CN108629333A - A kind of face image processing process of low-light (level), device, equipment and readable medium - Google Patents
A kind of face image processing process of low-light (level), device, equipment and readable medium Download PDFInfo
- Publication number
- CN108629333A CN108629333A CN201810516307.5A CN201810516307A CN108629333A CN 108629333 A CN108629333 A CN 108629333A CN 201810516307 A CN201810516307 A CN 201810516307A CN 108629333 A CN108629333 A CN 108629333A
- Authority
- CN
- China
- Prior art keywords
- image
- face
- dimensional
- images
- dimension human
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
- G06V40/166—Detection; Localisation; Normalisation using acquisition arrangements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention provides a kind of face image processing process of low-light (level), device, equipment and readable medium, this method to include:Each frame in video is handled, n two-dimension human face images for belonging to same person are extracted;Then n two-dimension human face images are generated into n three-dimensional face images according to three-dimensional face model, and n three-dimensional face images superpositions is generated into a mean value three-dimensional face images;Mean value three-dimensional face images are mapped into two-dimensional space again and obtain a mean value two-dimension human face image;The facial image that suitable brightness and comparison are obtained after image enhancement processing is done to mean value two-dimension human face image.The present invention carries out noise reduction process according to the characteristics of thermal noise to frame picture noise in video, and obtain change into 3-D view after two-dimension human face image after carry out superposition calculation, the signal-to-noise ratio of human face photo is improved with this, and the texture for the part that can be blocked by the continuous completion face of human face photo of different angle in all two-dimension human face images, improve the quality of facial image.
Description
Technical field
The present invention relates to technical field of image processing, especially a kind of face image processing process of low-light (level), is set device
Standby and readable medium.
Background technology
With video monitoring system be widely used and face recognition technology is rapidly developed in recent years, use monitoring
The application that camera carries out recognition of face is more and more extensive.But to realize recognition of face, to the human face photo of camera shooting
It is more demanding, the conditions such as more stringent facial angle, clarity, illumination are needed, once some condition is poor, recognition of face
Accuracy rate will have a greatly reduced quality.
Low light environment refers to that ambient light is weaker, and the object under environment causes object dark due to insufficient light.For
For monitor camera, under low-light (level) environment, it is often difficult to take clearly picture.In order to solve under low light environment
Shooting problem, camera release starlight grade hardware, but still limited to the promotion for shooting picture quality.
But in face recognition application, since in monitor video, people is movement, by the side for increasing the time for exposure
Formula can cause moving object in picture fuzzy;By way of increasing sensitivity, picture can be made a large amount of noise occur, and
By algorithm noise reduction process, and final picture can be made to lose a large amount of details.Therefore, simple to be difficult by adjustment camera parameter
It is well adapted for the application of low-light (level) human face identification.
Noise on monitoring image includes mainly random noise of the thermal noise with height under photosensitive.Monitoring camera due to it is long when
Between run, photosensitive element will produce thermal noise, and this partial noise does not do corresponding processing substantially in existing application, but
Under low-light (level) environment, an important factor for thermal noise is also a quality for influencing picture.
All also only it is to improve the brightness of picture and right using the method for image enhancement substantially at present in field of face identification
It is processed than degree, and just for single photo.And by simple image enhancement, improving lightness and contrast
Meanwhile picture noise can be also amplified, the detail section of picture is influenced, the accuracy rate of recognition of face is reduced.
Invention content
The present invention is the defects of for the above-mentioned prior art, it is proposed that following technical solution.
A kind of face image processing process of low-light (level), this method include:
Extraction step handles each frame in video, extracts n two-dimension human face figures for belonging to same person
Picture;
N two-dimension human face images are generated n three-dimensional face images by generation step according to three-dimensional face model, and by n
Three-dimensional face images superposition generates a mean value three-dimensional face images;
The mean value three-dimensional face images are mapped to two-dimensional space and obtain a mean value two-dimension human face figure by mapping step
Picture;
Processing step obtains suitable brightness and comparison after doing image enhancement processing to the mean value two-dimension human face image
Facial image;
Wherein, n is integer, n ﹥ ﹦ 2.
Further, the operation of the thermal noise obtaining step is:Lens cap is covered after opening camera unit, is kept just
1 image of acquisition per minute after often operation 1 hour, 1 hour, obtains 60 images, to each pixel of 60 images altogether
Value, which is overlapped, averages, and the average value using each pixel value generates an image as the distribution of the thermal noise of camera unit
Image, and preserve the thermal noise distributed image.
