmentation module is configured to perform segmentation processing on the image to separate the plurality of human faces from the background of the image. Specifically, the segmentation module 402 will perform accurate face segmentation on the face image obtained from the detection module 401 to finally obtain a pixel-level face image. Hereinafter, more details of the module 402 will be described in detail with reference to FIG. 5B .

Fig. 5B shows a schematic block diagram of a segmentation module according to an embodiment of the present disclosure. As shown, segmentation module 500B (ie, detection module 402 in FIG. 4 ) includes submodules 540-560.

The sub-module 540 is a second pre-processing sub-module, which is configured to perform pre-processing on the expanded face area, the pre-processing includes mean subtraction, normalization and re-adjustment of image size.

The sub-module 550 is an accurate segmentation sub-module, which is adapted to accurately segment the human face in the pre-processed human face area to generate a pixel-level human face image.

The sub-module 560 is a recording sub-module configured to record the pixels of the segmented face. Specifically, in order to facilitate subsequent restoration processing, the pixels of the segmented face are recorded in the file.

Returning to FIG. 4 , module 403 is an evaluation module, which is configured to generate a given evaluation vector for a given face among the multiple separated faces based on attributes of the given face.

Specifically, after the segmentation processing in module 402, an accurately segmented face image is obtained. Then, each of the plurality of faces needs to be evaluated and an evaluation vector is generated.

According to an embodiment of the present disclosure, there are multiple evaluation dimensions of the evaluation vector, such as sharpness, resolution, face angle, image brightness, image contrast, etc., wherein the face angle includes roll angle (roll), pitch angle (pitch ) and yaw angle (yaw).

As shown in FIG. 4 , module 404 is a processing module configured to perform corresponding blurring processing on a given face based on a given evaluation vector. Specifically, different levels of blurring are performed on each face image in the plurality of faces according to the evaluation vector.

According to an embodiment of the present disclosure, the blurring processing includes, for example, Gaussian filtering, contrast enhancement, image shaking, coding and other image processing means.

As shown in FIG. 4 , the module 405 is a restoration module, which is configured to restore the blurred faces into the image. Specifically, after the face image is processed at the processing module 404 , each face will be restored to the original image at the restoration module 405 .

According to an embodiment of the present disclosure, the segmented face pixels are recorded during the aforementioned segmentation processing, and the processed face image can be restored to the original image according to the recorded data (eg, segmentation location file).

Fig. 6 shows a block diagram of an apparatus 600 for processing multiple human faces in an image according to another implementation manner. As shown in the figure, the apparatus 600 includes modules 401 , 402 , 606 , 403 , 604 and 405 . The difference between the apparatus 600 shown in FIG. 6 and the apparatus 400 shown in FIG. 4 is that a deviation rate calculation module 606 is added, and a processing module is adjusted accordingly (ie, the processing module 404 in FIG. 4 is replaced with a module 604).

The increased deviation rate calculation module 606 is configured to determine the deviation rate of the image, the deviation rate indicates the difference between the plurality of faces and the background of the image; and the new processing module 604 is configured to: in addition to based on the given evaluation vector, The given face is also blurred accordingly based on the deviation rate.

According to an embodiment of the present disclosure, the deviation rate is obtained from the original image, and the deviation rate can indicate the difference between multiple human faces and the background of the image, so as to control the intensity of subsequent image processing.

According to an embodiment of the present disclosure, the deviation rate can be determined by: normalizing the pixel values of the image; selecting a plurality of sampling points from a plurality of faces and the background of the image based on a predetermined probability distribution; The sampling points determine the deviation rate.

The apparatus 400 in FIG. 4 and the apparatus 600 in FIG. 6 respectively correspond to the method 100 in FIG. 1 and the method 300 in FIG. 3 . Therefore, the description for methods 100 and 300 is equally applicable to apparatuses 400 and 600 .

According to an exemplary implementation of the present disclosure, there is provided an apparatus including one or more processors; and a storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method according to the present disclosure.

According to an exemplary implementation of the present disclosure, there is provided a computer-readable medium on which a computer program is stored, and when the program is executed by a processor, the method according to the present disclosure is implemented.

FIG. 7 shows a block diagram of a computing device 700 capable of implementing various embodiments of the present disclosure. As shown, the device 700 includes a central processing unit (CPU) 701 that can be programmed according to computer program instructions stored in a read-only memory (ROM) 702 or loaded from a storage unit 708 into a random-access memory (RAM) 703 program instructions to perform various appropriate actions and processes. In the RAM 703, various programs and data necessary for the operation of the device 700 can also be stored. The CPU 701 , ROM 702 , and RAM 703 are connected to each other via a bus 704 . An input/output (I/O) interface 705 is also connected to the bus 704 .

Multiple components in the device 700 are connected to the I/O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a magnetic disk, an optical disk, etc. ; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, and the like. The communication unit 709 allows the device 700 to exchange information/data with other devices over a computer network such as the Internet and/or various telecommunication networks.

The processing unit 701 executes various methods and processes described above, such as methods 100 , 200A, 200B and/or method 300 . For example, in some implementations, methods 100 , 200A, 200B, and/or method 300 may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as storage unit 708 . In some embodiments, part or all of the computer program may be loaded and/or installed on the device 700 via the ROM 702 and/or the communication unit 709 . When the computer program is loaded into RAM 703 and executed by CPU 701 , one or more blocks of methods 100 , 200A, 200B and/or method 300 described above may be performed. Alternatively, in other implementations, the CPU 701 may be configured to execute the methods 100 , 200A, 200B and/or the method 300 in any other suitable manner (for example, by means of firmware).

