A method and device for recovering a high-resolution face image from a low-resolution image

Abstract

The application discloses a method and device for restoring high-resolution face images from low-resolution face images, and belongs to the technical field of face image processing. In order to solve the problem that the resolution of collected images is low, and it is not easy to determine the blurred identity features of faces, when the distance between a collection device and a person is far, or the face activity amplitude is large during the collection process, and part of the people on the spot wear sunglasses, masks or other things carried by themselves, resulting in incomplete face collection, the application can control the intermittent rotation of an image collection device and a camera probe through the continuous rotation of a guide turntable, so that the face image can be collected when the image collection device and the camera probe stop. At the same time, the stability of the image collection device and the camera probe during the rotation process is improved through the connection of a limiting rotation shaft column and a limiting rotation slot hole, and the connection of a limiting connecting piece and an annular limiting slot. The face processing modes of four situations are mutually fused and assisted, so that the low-resolution face images can be maximally restored.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of face image processing, in particular to a method and device for restoring high-resolution face image from low-resolution face image. BACKGROUND

[0002] Face image processing refers to a computer technology for identity authentication by analyzing and comparing human visual feature information, which can be used in security verification systems, medical records management, bank and customs monitoring systems, and automatic access control systems, etc. Compared with other methods for identity recognition using fingerprints, irises and other human biological characteristics, face recognition is more friendly, convenient and concealed. Due to its huge application prospect and incomparable advantages, face recognition has become a hot spot in the field of current pattern recognition and artificial intelligence. Image preprocessing is an important link in the process of face recognition. Due to different image acquisition environments, input images often have noise, insufficient contrast and other shortcomings.

[0003] According to the application number CN105654033B, a face image verification method and device are known. The method includes receiving a face image to be verified, extracting a first face feature of the face image to be verified, and verifying the face image to be verified represented by the first face feature according to a second face feature corresponding to a registered face image and a third face feature corresponding to a face album grouping bound to the registered face image. This can effectively avoid the adverse effects of different factors such as light, expression, hairstyle and clothing on the accuracy of the verification result, and can improve the accuracy of the verification result. In large activity venues, many people come and go. In order to ensure the safety of personnel flow, the faces of each person are collected and recorded. However, the space of the on-site venue is large, and the images intercepted cannot collect the front faces of all the people. On the one hand, people who are far away from the collection device, or whose faces move greatly during the collection process, have low image resolution, and it is not easy to determine the identity features of the blurred face. On the other hand, some people on site wear sunglasses, masks or other personal belongings, which makes the face collection incomplete and it is difficult to determine the identity of the person.

[0004] In view of the above problems, the existing device is improved, and a method and device for restoring high-resolution face image from low-resolution face image are proposed. SUMMARY

[0005] The purpose of the present application is to provide a method and device for recovering a high-resolution face image from a low-resolution image, which solves the problem of determining the identity of a person whose face is blurred in a low-resolution image collected from a person who is far away from the collection device or whose face moves greatly during the collection process, and the problem of determining the identity of a person whose face is not completely collected due to wearing sunglasses, a mask or other personal items.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical solution: a method for recovering a high-resolution face image from a low-resolution image, comprising the following steps:

[0007] S1: image collection: the scanned image is taken one by one by an image collection device;

[0008] S2: face cutout: the corresponding face is cut out from the image by edge separation, and the soft edge problem of the face image separation is processed by the Laplace operator and the graph structure is constructed;

[0009] S3: face enlargement: the cutout of the face is enlarged for further processing;

[0010] S4: face anti-aliasing: the image anti-aliasing technology is used, the edge part of the face becomes smooth, and the aliasing is eliminated;

[0011] S5: face super-resolution: the high-resolution version of the image is predicted by analyzing the face image data, and the face is high-definition;

[0012] S6: face denoising: the noise in the image is eliminated by filtering and normalizing the image data, and thus all implementation steps are completed.

[0013] Further, the image collection device includes a face region occlusion processing module for processing face occlusion parts, a face symmetry cutting and completing module for completing face symmetry missing, and a same region cutting and copying completing module for completing non-feature parts of the face.