Further, the extraction step includes:
Thermal noise removal step extracts the first image from video, and described image and the thermal noise distributed image are done
The thermal noise of difference removal described first image obtains the second image;
Histogram treatment step carries out histogram equalization processing to promote the whole of second image to second image
Body brightness and contrast obtains third image;
Detecting step detects in the third image whether have face, if so, then obtaining people using face detection model
Multiple characteristic point positions of face position and face simultaneously determine the piece identity of face according to the face location detected, it is assumed that when
Before the face that detects be personage A;
Two-dimension human face image obtaining step, it includes only face to be gone out from second image zooming-out according to the face location
The first face two dimensional image M1 and multiple characteristic point position P1 of face;
Above-mentioned thermal noise removal step, histogram treatment step, detecting step and two-dimension human face image is repeated to obtain
Step obtains the n two-dimension human face images { M1, M2 ..., Mn } of personage A and n corresponding with n two-dimension human face images
The multiple characteristic point positions of face { P1, P2 ..., Pn };
Wherein, P1, P2 ..., Pn indicate the vectors of multiple characteristic point positions.
Further, the generation step includes:
Three-dimensional face images obtain step, by the n of personage A two-dimension human face images { M1, M2 ..., Mn } according to n
The set { P1, P2 ..., Pn } of the corresponding multiple characteristic point positions of n face of two-dimension human face image is mapped to three-dimensional people
Face model, the pixel value of shield portions are set to 0 and obtain the n three-dimensional face images;
Step is weighted, the pixel value at the same position of the n three-dimensional face images is weighted average computation, and
The pixel for being 0 for pixel value does not participate in weighted average calculation, obtains an initial mean value three-dimensional face images;
Aligning step judges that the initial mean value three-dimensional face images whether there is pixel value as 0 region, if,
According to facial symmetry, the pixel value of symmetric position is assigned to the pixel that pixel value is 0 and obtains a mean value three-dimensional face figure
Picture.
Further, the number of the characteristic point is 68.
Further, it is described to the mean value two-dimension human face image do image enhancement processing be logarithmic transformation processing.
The invention also provides a kind of face image processing device of low-light (level), which includes:
Extraction unit extracts n two-dimension human faces for belonging to same person to handle each frame in video
Image;
Generation unit, n two-dimension human face images are generated n three-dimensional face images according to three-dimensional face model, and
N three-dimensional face images superpositions are generated into a mean value three-dimensional face images;
Map unit obtains a mean value two-dimension human face the mean value three-dimensional face images are mapped to two-dimensional space
Image;
Processing unit obtains suitable brightness and right after image enhancement processing to be done to the mean value two-dimension human face image
The facial image of ratio
Wherein, n is integer, n ﹥ ﹦ 2.
Further, described device further includes:
Thermal noise acquiring unit, to obtain the thermal noise distribution map of camera unit before extraction unit work
Picture.
Further, the operation that the thermal noise acquiring unit executes is:
Lens cap is covered after opening camera unit, keeps normal operation 1 image of acquisition per minute after 1 hour, 1 hour,
60 images are obtained altogether, each pixel value of 60 images is overlapped and is averaged, and use the flat of each pixel value
Average generation one opens thermal noise distributed image of the image as camera unit, and preserves the thermal noise distributed image.
Further, the extraction unit includes:
Thermal noise removes module, to extract the first image from video, by described image and the thermal noise distribution map
Thermal noise as doing difference removal described first image obtains the second image;
Histogram processing module, to carry out histogram equalization processing to second image to promote second image
Overall brightness and contrast, obtain third image;
Detection module, to use face detection model to detect in the third image whether have face, if so, then obtaining
It takes multiple characteristic point positions of face location and face and determines the piece identity of face according to the face location detected, it is false
If currently detected face is personage A;
Two-dimension human face image acquisition module includes only people to be gone out from second image zooming-out according to the face location
The multiple characteristic point position P1 of the first face two dimensional image M1 and face of face;
Above-mentioned thermal noise removal module, histogram processing module, detection module and two-dimension human face image is reused to obtain
Module execute corresponding operation obtain personage A n two-dimension human face images { M1, M2 ..., Mn } and with n two-dimension human face figures
As the corresponding multiple characteristic point positions of n face { P1, P2 ..., Pn };
Wherein, P1, P2 ..., Pn indicate the vectors of multiple characteristic point positions.
Further, the generation unit includes:
Three-dimensional face images obtain module, by the n of personage A two-dimension human face images { M1, M2 ..., Mn } according to n
The set { P1, P2 ..., Pn } of the corresponding multiple characteristic point positions of n face of two-dimension human face image is mapped to three-dimensional people
Face model, the pixel value of shield portions are set to 0 and obtain the n three-dimensional face images;
Pixel value at the same position of the n three-dimensional face images is weighted average computation by weighting block, and
The pixel for being 0 for pixel value does not participate in weighted average calculation, obtains an initial mean value three-dimensional face images;
Correction module judges that the initial mean value three-dimensional face images whether there is pixel value as 0 region, if,
According to facial symmetry, the pixel value of symmetric position is assigned to the pixel that pixel value is 0 and obtains a mean value three-dimensional face figure
Picture.
Further, the number of the characteristic point is 68.
