WO2017114133A1 - 图像识别的方法和装置 - Google Patents

图像识别的方法和装置 Download PDF

Info

Publication number
WO2017114133A1
WO2017114133A1 PCT/CN2016/109137 CN2016109137W WO2017114133A1 WO 2017114133 A1 WO2017114133 A1 WO 2017114133A1 CN 2016109137 W CN2016109137 W CN 2016109137W WO 2017114133 A1 WO2017114133 A1 WO 2017114133A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
signal
analysis
face image
amplitude value
Prior art date
Application number
PCT/CN2016/109137
Other languages
English (en)
French (fr)
Inventor
刘若鹏
阮志锋
Original Assignee
深圳光启合众科技有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 深圳光启合众科技有限公司 filed Critical 深圳光启合众科技有限公司
Publication of WO2017114133A1 publication Critical patent/WO2017114133A1/zh

Links

Images

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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction

Definitions

  • the present invention relates to the field of face recognition, and in particular to a method and apparatus for image recognition.
  • Principal Component Analysis is a relatively accurate face recognition algorithm. Through certain training, the principal components of face images are extracted and unsupervised learning methods are implemented to realize face recognition and face recognition. .
  • PCA Principal Component Analysis
  • Embodiments of the present invention provide a method and apparatus for image recognition to solve at least the technical problem that the recognition rate of face recognition is relatively low.
  • a method for image recognition includes: obtaining an analytical signal from an analytical image of a face image, wherein an amplitude value of the analytical signal carries a representation in the facial image Illuminating information; performing normalization processing on the amplitude value in the parsing signal; and performing image recognition on the face image reconstructed by the amplitude value obtained by the normalization processing.
  • acquiring the parsing signal from the parsed image of the face image comprises: acquiring the 2D of the face image Parsing an image, and acquiring two Hilbert transform analysis signals from the two-dimensional analysis image; or acquiring a two-dimensional analysis image of the face image, and obtaining quaternion resolution from the two-dimensional analysis image signal.
  • the two Hilbert transform parsing signals are identical to each other.
  • f is a 2-dimensional analytical image of the face image
  • ⁇ 1 and ⁇ 3 are mapping values of the Hilbert transform of the 2-dimensional analytical image
  • a 1 and a 3 are the two Hilberts Special transform analysis amplitude value of the signal, with Is to analyze the phase value of the signal, i' is an imaginary unit
  • H ⁇ f ⁇ is the Hilbert transform of the two-dimensional analytical image f
  • H x ⁇ f ⁇ is the x-direction of the two-dimensional analytical image f
  • H y ⁇ f ⁇ is a Hilbert transform in the y direction of the 2-dimensional analytical image f
  • normalizing the amplitude value in the parsed signal includes: 1 and the a 3 are normalized.
  • the quaternion resolution signal is
  • f is a two-dimensional analytical image of the face image
  • ⁇ q is a mapping value of the Hilbert transform of the two-dimensional analytical image
  • a q is an amplitude value of the quaternion analysis signal
  • Is a phase value
  • i, j, and k are imaginary units of quaternions
  • H ⁇ f ⁇ is a Hilbert transform of the 2-dimensional analytical image f
  • H x ⁇ f ⁇ is the 2-dimensional analytical image f
  • H y ⁇ f ⁇ is a Hilbert transform in the y direction of the two-dimensional analytical image f
  • normalizing the amplitude values in the analysis signal includes: The a q is normalized.
  • performing image recognition on the face image reconstructed by using the amplitude value obtained after the normalization process includes: Inputting the amplitude value obtained by the normalization process into the analysis signal; reconstructing a face image by using an analysis signal inputting the amplitude value obtained by the normalization process; and performing a human face image on the reconstructed face image Face recognition.
  • the facial image is reconstructed by using the parsing signal input with the amplitude value obtained by the normalization process, and the obtained facial image is as follows:
  • f new (x, y) is the reconstructed face image
  • a 1nor and a 3nor are the normalized amplitude values of the two Hilbert transform analysis signals, with Is to analyze the phase value of the signal.