[0014] Further, the face region occlusion processing module includes a face occlusion contour determining element for determining the face occlusion contour, a face occlusion part smearing element for smearing the face occlusion part, a same color system covering smearing area element for covering the smearing area, a face skeleton contour shaping element for shaping according to the face contour, a feature part completing element for completing the feature part, and a face image recovering element for recovering the face image.

[0015] The utility model provides a kind of low-resolution restores high-resolution face image's device, including image acquisition device and the camera probe being arranged in the side surface of image acquisition device, the bottom end of image acquisition device is equipped with the placing base plate, the lower bottom surface of image acquisition device is equipped with the first rotating column, the upper surface of placing base plate is equipped with the second rotating column, the first rotating column and the second rotating column are opposite, and the first rotating column and the second rotating column size match.

[0016] Further, the first rotating column includes a first cylinder disposed on the lower surface of the image acquisition device and a movable recessed chamber opened on the surface of the first cylinder. The movable recessed chamber is open downward. An annular limiting groove is opened on the outer side of the movable recessed chamber. A guide disc is disposed on the inner bottom surface of the movable recessed chamber. An annular tooth surface is disposed on the inner side wall of the guide disc. A limiting rotating shaft column is installed at the center of the inner bottom surface of the guide disc. The first rotating column is connected to the second rotating column through the limiting rotating shaft column.

[0017] The second rotating column includes a second cylinder disposed on the upper surface of the placing base plate and a limiting rotating slot hole opened at the center of the second cylinder. A guide rotating disc is installed on the side end of the upper surface of the second cylinder. The guide rotating disc is disposed on the inner side of the guide disc. A driving rotating shaft is installed at the center of the guide rotating disc. The guide rotating disc is connected to the surface of the second cylinder through the driving rotating shaft. An arc-shaped protrusion is installed on the side surface of the guide rotating disc. An arc-shaped tooth surface is disposed on the outer side surface of the arc-shaped protrusion. The guide rotating disc is connected to the annular tooth surface through the arc-shaped tooth surface. A limiting connector is disposed on the circumferential end of the upper surface of the second cylinder. The second cylinder is connected to the annular limiting groove through the limiting connector.

[0018] Compared with the prior art, the utility model has the following beneficial effects:

[0019] 1.The method and device for recovering high-resolution face image from low-resolution face image are provided, and the method includes the following steps: in the first case, when the resolution of the collected image is low, the low-resolution face image is recovered to normal resolution through face matting, face enlargement, face anti-aliasing, face super-resolution and face denoising, so as to identify the identity; in the second case, when the face image collected by the camera probe is missing in the feature part, the contour of the face shielding area is first delineated to define the processing range, when the cheek area is shielded, the shielding part is smeared through the face shielding part smearing element, and the symmetric part of the face is intercepted to fill the covering smearing area through the same color system covering smearing element, so as to restore the symmetry, when the face contour area is shielded, the face shielding contour element is determined to repair and restore the face according to the comprehensive contour of the face, when the feature part of the face is missing, such as eyes, eyebrows, nose and mouth, if the missing part is the symmetric facial feature such as eyes and eyebrows, the symmetric copying repair module is used for symmetric copying repair, if the missing part is the single facial feature such as nose and mouth, secondary image collection is needed to obtain the complete face image; in the third case, when the collected face image is missing in a large range, the face is divided by a division line, and the left and right faces are symmetrically repaired according to the division line and the missing condition; in the fourth case, when the face is missing in a large range, the same color skin is intercepted to repair and restore the missing shielding part, and the face processing modes of the four cases are mutually integrated and assisted to maximize the restoration of the low-resolution face image.

[0020] 2.The method and device for recovering high-resolution face image from low-resolution face image are provided, and the method includes the following steps: through the continuous rotation of the guide turntable, the intermittent rotation of the image collection device and the camera probe can be controlled, so that the face image can be collected when the image collection device and the camera probe stop, and at the same time, the stability of the image collection device and the camera probe in the rotation process is improved through the connection of the limiting rotation shaft column and the limiting rotation slot hole and the connection of the limiting connecting element and the annular limiting slot. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 It is a whole flowchart of the method for recovering high-resolution face image from low-resolution face image.