Further, it is described to the mean value two-dimension human face image do image enhancement processing be logarithmic transformation processing.
The invention also provides a kind of face image processing equipment of low-light (level), the equipment includes processor, memory,
The processor is connected with the memory by bus, and machine readable code, the processor are stored in the memory
The machine readable code in memory is executed to execute above-mentioned any method.
The invention also provides a kind of computer readable storage medium, computer program generation is stored on the storage medium
Code can perform above-mentioned any method when the computer program code is computer-executed.
The present invention technique effect be:The present invention, which is proposed, carries out frame picture noise in video according to the characteristics of thermal noise
Noise reduction process, while using the video sequence in video, obtaining all two-dimension human face images of same person and changing into graphics
Carry out three-dimensional overlay calculating as after, improve the signal-to-noise ratio of human face photo with this, and can by all two-dimension human face images not
The continuous completion face of human face photo with angle is blocked the texture of part, improves the quality of facial image, face is promoted with this
The accuracy rate of identification.
Description of the drawings
Fig. 1 is a kind of flow chart of the face image processing process of low-light (level) according to an embodiment of the invention.
Fig. 2 is the flow chart of extraction step according to an embodiment of the invention.
Fig. 3 is the flow chart of generation system according to an embodiment of the invention.
Fig. 4 is a kind of structural schematic diagram of the face image processing device of low-light (level) according to an embodiment of the invention.
Fig. 5 is the structural schematic diagram of extraction unit according to an embodiment of the invention.
Fig. 6 is the structural schematic diagram of generation unit according to an embodiment of the invention.
Fig. 7 is the structural schematic diagram of the face image processing equipment of low-light (level) according to an embodiment of the invention.
Specific implementation mode
1-7 is specifically described below in conjunction with the accompanying drawings.
Fig. 1 shows that a kind of face image processing process of low-light (level) of the present invention, this method include:
Extraction step S1 handles each frame in video, extracts n two-dimension human face figures for belonging to same person
Picture;
N two-dimension human face images are generated n three-dimensional face images by generation step S2 according to three-dimensional face model, and by n
It opens three-dimensional face images superposition and generates a mean value three-dimensional face images;
The mean value three-dimensional face images are mapped to two-dimensional space and obtain a mean value two-dimension human face figure by mapping step S3
Picture;
Processing step S4 obtains suitable brightness and comparison after doing image enhancement processing to the mean value two-dimension human face image
Facial image;
Wherein, n is integer, n ﹥ ﹦ 2.
Video comes from monitor video in the embodiment of the present invention, and the camera unit in the present invention can be monitor, monitoring
Camera, camera etc., since monitoring camera is due to long-play, photosensitive element will produce heat and make an uproar, and this part is made an uproar
Sound does not do corresponding processing substantially in existing application, but under low-light (level) environment, and heat is made an uproar and one influences picture
Quality an important factor for.The present invention propose removal image in thermal noise method, this be the application important inventive point it
One, as shown in Figure 1, the present processes further include:Thermal noise obtaining step S0, to obtain the thermal noise distribution of camera unit
Image;Wherein, thermal noise obtaining step S0 is before the extraction step S1.
The operation of the thermal noise obtaining step S0 is:Lens cap is covered after opening camera unit, keeps normal operation 1 small
When, 1 image of acquisition per minute after 1 hour obtains 60 images, is overlapped to each pixel value of 60 images altogether
It averages, the average value using each pixel value generates an image as the thermal noise distributed image of camera unit, and protects
Deposit the thermal noise distributed image.Thermal noise obtaining step S0 only needs to use preceding execution in camera, with camera shooting usage time
Lengthening, camera electronic device can aging, be periodically executed the step, such as once per week, thermal noise will be obtained
Distributed image uses after preserving for follow-up.
How facial image to be extracted from video sequence, is one of the key point of the present invention, as shown in Fig. 2, institute
Stating extraction step S1 includes:
Thermal noise removal step S11, extracts the first image from video, by described image and the thermal noise distributed image
The thermal noise for doing difference removal described first image obtains the second image;
Histogram treatment step S12 carries out histogram equalization processing to promote second image to second image
Overall brightness and contrast, obtain third image;
Detecting step S13 detects in the third image whether have face, if so, then obtaining using face detection model
It takes multiple characteristic point positions of face location and face and determines the piece identity of face according to the face location detected, it is false
If currently detected face is personage A;
Two-dimension human face image obtaining step S14, it includes only people to be gone out from second image zooming-out according to the face location
The multiple characteristic point position P1 of the first face two dimensional image M1 and face of face;
Repeat above-mentioned thermal noise removal step S11, histogram treatment step S12, detecting step S13 and two-dimension human face
Image acquisition step S14 obtain personage A n two-dimension human face images { M1, M2 ..., Mn } and with n two-dimension human face images point
Not corresponding multiple characteristic point positions of n face { P1, P2 ..., Pn };When repeating, if personage's location A does not have nearby
Detect that face, i.e. personage A leave monitoring area, then the extraction for completing two-dimension human face image enters 3-D view generation.