  • the facial image is reconstructed by using the parsing signal input with the amplitude value obtained by the normalization process, and the obtained facial image is as follows:
  • f new (x, y) is the reconstructed face image
  • a qnor is the normalized amplitude value of the quaternion analysis signal
  • an apparatus for image recognition comprising: an obtaining unit, configured to acquire an analytical signal from a parsed image of a face image, wherein an amplitude value of the parsing signal carries the The information indicating the illumination in the face image; the processing unit configured to normalize the amplitude value in the parsed signal; and the recognition unit, configured to reconstruct the amplitude value obtained by using the normalization process
  • the face image is image recognized.
  • the acquiring unit includes: a first acquiring module, configured to acquire a 2-dimensional analytical image of the face image, and obtain two Hilbert transform analysis signals from the 2-dimensional analytical image; or a second And an acquiring module, configured to acquire a 2-dimensional analysis image of the face image, and obtain a quaternion analysis signal from the 2-dimensional analysis image.
  • the two Hilbert transform parsing signals are identical to each other.
  • f is a 2-dimensional analytical image of the face image
  • ⁇ 1 and ⁇ 3 are mapping values of the Hilbert transform of the 2-dimensional analytical image
  • a 1 and a 3 are the two Hilberts Special transform analysis amplitude value of the signal, with Is to analyze the phase value of the signal, i' is an imaginary unit
  • H ⁇ f ⁇ is the Hilbert transform of the two-dimensional analytical image f
  • H x ⁇ f ⁇ is the x-direction of the two-dimensional analytical image f
  • H y ⁇ f ⁇ is a Hilbert transform in the y direction of the 2-dimensional analytical image f
  • the processing unit is further configured to normalize the a 1 and the a 3 respectively deal with.
  • the quaternion resolution signal is
  • f is a two-dimensional analytical image of the face image
  • ⁇ q is a mapping value of the Hilbert transform of the two-dimensional analytical image
  • a q is an amplitude value of the quaternion analysis signal
  • Is a phase value
  • i, j, and k are imaginary units of quaternions
  • H ⁇ f ⁇ is a Hilbert transform of the 2-dimensional analytical image f
  • H x ⁇ f ⁇ is the 2-dimensional analytical image f
  • H y ⁇ f ⁇ is a Hilbert transform in the y direction of the 2D analytical image f
  • the processing unit is further configured to normalize the a q .
  • the identifying unit includes: an input module, configured to input the amplitude value obtained by the normalization process into the parsing signal; and a reconstruction module, configured to perform the normalization process by using the input The obtained analytical signal of the amplitude value reconstructs the face image; and the recognition module is configured to perform face recognition on the reconstructed face image.
  • the reconstructing module reconstructs a face image by using an parsing signal input with the amplitude value obtained by the normalization process, and the obtained person is obtained.
  • the face image is as follows:
  • f new (x, y) is the reconstructed face image
  • a 1nor and a 3nor are the normalized amplitude values of the two Hilbert transform analysis signals, with Is to analyze the phase value of the signal.
  • the reconstructing module reconstructs the facial image by using the parsing signal input with the amplitude value obtained by the normalization process, and the obtained facial image is as follows:
  • f new (x, y) is the reconstructed face image
  • a qnor is the normalized amplitude value of the quaternion analysis signal
  • the parsing signal is obtained from the parsed image of the face image, wherein the amplitude value of the parsing signal carries the information representing the illumination in the human face image; and the amplitude value in the parsed signal is normalized; Performing image recognition on the face image reconstructed by the amplitude value obtained by the normalization process, before analyzing the collected face image, analyzing the amplitude value of the information carrying the characteristic light in the new model, and the amplitude The value continues to be normalized, reducing the amplitude difference between the multiple amplitude values, thereby eliminating the luminance difference between the higher brightness and the lower brightness, improving the robustness of the brightness conversion, and reducing the illumination pair acquisition. The effect of the face image.