[0022] Figure 2 It is a whole structure diagram of the device for recovering high-resolution face image from low-resolution face image.

[0023] Figure 3 It is a system block diagram of the method for recovering high-resolution face image from low-resolution face image.

[0024] Figure 4The structure schematic diagram of the face area shielding processing module of the method for recovering high-resolution face images from low-resolution images according to the present application;

[0025] Figure 5 The structure schematic diagram of the assembly of the first rotating column and the second rotating column of the device for recovering high-resolution face images from low-resolution images according to the present application;

[0026] Figure 6 The structure schematic diagram of the first rotating column of the device for recovering high-resolution face images from low-resolution images according to the present application;

[0027] Figure 7 The structure schematic diagram of the second rotating column of the device for recovering high-resolution face images from low-resolution images according to the present application;

[0028] Figure 8 The structure schematic diagram of the limiting connecting piece of the device for recovering high-resolution face images from low-resolution images according to the present application.

[0029] In the figure: 1, image acquisition device; 11, face area shielding processing module; 111, face shielding contour determining element; 112, face shielding part smearing element; 113, same color system covering smearing area element; 114, face skeleton contour shaping element; 115, feature part completing element; 116, recovered face image element; 12, face symmetry intercepting and completing module; 13, same area intercepting and copying completing module; 2, camera probe; 3, placement base plate; 4, first rotating column; 41, first column body; 42, movable concave cavity; 43, annular limiting groove; 44, guide disc; 45, annular tooth surface; 46, limiting rotating shaft column; 5, second rotating column; 51, second column body; 52, limiting rotating groove hole; 53, guide rotating disc; 54, driving rotating shaft; 55, arc-shaped protrusion; 56, arc-shaped tooth surface; 57, limiting connecting piece. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0031] Please refer to Figures 1-8 In order to solve the technical problem that the resolution of the collected image is low, it is difficult to determine the blurred identity features of the face, and part of the personnel on the spot wears sunglasses, masks or other things carried by the body, causing the face collection to be incomplete and it is difficult to determine the identity of the personnel, the following preferred technical solutions are provided:

[0032] A method for recovering high-resolution human face images from low-resolution images, comprising the following steps:

[0033] S1: image acquisition: the scanned images are captured one by one by the image acquisition device 1;

[0034] S2: face cutout: the corresponding face is cut out from the image by edge separation, and the soft edge problem of the face image separation is processed by Laplace operator and the graph structure is constructed;

[0035] S3: face enlargement: the cutout of the face is enlarged for further processing;

[0036] S4: face anti-aliasing: using image anti-aliasing technology, the edge part of the face becomes smooth, and the aliasing is eliminated;

[0037] S5: face super-resolution: by analyzing the face image data, the high-resolution version of the image is predicted, and the face is high-definition;

[0038] S6: face denoising: by filtering and normalizing the image data, the noise in the image is eliminated, and thus all the implementation steps are completed.

[0039] The image acquisition device 1 includes a face area occlusion processing module 11 for processing face occlusion parts, a face symmetry cutting completion module 12 for completing face symmetry missing, and a same area cutting copy completion module 13 for completing non-feature parts of the face. The face area occlusion processing module 11 includes a face occlusion contour determining element 111 for determining the face occlusion contour, a face occlusion part smearing element 112 for smearing the face occlusion part, and a same color system covering smearing area element 113 for covering the smearing area. The face area occlusion processing module 11 further includes a face skeleton contour shaping element 114 shaped according to the face contour, a feature part completion element 115 for completing the feature part, and a face image recovery element 116 for recovering the face image.

[0040] Specifically, the first case: when the resolution of the collected image is low, the low-resolution face image is restored to normal resolution by face cutout, face enlargement, face anti-aliasing, face super-resolution and face denoising for identity recognition; the second case: when the face image collected by the camera probe 2 has missing feature parts, the contour of the face occlusion area is first delineated to define the processing range, when the cheek area is occluded, the occluded part is smeared by the face occlusion part smearing element 112, and the symmetric part of the face is intercepted to fill the covered smearing area by the same color system covering smearing area element 113 for symmetry restoration, when the face contour area is occluded, the face occlusion contour element 111 is used to repair and restore the face according to the comprehensive contour of the face, when the feature parts of the face are missing, such as eyes, eyebrows, nose and mouth, if the missing parts are eyes and eyebrows, which are symmetric facial features, the face symmetry interception repair module 12 is used for symmetric copying repair, if the missing parts are nose and mouth, which are single facial features, secondary image acquisition is needed to obtain complete face image; the third case: when the collected face image has large range missing, the face is first divided by a division line, and the left and right faces are symmetrically repaired according to the division line and the missing condition; the fourth case: when the face has large range missing, the same color skin is intercepted to repair and restore the missing occluded part, the face processing methods of the four cases are mutually integrated and assisted to maximize the restoration of low-resolution face image.