Wherein, P1, P2 ..., Pn indicate the vectors of multiple characteristic point positions.
In one embodiment, the number of human face characteristic point is selected as 68.Characteristic point is excessive, and calculating speed is slower, characteristic point
Less, the three-dimensional face images being subsequently generated are inaccurate.For example, 68 characteristic points of selection, each vector PXAll be (x1, y1),
(x2, y2) ..., (x68, y68), 1=<X<=n, i.e., each vector PXInclude the coordinate of 68 characteristic points.
How the facial image relatively sharp in the case where generating low-light (level), the embodiment of the present invention use method be will be two-dimentional
Human face image sequence is converted to three-dimensional face images sequence, and generates accurate three-dimensional face images after carrying out overlap-add procedure, this
It is one of the important inventive point of the present invention, as shown in figure 3, the generation step S2 includes:
Three-dimensional face images obtain step S21, by the n of personage A two-dimension human face images { M1, M2 ..., Mn } according to
The set { P1, P2 ..., Pn } of n corresponding multiple characteristic point positions of n face of two-dimension human face image is mapped to three-dimensional
Faceform, the pixel value of shield portions are set to 0 and obtain the n three-dimensional face images;
Step S22 is weighted, the pixel value at the same position of the n three-dimensional face images is weighted average meter
It calculates, and the pixel for being 0 for pixel value does not participate in weighted average calculation, obtains an initial mean value three-dimensional face images;
Aligning step S23 judges that the initial mean value three-dimensional face images whether there is pixel value for 0 region, such as
The pixel value of symmetric position is assigned to the pixel that pixel value is 0 and obtains a mean value three-dimensional face by fruit according to facial symmetry
Image.
Image enhancement processing (such as logarithmic transformation is carried out again to mapping generation two dimensional image by 3-D view in the present invention
Reason), all relatively good face two dimensional image of brightness and illuminance is obtained, to improve the accuracy rate of low-light (level) human face identification.
Fig. 4 shows that a kind of face image processing device of low-light (level) of the present invention, the device include:
Extraction unit 41 extracts n two-dimentional people for belonging to same person to handle each frame in video
Face image;
Generation unit 42, n two-dimension human face images are generated n three-dimensional face images according to three-dimensional face model,
And n three-dimensional face images superpositions are generated into a mean value three-dimensional face images;
Map unit 43 obtains a mean value two dimension people the mean value three-dimensional face images are mapped to two-dimensional space
Face image;
Processing unit 44, to the mean value two-dimension human face image is done obtain after image enhancement processing suitable brightness and
The facial image of comparison
Wherein, n is integer, n ﹥ ﹦ 2.
Video comes from monitor video in the embodiment of the present invention, and the camera unit in the present invention can be monitor, monitoring
Camera, camera etc., since monitoring camera is due to long-play, photosensitive element will produce heat and make an uproar, and this part is made an uproar
Sound does not do corresponding processing substantially in existing application, but under low-light (level) environment, and heat is made an uproar and one influences picture
Quality an important factor for.The present invention propose removal image in thermal noise method, this be the application important inventive point it
One, as shown in figure 4, the described device of the application further includes:Thermal noise acquiring unit 40, to work in the extraction unit 41
The thermal noise distributed image of camera unit is obtained before.
The operation that the thermal noise acquiring unit 40 executes is:Lens cap is covered after opening camera unit, keeps normal fortune
Row 1 image of acquisition per minute after 1 hour, 1 hour obtains 60 images altogether, to each pixel values of 60 images into
Row superposition is averaged, and the average value using each pixel value generates an image as the thermal noise distribution map of camera unit
Picture, and preserve the thermal noise distributed image.Thermal noise obtaining step S0 only needs to use preceding execution in camera, as camera shooting makes
With the lengthening of time, the meeting aging of camera electronic device is periodically executed the step, such as once per week, will obtain
Thermal noise distributed image uses after preserving for follow-up.
How facial image to be extracted from video sequence, is one of the key point of the present invention, as shown in figure 5, institute
Stating extraction unit 41 includes:
Thermal noise removes module 411, and to extract the first image from video, described image and the thermal noise are distributed
The thermal noise that image does difference removal described first image obtains the second image;
Histogram processing module 412, to carry out histogram equalization processing to second image to promote described second
The overall brightness and contrast of image, obtain third image;
Detection module 413, to use face detection model to detect in the third image whether have face, if so,
It then obtains multiple characteristic point positions of face location and face and determines personage's body of face according to the face location detected
Part, it is assumed that currently detected face is personage A;
Two-dimension human face image acquisition module 414 only wraps to be gone out from second image zooming-out according to the face location
The first face two dimensional image M1 containing the face and multiple characteristic point position P1 of face;
Reuse above-mentioned thermal noise removal module 411, histogram processing module 412, detection module 413 and two-dimension human face
Image collection module 414 execute corresponding operation obtain personage A n two-dimension human face images { M1, M2 ..., Mn } and with n
The corresponding multiple characteristic point positions of n face of two-dimension human face image { P1, P2 ..., Pn };When repeating, if people
Object location A does not nearby detect that face, i.e. personage A leave monitoring area, then the extraction for completing two-dimension human face image enters three
Image is tieed up to generate.