  • the face image is reconstructed by using the normalized amplitude value, and the reconstructed face image eliminates the influence of the illumination change on the captured face image on the face recognition, so that when the reconstructed face image is used for image recognition
  • the invention can improve the accuracy of face recognition, thereby solving the technical problem that the recognition rate of face recognition in the prior art is relatively low, and the effect of improving the recognition rate of face recognition is achieved.
  • FIG. 1 is a flow chart of a method of image recognition according to an embodiment of the present invention.
  • FIG. 2 is a schematic diagram of an apparatus for image recognition according to an embodiment of the present invention.
  • FIG. 3 is a schematic diagram of an acquisition unit including a first acquisition module in an image recognition apparatus according to an embodiment of the invention
  • FIG. 4 is a schematic diagram of an acquisition unit including a second acquisition module in an image recognition apparatus according to another embodiment of the present invention.
  • FIG. 5 is a schematic diagram of an identification unit including a plurality of modules in an image recognition apparatus according to an embodiment of the present invention.
  • an embodiment of a method of image recognition is provided, it being noted that the steps illustrated in the flowchart of the figures may be performed in a computer system such as a set of computer executable instructions, and, although The logical order is shown in the flowcharts, but in some cases the steps shown or described may be performed in a different order than the ones described herein.
  • the method for image recognition is to preprocess the obtained face image after obtaining the face image, and normalize the amplitude value of the face information in the face image to eliminate the influence of brightness on the face image. . Then, the reconstructed face image is reconstructed by using the processed amplitude value, and the reconstructed face image is recognized, which improves the robustness of the brightness change in the face recognition process, thereby solving the problem of changing the illumination in the face recognition process.
  • the technical problem that the recognition rate of face recognition is relatively low has reached the technical effect of improving the face recognition rate.
  • the processing of the face image in the image recognition method is applied before the recognition, it is a pre-processing of the face image, and therefore, it is not limited to which face recognition algorithm is adopted, and the PCA can be applied as a face.
  • face recognition of the recognition algorithm face recognition using other face recognition algorithms can also be applied.
  • FIG. 1 is a flow chart of a method for image recognition according to an embodiment of the present invention. As shown in FIG. 1, the method includes the following steps:
  • Step S102 Acquire an analytical signal from the analytical image of the facial image, wherein the amplitude value of the analytical signal carries information representing the illumination in the human face image.
  • step S104 the amplitude value in the analysis signal is normalized.
  • Step S106 performing image recognition on the face image reconstructed by using the amplitude value obtained after the normalization process.
  • the parsed image of the face image is calculated, and the parsing signal is extracted from the parsed image, and the parsing signal is used to represent the collected face image.
  • the magnitude value of the information carrying the characterization light in the new model is analyzed, and the amplitude value is further normalized to reduce the amplitude difference between the plurality of amplitude values, thereby eliminating the difference.
  • the difference in brightness between higher brightness and lower brightness improves the robustness of the brightness conversion and reduces the effect of illumination on the captured face image.
  • the face image is reconstructed by using the normalized amplitude value, and the reconstructed face image eliminates the influence of the illumination change on the captured face image on the face recognition, so that when the reconstructed face image is used for image recognition
  • the invention can improve the accuracy of face recognition, thereby solving the technical problem that the recognition rate of face recognition in the prior art is relatively low, and the effect of improving the recognition rate of face recognition is achieved.
  • the 2D-analyzed image can be calculated for the collected face image, and the 2D-analyzed image can be defined in the following two manners: that is, acquiring the parsed signal from the parsed image of the face image includes: acquiring the face image 2 Dimensional analysis of the image, and obtaining two Hilbert transform analysis signals from the two-dimensional analytical image; or acquiring a two-dimensional analytical image of the face image, and acquiring the quaternion analysis signal from the two-dimensional analytical image.
  • f is a 2-dimensional analytical image of the face image
  • ⁇ 1 and ⁇ 3 are the mapping values of the Hilbert transform of the 2D analytical image
  • a 1 and a 3 are the amplitudes of the two Hilbert transform analytical signals.