The functions described herein above may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate array (FPGA), application specific integrated circuit (ASIC), application specific standard product (ASSP), system on a chip (SOC), load programmable logic device (CPLD), etc.

Program code for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special purpose computer, or other programmable data processing devices, so that the program codes, when executed by the processor or controller, make the functions/functions specified in the flow diagrams and/or block diagrams Action is implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media would include one or more wire-based electrical connections, portable computer discs, hard drives, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, compact disk read only memory (CD-ROM), optical storage, magnetic storage, or any suitable combination of the foregoing.

In addition, while operations are depicted in a particular order, this should be understood to require that such operations be performed in the particular order shown, or in sequential order, or that all illustrated operations should be performed to achieve desirable results. Under certain circumstances, multitasking and parallel processing may be advantageous. Likewise, while the above discussion contains several specific implementation details, these should not be construed as limitations on the scope of the disclosure. Certain features that are described in the context of separate implementations can also be implemented in combination in a single implementation. Conversely, various features that are described in the context of a single implementation can also be implemented in multiple implementations separately or in any suitable subcombination.

Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are merely example forms of implementing the claims.

Claims (18)

1. A method for processing multiple faces in an image, comprising: detecting a plurality of human faces in the image; performing segmentation processing on the image to separate the plurality of human faces from the background of the image; For a given face among the separated faces, generating a given evaluation vector based on attributes of the given face; performing corresponding blurring processing on the given face based on the given evaluation vector; and Restoring the blurred faces to the image. 2. The method according to claim 1, wherein the attributes of the given face include at least any one of the following: sharpness, resolution, face angle, image brightness, and image Contrast, wherein the face angle includes roll angle, pitch angle and yaw angle. 3. The method of claim 1, further comprising: determining a deviation rate of the image, the deviation rate indicating a difference between the plurality of human faces and the background of the image; and Perform corresponding blurring processing on the given face based on the deviation rate. 4. The method of claim 3, wherein said determining a rate of deviation of said image comprises: normalizing the pixel values of the image; selecting a plurality of sampling points from the plurality of faces and the background of the image based on a predetermined probability distribution; and The deviation rate is determined based on the selected plurality of sampling points. 5. The method according to claim 1, wherein said detecting a plurality of human faces in an image comprises: performing a first preprocessing on the image, said first preprocessing comprising subtracting the mean, normalizing and resizing the image; Perform detection on preprocessed images to generate face regions; and Extend the face area outward by a certain number of pixels. 6. The method according to claim 5, wherein said image segmentation process comprises: Performing a second preprocessing on the expanded face area, the second preprocessing includes subtracting the mean value, normalizing and readjusting the image size; Accurately segment faces in the preprocessed face region to generate pixel-level face images; and Record the pixels of the segmented face. 7. The method according to claim 6, wherein said reverting to an image comprises: Restoring the human face into the image based on the recorded pixels. 8. The method according to claim 1, wherein the blurring process comprises at least any one of the following: Gaussian filtering, contrast enhancement, image dithering and coding. 9. A device for processing multiple faces in an image, comprising: a detection module configured to detect a plurality of human faces in the image; a segmentation module configured to perform segmentation processing on the image to separate the plurality of human faces from the background of the image; An evaluation module configured to generate a given evaluation vector based on attributes of the given face for a given face among the separated multiple faces; A processing module configured to perform corresponding blurring processing on the given face based on the given evaluation vector; and The recovery module recovers the blurred faces into the image. 10. The device according to claim 9, wherein the attributes of the given face include at least any one of the following: sharpness, resolution, face angle, image brightness, and image Contrast, wherein the face angle includes roll angle, pitch angle and yaw angle. 11. The apparatus of claim 9, further comprising: a deviation rate calculation module configured to determine a deviation rate of the image, the deviation rate indicating a difference between the plurality of human faces and the background of the image; and The processing module is further configured to perform corresponding blurring processing on the given face based on the deviation rate. 12. The apparatus of claim 11, wherein said determining a rate of deviation of said image comprises: normalizing the pixel values of the image; selecting a plurality of sampling points from the plurality of faces and the background of the image based on a predetermined probability distribution; and The deviation rate is determined based on the selected plurality of sampling points. 13. The apparatus of claim 9, wherein the detection module comprises: The first pre-processing sub-module is configured to perform a first pre-processing on the image, the first pre-processing includes subtracting the mean value, normalizing and re-adjusting the image size; A detection submodule configured to detect the preprocessed image to generate a face area; and The extension submodule is configured to extend the face area outward by a certain number of pixels. 14. The apparatus of claim 13, wherein the segmentation module comprises: The second preprocessing submodule is configured to perform a second preprocessing on the expanded face area, the second preprocessing includes subtracting the mean value, normalizing and readjusting the image size; an accurate segmentation submodule configured to accurately segment the human face in the preprocessed human face region to generate a pixel-level human face image; and The recording submodule is configured to record the pixels of the segmented face. 15. The apparatus of claim 14, wherein said reverting to an image comprises: Restoring the human face into the image based on the recorded pixels. 16. The device according to claim 9, wherein the blurring process comprises at least any one of the following: Gaussian filtering, contrast enhancement, image dithering and coding. 17. A device, said device comprising: one or more processors; and A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, so that the one or more processors implement the process according to any one of claims 1-8 The method for processing multiple human faces in an image described in the item. 18. A computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the method for processing multiple human faces in an image according to any one of claims 1-8 is realized .