Wherein, P1, P2 ..., Pn indicate the vectors of multiple characteristic point positions.
In one embodiment, the number of human face characteristic point is selected as 68.Characteristic point is excessive, and calculating speed is slower, characteristic point
Less, the three-dimensional face images being subsequently generated are inaccurate.For example, 68 characteristic points of selection, each vector PXAll be (x1, y1),
(x2, y2) ..., (x68, y68), 1=<X<=n, i.e., each vector PXInclude the coordinate of 68 characteristic points.
How the facial image relatively sharp in the case where generating low-light (level), the embodiment of the present invention use method be will be two-dimentional
Human face image sequence is converted to three-dimensional face images sequence, and generates accurate three-dimensional face images after carrying out overlap-add procedure, this
It is one of the important inventive point of the present invention, as shown in fig. 6, the generation unit 42 includes:
Three-dimensional face images obtain module 421, by the n of personage A two-dimension human face images { M1, M2 ..., Mn } according to
The set { P1, P2 ..., Pn } of n corresponding multiple characteristic point positions of n face of two-dimension human face image is mapped to three-dimensional
Faceform, the pixel value of shield portions are set to 0 and obtain the n three-dimensional face images;
Pixel value at the same position of the n three-dimensional face images is weighted average meter by weighting block 422
It calculates, and the pixel for being 0 for pixel value does not participate in weighted average calculation, obtains an initial mean value three-dimensional face images;
Correction module 423 judges that the initial mean value three-dimensional face images whether there is pixel value for 0 region, such as
The pixel value of symmetric position is assigned to the pixel that pixel value is 0 and obtains a mean value three-dimensional face by fruit according to facial symmetry
Image.
Image enhancement processing (such as logarithmic transformation is carried out again to mapping generation two dimensional image by 3-D view in the present invention
Reason), all relatively good face two dimensional image of brightness and illuminance is obtained, to improve the accuracy rate of low-light (level) human face identification.
The present invention is based on video sequence, 3D is carried out to the facial image in entire sequence and is superimposed noise reduction, after the completion of noise reduction again
2D face pictures are changed into, then carry out image enhancement.By taking a people takes 4 seconds by camera head monitor region as an example, camera 1 second
25 frame images are acquired, i.e., entire video sequence can at least collect 100 faces of same person, after superposition, noise
Value decline close to 1/100, and the value of valid data is then held essentially constant, therefore is effectively improved the noise of picture
Than not only having eliminated noise but also utmostly having remained the details of face, and improved the quality of final human face photo, ensure that follow-up
The accuracy rate of recognition of face.
Low light environment is widely present in reality scene, such as dim interior or night etc..Current face is known
Preferable illumination condition has not been required, strongly limits the use of recognition of face under low-light (level) environment.It, can using the present invention
To widen the application scenarios of recognition of face, has under video monitoring scene especially under low-light (level) environment and greatly apply valence
Value.
The equipment that Fig. 7 shows a kind of operation electronic signature of the present invention, including:Memory a and processor b, it is described to deposit
Computer program is stored in reservoir a, when the computer program is executed by the processor b, the processor b executes memory
Machine readable code in a is to execute method.
The invention also provides a kind of computer readable storage medium, computer program generation is stored on the storage medium
Code, one of above-mentioned method is can perform when the computer program code is computer-executed.
For convenience of description, it is divided into various units when description apparatus above with function to describe respectively.Certainly, implementing this
The function of each unit is realized can in the same or multiple software and or hardware when application, the present invention in so-called client,
Client refers to identical content, and the server-side, server, server end in the present invention refer to identical content.
As seen through the above description of the embodiments, those skilled in the art can be understood that the application can
It is realized by the mode of software plus required general hardware platform.Based on this understanding, the technical solution essence of the application
On in other words the part that contributes to existing technology can be expressed in the form of software products, the computer software product
It can be stored in a storage medium, such as ROM/RAM, magnetic disc, CD, including some instructions are used so that a computer equipment
(can be personal computer, server either network equipment etc.) executes the certain of each embodiment of the application or embodiment
Method described in part.
It should be noted last that:Above example only illustrates and not to limitation technical scheme of the present invention, although reference
Above-described embodiment describes the invention in detail, it will be understood by those of ordinary skill in the art that:It still can be to this hair
It is bright to be modified or replaced equivalently, it without departing from the spirit or scope of the invention, or any substitutions, should all
Cover in the scope of the claims of the present invention.