  • Is to analyze the phase value of the signal i' is the imaginary unit
  • H ⁇ f ⁇ is the Hilbert transform of the 2D analytical image f
  • H x ⁇ f ⁇ is the Hilbert transform in the x direction of the 2D analytical image f
  • H y ⁇ f ⁇ is a Hilbert transform in the y direction of the two-dimensional analytical image f
  • normalizing the amplitude values in the analysis signal includes normalizing the a 1 and a 3 .
  • a 1 and a 3 are amplitude values of two Hilbert transform analysis signals, and a 1 and a 3 are normalized to make 2 Hilbert
  • the amplitude values of the special transform analysis signal are all converted into relative values, which reduces the amplitude difference between different amplitude values, thereby narrowing the luminance difference between different luminances, thereby improving the robustness to the luminance variation and avoiding the acquisition.
  • the face image has a large change in the brightness change.
  • f is a 2-dimensional analytical image of the face image
  • ⁇ q is the mapping value of the Hilbert transform of the 2D analytical image
  • a q is the amplitude value of the quaternion analysis signal.
  • Is the phase value, i, j, and k are the imaginary units of the quaternion
  • H ⁇ f ⁇ is the Hilbert transform of the 2D analytical image f
  • H x ⁇ f ⁇ is the x direction of the 2D analytical image f
  • the Albert transform, H y ⁇ f ⁇ is a Hilbert transform in the y direction of the two-dimensional analytical image f
  • normalizing the amplitude values in the analytical signal includes normalizing the a q .
  • aq is the amplitude value of the quaternion analysis signal, normalizes the aq , and eliminates the information of the partial brightness change, and the information of the brightness change is for the face recognition. Negative influences, information that causes inaccurate face recognition. Therefore, face images reconstructed using normalized amplitude values no longer carry the information of these eliminated brightness changes, thereby improving the recognition rate and accuracy of face recognition. rate.
  • the collected face image may also adopt other forms of two-dimensional analysis images, and normalize the amplitude values in the analysis signals of the two-dimensional analysis images to perform face recognition.
  • performing image recognition on the face image reconstructed by using the amplitude value obtained by the normalization process includes: inputting the amplitude value obtained by the normalization process into the parsing signal; and obtaining the normalized processing by using the input
  • the analytic signal of the amplitude value reconstructs the face image; the face image is reconstructed for the reconstructed face image.
  • the amplitude value obtained after the normalization process is input to the above-mentioned analysis signal.
  • a 1 and a 3 are normalized to obtain a 1nor and a 3nor
  • a q is normalized to obtain a qnor .
  • f new (x, y) is the reconstructed face image
  • a 1nor and a 3nor are the normalized amplitude values of the two Hilbert transform analysis signals. with Is to analyze the phase value of the signal.
  • a qnor is brought into the above quaternion resolution signal to replace the a q before normalization, and the reconstructed image is brought as follows:
  • f new (x, y) is the reconstructed face image
  • a qnor is the normalized amplitude value of the quaternion resolution signal
  • An embodiment of the present invention further provides an apparatus for image recognition, which may perform the above method for image recognition.
  • the image recognition apparatus includes: an acquisition unit 10, a processing unit 20, and an identification unit. 30.
  • the obtaining unit 10 is configured to obtain an analytical signal from the analytical image of the facial image, wherein the amplitude value of the analytical signal carries information representing the illumination in the human face image.
  • the processing unit 20 is configured to normalize the amplitude values in the parsed signal.
  • the identification unit 30 is configured to perform image recognition on the face image reconstructed using the amplitude value obtained after the normalization process.
  • the parsed image of the face image is calculated, and the parsing signal is extracted from the parsed image, and the parsing signal is used to represent the collected face image.
  • the magnitude value of the information carrying the characterization light in the new model is analyzed, and the amplitude value is further normalized to reduce the amplitude difference between the plurality of amplitude values, thereby eliminating the difference.
  • the difference in brightness between higher brightness and lower brightness improves the robustness of the brightness conversion and reduces the effect of illumination on the captured face image.