[0041] In order to solve the technical problems of smoothness and time interval during face image collection, the following preferred technical solutions are provided:

[0042] A device for restoring high-resolution face image from low-resolution face image, comprising an image collection device 1 and a camera probe 2 arranged on the side surface of the image collection device 1, the bottom end of the image collection device 1 is provided with a placing base 3, the lower bottom surface of the image collection device 1 is provided with a first rotating column 4, the upper surface of the placing base 3 is provided with a second rotating column 5, the first rotating column 4 and the second rotating column 5 are connected, and the sizes of the first rotating column 4 and the second rotating column 5 are matched, the first rotating column 4 can drive the image collection device 1 and the camera probe 2 to rotate, and the intermittent rotation of the image collection device 1 and the camera probe 2 is controlled to collect face image.

[0043] The first rotating column 4 comprises a first column body 41 arranged on the lower surface of the image acquisition device 1 and a movable concave cavity 42 formed on the surface of the first column body 41, the movable concave cavity 42 is open downward, an annular limiting groove 43 is formed on the outer side of the movable concave cavity 42, a guide disc 44 is arranged on the inner bottom surface of the movable concave cavity 42, an annular tooth surface 45 is arranged on the inner side wall of the guide disc 44, a limiting rotating shaft column 46 is arranged on the inner bottom surface of the guide disc 44, and the first rotating column 4 is connected with the second rotating column 5 through the limiting rotating shaft column 46.

[0044] The second rotating column 5 comprises a second column body 51 arranged on the upper surface of the placing disc 3 and a limiting rotating groove hole 52 formed in the center of the second column body 51, a guide rotating disc 53 is arranged on the side end of the upper surface of the second column body 51, the guide rotating disc 53 is arranged on the inner side of the guide disc 44, a driving rotating shaft 54 is arranged in the center of the guide rotating disc 53, a driving motor is arranged in the driving rotating shaft 54, the guide rotating disc 53 is connected with the surface of the second column body 51 through the driving rotating shaft 54, an arc-shaped protrusion 55 is arranged on the side surface of the guide rotating disc 53, an arc-shaped tooth surface 56 is arranged on the outer side surface of the arc-shaped protrusion 55, the guide rotating disc 53 is connected with the annular tooth surface 45 through the meshing connection of the arc-shaped tooth surface 56, and a limiting connecting piece 57 is arranged on the circumferential end of the upper surface of the second column body 51.

[0045] Specifically, the driving rotating shaft 54 is started to drive the guide rotating disc 53 to rotate on the upper surface of the second column body 51, the guide rotating disc 53 is arranged on the inner side of the guide disc 44, and there is a certain gap between the guide rotating disc 53 and the inner wall of the guide disc 44, with the circumferential movement of the arc-shaped protrusion 55, when the arc-shaped protrusion 55 contacts with the inner wall of the guide disc 44, the meshing connection of the arc-shaped tooth surface 56 and the annular tooth surface 45 is utilized, the first column body 41 is driven to rotate as a whole according to the fixed connection between the guide disc 44 and the first column body 41, the image acquisition device 1 and the camera probe 2 are driven to rotate by the first rotating column 4, when the arc-shaped protrusion 55 is separated from the inner wall of the guide disc 44, the image acquisition device 1 and the camera probe 2 stop rotating, and the intermittent rotation of the image acquisition device 1 and the camera probe 2 can be controlled through the continuous rotation of the guide rotating disc 53, when the image acquisition device 1 and the camera probe 2 stop, the face image can be collected, and meanwhile, the stability of the image acquisition device 1 and the camera probe 2 in the rotating process is improved through the connection of the limiting rotating shaft column 46 and the limiting rotating groove hole 52 and the connection of the limiting connecting piece 57 and the annular limiting groove 43.