Claims (16)
1. a kind of face image processing process of low-light (level), which is characterized in that this method includes:
Extraction step handles each frame in video, extracts n two-dimension human face images for belonging to same person;
N two-dimension human face images are generated n three-dimensional face images according to three-dimensional face model, and n Zhang San are tieed up by generation step
Facial image superposition generates a mean value three-dimensional face images;
The mean value three-dimensional face images are mapped to two-dimensional space and obtain a mean value two-dimension human face image by mapping step;
Processing step does the mean value two-dimension human face image face that suitable brightness and comparison are obtained after image enhancement processing
Image;
Wherein, n is integer, and n ﹥ ﹦ 2.
2. the method according to claim 1, which is characterized in that the method further includes:
Thermal noise obtaining step, to obtain the thermal noise distributed image of camera unit;
Wherein, thermal noise obtaining step is before the extraction step.
3. method according to claim 2, which is characterized in that the operation of the thermal noise obtaining step is:
Lens cap is covered after opening camera unit, holding normal operation 1 image of acquisition per minute after 1 hour, 1 hour obtains altogether
60 images are taken, each pixel value of 60 images is overlapped and is averaged, the average value of each pixel value is used
Thermal noise distributed image of the image as camera unit is generated, and preserves the thermal noise distributed image.
4. method according to claim 3, which is characterized in that the extraction step includes:
Thermal noise removal step extracts the first image from video, and described image and the thermal noise distributed image are done difference
The thermal noise of removal described first image obtains the second image;
It is bright to promote the entirety of second image to carry out histogram equalization processing to second image for histogram treatment step
Degree and contrast, obtain third image;
Detecting step detects in the third image whether have face, if so, then obtaining face position using face detection model
It sets multiple characteristic point positions with face and determines the piece identity of face according to the face location detected, it is assumed that current inspection
The face measured is personage A;
Two-dimension human face image obtaining step goes out from second image zooming-out according to the face location and only includes the first of face
The face two dimensional image M1 and multiple characteristic point position P1 of face;
Repeat above-mentioned thermal noise removal step, histogram treatment step, detecting step and two-dimension human face image obtaining step
Obtain the n two-dimension human face images { M1, M2 ..., Mn } of personage A and n face corresponding with n two-dimension human face images
Multiple characteristic point positions { P1, P2 ..., Pn };
Wherein, P1, P2 ..., Pn indicate the vectors of multiple characteristic point positions.
5. method according to claim 4, which is characterized in that the generation step includes:
Three-dimensional face images obtain step, by the n of personage A two-dimension human face images { M1, M2 ..., Mn } according to and n two dimensions
The set { P1, P2 ..., Pn } of the corresponding multiple characteristic point positions of n face of facial image is mapped to three-dimensional face mould
Type, the pixel value of shield portions are set to 0 and obtain the n three-dimensional face images;
Step is weighted, the pixel value at the same position of the n three-dimensional face images is weighted average computation, and for
The pixel that pixel value is 0 does not participate in weighted average calculation, obtains an initial mean value three-dimensional face images;
Aligning step judges that the initial mean value three-dimensional face images whether there is pixel value as 0 region, if, according to
The pixel value of symmetric position is assigned to the pixel that pixel value is 0 and obtains a mean value three-dimensional face images by facial symmetry.
6. method according to claim 5, which is characterized in that the number of the characteristic point is 68.
7. the method according to claim 1, which is characterized in that described to do image enhancement processing to the mean value two-dimension human face image
It is logarithmic transformation processing.
8. a kind of face image processing device of low-light (level), which is characterized in that the device includes:
Extraction unit extracts n two-dimension human face figures for belonging to same person to handle each frame in video
Picture;
Generation unit, n two-dimension human face images are generated n three-dimensional face images according to three-dimensional face model, and by n
Three-dimensional face images superposition generates a mean value three-dimensional face images;
Map unit obtains a mean value two-dimension human face figure the mean value three-dimensional face images are mapped to two-dimensional space
Picture;
Processing unit obtains suitable brightness and comparison after image enhancement processing to be done to the mean value two-dimension human face image
Facial image
Wherein, n is integer, and n ﹥ ﹦ 2.
9. device according to claim 8, which is characterized in that described device further includes:
Thermal noise acquiring unit, to obtain the thermal noise distributed image of camera unit before extraction unit work.
10. device according to claim 9, which is characterized in that the operation that the thermal noise acquiring unit executes is:
Lens cap is covered after opening camera unit, holding normal operation 1 image of acquisition per minute after 1 hour, 1 hour obtains altogether
60 images are taken, each pixel value of 60 images is overlapped and is averaged, the average value of each pixel value is used
Thermal noise distributed image of the image as camera unit is generated, and preserves the thermal noise distributed image.