  • the face image is reconstructed by using the normalized amplitude value, and the reconstructed face image eliminates the influence of the illumination change on the captured face image on the face recognition, so that when the reconstructed face image is used for image recognition
  • the invention can improve the accuracy of face recognition, thereby solving the technical problem that the recognition rate of face recognition in the prior art is relatively low, and the effect of improving the recognition rate of face recognition is achieved.
  • the 2D-analyzed image can be calculated for the captured face image.
  • the 2D-analyzed image can be defined in the following two manners: As shown in FIG. 3 and FIG. 4, the obtaining unit 10 includes: a first acquiring module 102. And acquiring a two-dimensional analysis image of the face image, and acquiring two Hilbert transform analysis signals from the two-dimensional analysis image; or a second acquisition module 104, configured to acquire a two-dimensional analysis image of the face image, And the quaternion analysis signal is obtained from the 2D analytical image.
  • f is a 2-dimensional analytical image of the face image
  • ⁇ 1 and ⁇ 3 are the mapping values of the Hilbert transform of the 2D analytical image
  • a 1 and a 3 are the amplitudes of the two Hilbert transform analytical signals.
  • Is to analyze the phase value of the signal i' is the imaginary unit
  • H ⁇ f ⁇ is the Hilbert transform of the 2D analytical image f
  • H x ⁇ f ⁇ is the Hilbert transform of the x-direction of the 2D analytical image f
  • H y ⁇ f ⁇ is a Hilbert transform in the y direction of the two-dimensional analytical image f
  • the processing unit is also used to perform normalization processing on a 1 and a 3 , respectively.
  • a 1 and a 3 are amplitude values of two Hilbert transform analysis signals, and a 1 and a 3 are normalized to make 2 Hilbert
  • the amplitude values of the special transform analysis signal are all transformed into relative values with respect to 1, the larger amplitude value is reduced, the smaller amplitude value is increased, and the amplitude difference between the different amplitude values is reduced, thereby reducing the difference between the different brightness values.
  • the difference in brightness also improves the robustness to the change in brightness, and avoids a large change in the acquired face image under the change of brightness.
  • f is a 2-dimensional analytical image of the face image
  • ⁇ q is the mapping value of the Hilbert transform of the 2D analytical image
  • a q is the amplitude value of the quaternion analysis signal.
  • Is the phase value, i, j, and k are the imaginary units of the quaternion
  • H ⁇ f ⁇ is the Hilbert transform of the 2D analytical image f
  • H x ⁇ f ⁇ is the x direction of the 2D analytical image f
  • the Albert transform, H y ⁇ f ⁇ is a Hilbert transform in the y direction of the two-dimensional analytical image f
  • the processing unit is also used to normalize a q .
  • aq is the amplitude value of the quaternion analysis signal, normalizes the aq , and eliminates the information of the partial brightness change, and the information of the brightness change is for the face recognition. Negative influences, information that causes inaccurate face recognition. Therefore, face images reconstructed using normalized amplitude values no longer carry the information of these eliminated brightness changes, thereby improving the recognition rate and accuracy of face recognition. rate.
  • the collected face image may also adopt other forms of two-dimensional analysis images, and normalize the amplitude values in the analysis signals of the two-dimensional analysis images to perform face recognition.
  • the identification unit 30 includes: an input module 302, configured to input the amplitude value obtained by the normalization process into the parsing signal; and a reconstruction module 304, configured to perform normalization processing by using the input The analytic signal of the obtained amplitude value reconstructs the face image; the recognition module 306 is configured to perform face recognition on the reconstructed face image.
  • the amplitude value obtained after the normalization process is input into the above-mentioned analytical signal.
  • a 1 and a 3 are normalized to obtain a 1nor and a 3nor respectively
  • a q is normalized to obtain a qnor .
  • f new (x, y) is the reconstructed face image
  • a 1nor and a 3nor are the normalized amplitude values of the two Hilbert transform analysis signals. with Is to analyze the phase value of the signal.