[0046] The above is only the preferred specific embodiment of the present application, but the protection scope of the present application is not limited to this, any person skilled in the art can make equivalent replacement or change according to the technical scheme and the inventive concept of the present application within the technical range disclosed by the present application, which should be covered in the protection scope of the present application.

Claims

1. A method for recovering a high-resolution human face image from a low-resolution image, characterized in that: Includes the following steps: S1: Image acquisition: The device that recovers high-resolution facial images from low resolution captures each segment of the scanned image; S2: Face matting: The corresponding face is extracted from the image by edge segmentation, and the soft edge problem of face image segmentation is handled by the Laplacian operator and the graphic structure is constructed. S3: Face Enlargement: Enlarges the cutout of the face for further processing; S4: Face Anti-aliasing: Using image anti-aliasing technology, the edges of the face are smoothed to eliminate jagged edges; S5: Face Super-Resolution: By analyzing face image data, it predicts a high-resolution version of the image, promoting face high-definition. S6: Face Denoising: By filtering and normalizing the image data, noise in the image is eliminated. This completes all implementation steps. The device for recovering high-resolution face images from low resolution includes an image acquisition device (1) and a camera probe (2) disposed on the side surface of the image acquisition device (1). A placement base (3) is installed at the bottom of the image acquisition device (1). A first rotating column (4) is disposed on the bottom surface of the image acquisition device (1). A second rotating column (5) is disposed on the upper surface of the placement base (3). The first rotating column (4) and the second rotating column (5) are connected to each other, and the dimensions of the first rotating column (4) and the second rotating column (5) are matched. The first rotating column (4) includes a first column (41) disposed on the lower surface of the image acquisition device (1) and a movable recessed cavity (42) opened on the surface of the first column (41). The movable recessed cavity (42) opens downward, and an annular limiting groove (43) is provided on the outer side of the movable recessed cavity (42). A guide plate (44) is provided on the inner bottom surface of the movable recessed cavity (42). An annular toothed surface (45) is provided on the inner side wall of the guide plate (44). A limiting rotating shaft (46) is installed in the center of the inner bottom surface of the guide plate (44). The first rotating column (4) is connected to the second rotating column (5) through the limiting rotating shaft (46). The second rotating column (5) includes a second column (51) disposed on the upper surface of the placement chassis (3) and a limiting rotating slot (52) opened in the center of the second column (51). A guide turntable (53) is installed on the side end of the upper surface of the second column (51). The guide turntable (53) is disposed inside the guide plate (44). A drive shaft (54) is installed in the center of the guide turntable (53). The guide turntable (53) is connected to the surface of the second column (51) through the drive shaft (54). An arc-shaped protrusion (55) is installed on the side surface of the guide turntable (53). An arc-shaped tooth surface (56) is provided on the outer surface of the arc-shaped protrusion (55). The guide turntable (53) is meshed with the annular tooth surface (45) through the arc-shaped tooth surface (56). A limiting connector (57) is provided at the circumferential end of the upper surface of the second column (51). The second column (51) is connected to the annular limiting slot (43) through the limiting connector (57).

2. The method for recovering a high-resolution face image from a low-resolution image as described in claim 1, characterized in that: The image acquisition device (1) includes a face region occlusion processing module (11) for processing face occlusion parts, a face symmetry cropping and completion module (12) for completing face symmetry missing parts, and a same region cropping and copying completion module (13) for completing non-feature parts of the face.

3. The method for recovering a high-resolution face image from a low-resolution image as described in claim 2, characterized in that: The face region occlusion processing module (11) includes a face occlusion contour determination element (111) for determining the face occlusion contour and a face occlusion part smearing element (112) for smearing the face occlusion part, and a same color cover smearing element (113) for covering the smearing area. The face region occlusion processing module (11) also includes a face bone contour shaping element (114) shaped according to the face contour and a feature part completion element (115) for completing the feature parts, and a face image recovery element (116) for recovering the face image.

Images