11. device according to claim 10, which is characterized in that the extraction unit includes:
Thermal noise removal module does described image and the thermal noise distributed image to extract the first image from video
The thermal noise of difference removal described first image obtains the second image;
Histogram processing module, to carry out histogram equalization processing to second image to promote the whole of second image
Body brightness and contrast obtains third image;
Detection module, to use face detection model to detect in the third image whether have face, if so, then obtaining people
Multiple characteristic point positions of face position and face simultaneously determine the piece identity of face according to the face location detected, it is assumed that when
Before the face that detects be personage A;
Two-dimension human face image acquisition module includes only face to be gone out from second image zooming-out according to the face location
The first face two dimensional image M1 and multiple characteristic point position P1 of face;
Reuse above-mentioned thermal noise removal module, histogram processing module, detection module and two-dimension human face image acquisition module
Execute corresponding operation obtain personage A n two-dimension human face images { M1, M2 ..., Mn } and with n two-dimension human face images point
Not corresponding multiple characteristic point positions of n face { P1, P2 ..., Pn };
Wherein, P1, P2 ..., Pn indicate the vectors of multiple characteristic point positions.
12. device according to claim 11, which is characterized in that the generation unit includes:
Three-dimensional face images obtain module, by the n of personage A two-dimension human face images { M1, M2 ..., Mn } according to and n two dimensions
The set { P1, P2 ..., Pn } of the corresponding multiple characteristic point positions of n face of facial image is mapped to three-dimensional face mould
Type, the pixel value of shield portions are set to 0 and obtain the n three-dimensional face images;
Pixel value at the same position of the n three-dimensional face images is weighted average computation by weighting block, and for
The pixel that pixel value is 0 does not participate in weighted average calculation, obtains an initial mean value three-dimensional face images;
Correction module judges that the initial mean value three-dimensional face images whether there is pixel value as 0 region, if, according to
The pixel value of symmetric position is assigned to the pixel that pixel value is 0 and obtains a mean value three-dimensional face images by facial symmetry.
13. device according to claim 12, which is characterized in that the number of the characteristic point is 68.
14. device according to claim 8, which is characterized in that described to be done at image enhancement to the mean value two-dimension human face image
Reason is logarithmic transformation processing.
15. a kind of equipment of operation electronic signature, which is characterized in that the equipment includes processor, memory, the processor
It is connected by bus with the memory, machine readable code is stored in the memory, the processor executes memory
In machine readable code with perform claim require 1-8 any one of them method.
16. a kind of computer readable storage medium, which is characterized in that it is stored with computer program code on the storage medium,
Any method of 1-8 is required with perform claim when the computer program code is computer-executed.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810516307.5A CN108629333A (en) | 2018-05-25 | 2018-05-25 | A kind of face image processing process of low-light (level), device, equipment and readable medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810516307.5A CN108629333A (en) | 2018-05-25 | 2018-05-25 | A kind of face image processing process of low-light (level), device, equipment and readable medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108629333A true CN108629333A (en) | 2018-10-09 |
Family
ID=63690439
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810516307.5A Pending CN108629333A (en) | 2018-05-25 | 2018-05-25 | A kind of face image processing process of low-light (level), device, equipment and readable medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108629333A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111523776A (en) * | 2020-04-08 | 2020-08-11 | 辽宁百思特达半导体科技有限公司 | Smart city system and method based on illegal target risk assessment |
CN112509144A (en) * | 2020-12-09 | 2021-03-16 | 深圳云天励飞技术股份有限公司 | Face image processing method and device, electronic equipment and storage medium |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100613912B1 (en) * | 2004-01-26 | 2006-08-17 | 엘지전자 주식회사 | Image processing device and method thereof |
CN101360246A (en) * | 2008-09-09 | 2009-02-04 | 西南交通大学 | Video error masking method combined with 3D human face model |
CN104077804A (en) * | 2014-06-09 | 2014-10-01 | 广州嘉崎智能科技有限公司 | Method for constructing three-dimensional human face model based on multi-frame video image |
CN105205480A (en) * | 2015-10-31 | 2015-12-30 | 潍坊学院 | Complex scene human eye locating method and system |
US20160148041A1 (en) * | 2014-11-21 | 2016-05-26 | Korea Institute Of Science And Technology | Method for face recognition through facial expression normalization, recording medium and device for performing the method |
CN105844276A (en) * | 2015-01-15 | 2016-08-10 | 北京三星通信技术研究有限公司 | Face posture correction method and face posture correction device |