  • a qnor is brought into the above quaternion resolution signal to replace the a q before normalization, and the reconstructed image is brought as follows:
  • f new (x, y) is the reconstructed face image
  • a qnor is the normalized amplitude value of the quaternion resolution signal
  • the disclosed technical contents may be implemented in other manners.
  • the device embodiments described above are only schematic.
  • the division of the unit may be a logical function division.
  • there may be another division manner for example, multiple units or components may be combined or may be Integrate into another system, or some features can be ignored or not executed.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, unit or module, and may be electrical or otherwise.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
  • the above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.
  • the integrated unit if implemented in the form of a software functional unit and sold or used as a standalone product, may be stored in a computer readable storage medium.
  • the technical solution of the present invention which is essential or contributes to the prior art, or all or part of the technical solution, may be embodied in the form of a software product stored in a storage medium. , including a number of instructions to make one
  • the computer device (which may be a personal computer, server or network device, etc.) performs all or part of the steps of the methods described in various embodiments of the present invention.
  • the foregoing storage medium includes: a U disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic disk, or an optical disk, and the like. .

Abstract

本发明公开了一种图像识别的方法和装置。其中，该方法包括：从人脸图像的解析图像中获取解析信号，其中，解析信号的幅度值携带有人脸图像中表征光照的信息；对解析信号中的幅度值进行归一化处理；对利用归一化处理后得到的幅度值重建的人脸图像进行图像识别。本发明解决了人脸识别的识别率比较低的技术问题。

Description

图像识别的方法和装置 技术领域
本发明涉及人脸识别领域，具体而言，涉及一种图像识别的方法和装置。
背景技术
主成分分析(Principal Component Analysis,PCA)是一种比较准确的人脸识别算法，通过一定的训练,提取人脸图像的主成分,进行非监督的学习方法,进而实现人脸识别和人脸辨别。
现有的主成分分析PCA人脸识别方法,受到光照的影响比较大，对亮度变换的鲁棒性较低。在图像光照有变化的时候,识别率会比较低。
针对上述的问题，目前尚未提出有效的解决方案。
发明内容
本发明实施例提供了一种图像识别的方法和装置，以至少解决人脸识别的识别率比较低的技术问题。
根据本发明实施例的一个方面，提供了一种图像识别的方法，包括：从人脸图像的解析图像中获取解析信号，其中，所述解析信号的幅度值携带有所述人脸图像中表征光照的信息；对所述解析信号中的所述幅度值进行归一化处理；对利用归一化处理后得到的幅度值重建的人脸图像进行图像识别。
可选地，从人脸图像的解析图像中获取解析信号包括：获取所述人脸图像的2维 解析图像，并从所述2维解析图像中获取2个希尔伯特变换解析信号；或者获取所述人脸图像的2维解析图像，并从所述2维解析图像中获取四元数解析信号。