CN105913404A (en) * | 2016-07-01 | 2016-08-31 | 湖南源信光电科技有限公司 | Low-illumination imaging method based on frame accumulation |
US20160335480A1 (en) * | 2015-05-15 | 2016-11-17 | Toshiba Tec Kabushiki Kaisha | Video Based Facial Recognition for Customer Verification at Touchless Checkout |
CN106446872A (en) * | 2016-11-07 | 2017-02-22 | 湖南源信光电科技有限公司 | Detection and recognition method of human face in video under low-light conditions |
CN106778563A (en) * | 2016-12-02 | 2017-05-31 | 江苏大学 | A kind of quick any attitude facial expression recognizing method based on the coherent feature in space |
CN107633165A (en) * | 2017-10-26 | 2018-01-26 | 深圳奥比中光科技有限公司 | 3D face identity authentications and device |
-
2018
- 2018-05-25 CN CN201810516307.5A patent/CN108629333A/en active Pending
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100613912B1 (en) * | 2004-01-26 | 2006-08-17 | 엘지전자 주식회사 | Image processing device and method thereof |
CN101360246A (en) * | 2008-09-09 | 2009-02-04 | 西南交通大学 | Video error masking method combined with 3D human face model |
CN104077804A (en) * | 2014-06-09 | 2014-10-01 | 广州嘉崎智能科技有限公司 | Method for constructing three-dimensional human face model based on multi-frame video image |
US20160148041A1 (en) * | 2014-11-21 | 2016-05-26 | Korea Institute Of Science And Technology | Method for face recognition through facial expression normalization, recording medium and device for performing the method |
CN105844276A (en) * | 2015-01-15 | 2016-08-10 | 北京三星通信技术研究有限公司 | Face posture correction method and face posture correction device |
US20160335480A1 (en) * | 2015-05-15 | 2016-11-17 | Toshiba Tec Kabushiki Kaisha | Video Based Facial Recognition for Customer Verification at Touchless Checkout |
CN105205480A (en) * | 2015-10-31 | 2015-12-30 | 潍坊学院 | Complex scene human eye locating method and system |
CN105913404A (en) * | 2016-07-01 | 2016-08-31 | 湖南源信光电科技有限公司 | Low-illumination imaging method based on frame accumulation |
CN106446872A (en) * | 2016-11-07 | 2017-02-22 | 湖南源信光电科技有限公司 | Detection and recognition method of human face in video under low-light conditions |
CN106778563A (en) * | 2016-12-02 | 2017-05-31 | 江苏大学 | A kind of quick any attitude facial expression recognizing method based on the coherent feature in space |
CN107633165A (en) * | 2017-10-26 | 2018-01-26 | 深圳奥比中光科技有限公司 | 3D face identity authentications and device |
Non-Patent Citations (1)
Title |
---|
BO-CHING HUANG等: "Image Pipeline Algorithms For Standard Mobile Imaging Architecure Sensors", 《CVGIP 2005》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111523776A (en) * | 2020-04-08 | 2020-08-11 | 辽宁百思特达半导体科技有限公司 | Smart city system and method based on illegal target risk assessment |
CN112509144A (en) * | 2020-12-09 | 2021-03-16 | 深圳云天励飞技术股份有限公司 | Face image processing method and device, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
KR102319177B1 (en) | Method and apparatus, equipment, and storage medium for determining object pose in an image | |
WO2018176938A1 (en) | Method and device for extracting center of infrared light spot, and electronic device | |
CN106372629B (en) | Living body detection method and device | |
CN111652086B (en) | Face living body detection method and device, electronic equipment and storage medium | |
CN107025660B (en) | Method and device for determining image parallax of binocular dynamic vision sensor | |
CN112487921B (en) | Face image preprocessing method and system for living body detection | |
CN107798702B (en) | Real-time image superposition method and device for augmented reality | |
WO2016107638A1 (en) | An image face processing method and apparatus | |
CN111368601B (en) | Living body detection method and apparatus, electronic device, and computer-readable storage medium | |
CN109086724A (en) | A kind of method for detecting human face and storage medium of acceleration | |
CN111340749B (en) | Image quality detection method, device, equipment and storage medium | |
CN106412441B (en) | A kind of video stabilization control method and terminal | |
KR20110014067A (en) | Method and system for transformation of stereo content | |
CN110505398B (en) | Image processing method and device, electronic equipment and storage medium | |
CN111680574B (en) | Face detection method and device, electronic equipment and storage medium | |
CN107820019B (en) | Blurred image acquisition method, blurred image acquisition device and blurred image acquisition equipment | |
CN111008935A (en) | Face image enhancement method, device, system and storage medium | |
CN112580434A (en) | Face false detection optimization method and system based on depth camera and face detection equipment | |
CN108629333A (en) | A kind of face image processing process of low-light (level), device, equipment and readable medium | |
CN114419102A (en) | Multi-target tracking detection method based on frame difference time sequence motion information | |
JP2009211122A (en) | Image processor and object estimation program | |
CN113362221A (en) | Face recognition system and face recognition method for entrance guard | |
CN109344758B (en) | Face recognition method based on improved local binary pattern | |
CN111382646A (en) | Living body identification method, storage medium and terminal equipment | |
JP2009258770A (en) | Image processing method, image processor, image processing program, and imaging device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181009 |
|
RJ01 | Rejection of invention patent application after publication |