可选地，所述2个希尔伯特变换解析信号为
Figure PCTCN2016109137-appb-000001
Figure PCTCN2016109137-appb-000002
其中，f是所述人脸图像的2维解析图像，ψ1和ψ3为所述2维解析图像的希尔伯特变换的映射值，a1和a3为所述2个希尔伯特变换解析信号的幅度值，
Figure PCTCN2016109137-appb-000003
和
Figure PCTCN2016109137-appb-000004
是解析信号的相位值，i'是虚数单位，H{f}是所述2维解析图像f的希尔伯特变换，Hx{f}是所述2维解析图像f的x方向的希尔伯特变换，Hy{f}是所述2维解析图像f的y方向的希尔伯特变换，对所述解析信号中的所述幅度值进行归一化处理包括：对所述a1和所述a3进行归一化处理。
可选地，所述四元数解析信号为
Figure PCTCN2016109137-appb-000005
其中，f是所述人脸图像的2维解析图像，ψq为所述2维解析图像的希尔伯特变换的映射值，aq为四元数解析信号的幅度值，
Figure PCTCN2016109137-appb-000006
和
Figure PCTCN2016109137-appb-000007
是相位值，i、j和k是四元数的虚数单位，H{f}是所述2维解析图像f的希尔伯特变换，Hx{f}是所述2维解析图像f的x方向的希尔伯特变换，Hy{f}是所述2维解析图像f的y方向的希尔伯特变换，对所述解析信号中的所述幅度值进行归一化处理包括：对所述aq进行归一化处理。
可选地，对利用归一化处理后得到的幅度值重建的人脸图像进行图像识别包括： 将所述归一化处理后得到的幅度值输入到所述解析信号中；利用输入有所述归一化处理后得到的幅度值的解析信号重建人脸图像；对重建的人脸图像进行人脸识别。
可选地，在所述解析信号为2个希尔伯特变换解析信号时，利用输入有所述归一化处理后得到的幅度值的解析信号重建人脸图像，得到的人脸图像如下：
Figure PCTCN2016109137-appb-000008
其中，fnew(x,y)为重建的人脸图像，a1nor和a3nor为所述2个希尔伯特变换解析信号的归一化后的幅度值，
Figure PCTCN2016109137-appb-000009
和
Figure PCTCN2016109137-appb-000010
是解析信号的相位值。
可选地，在所述解析信号为四元数解析信号时，利用输入有所述归一化处理后得到的幅度值的解析信号重建人脸图像，得到的人脸图像如下：
Figure PCTCN2016109137-appb-000011
其中，fnew(x,y)为重建的人脸图像，aqnor为所述四元数解析信号的归一化后的幅度值。
根据本发明实施例的一个方面，提供了一种图像识别的装置，包括：获取单元，用于从人脸图像的解析图像中获取解析信号，其中，所述解析信号的幅度值携带有所述人脸图像中表征光照的信息；处理单元，用于对所述解析信号中的所述幅度值进行归一化处理；识别单元，用于对利用归一化处理后得到的幅度值重建的人脸图像进行图像识别。
可选地，获取单元包括：第一获取模块，用于获取所述人脸图像的2维解析图像，并从所述2维解析图像中获取2个希尔伯特变换解析信号；或者第二获取模块，用于获取所述人脸图像的2维解析图像，并从所述2维解析图像中获取四元数解析信号。
可选地，所述2个希尔伯特变换解析信号为
Figure PCTCN2016109137-appb-000012
Figure PCTCN2016109137-appb-000013
其中，f是所述人脸图像的2维解析图像，ψ1和ψ3为所述2维解析图像的希尔伯特变换的映射值，a1和a3为所述2个希尔伯特变换解析信号的幅度值，
Figure PCTCN2016109137-appb-000014
和
Figure PCTCN2016109137-appb-000015
是解析信号的相位值，i'是虚数单位，H{f}是所述2维解析图像f的希尔伯特变换，Hx{f}是所述2维解析图像f的x方向的希尔伯特变换，Hy{f}是所述2维解析图像f的y方向的希尔伯特变换，所述处理单元还用于对所述a1和所述a3分别进行归一化处理。
可选地，所述四元数解析信号为
Figure PCTCN2016109137-appb-000016
其中，f是所述人脸图像的2维解析图像，ψq为所述2维解析图像的希尔伯特变换的映射值，aq为四元数解析信号的幅度值，
Figure PCTCN2016109137-appb-000017
和
Figure PCTCN2016109137-appb-000018
是相位值，i、j和k是四元数的虚数单位，H{f}是所述2维解析图像f的希尔伯特变换，Hx{f}是所述2维解析图像f的x方向的希尔伯特变换，Hy{f}是所述2维解析图像f的y方向的希尔伯特变换，所述处理单元还用于对所述aq进行归一化处理。
可选地，所述识别单元包括：输入模块，用于将所述归一化处理后得到的幅度值输入到所述解析信号中；重建模块，用于利用输入有所述归一化处理后得到的幅度值的解析信号重建人脸图像；识别模块，用于对重建的人脸图像进行人脸识别。
可选地，在所述解析信号为2个希尔伯特变换解析信号时，所述重建模块利用输入有所述归一化处理后得到的幅度值的解析信号重建人脸图像，得到的人脸图像如下：
Figure PCTCN2016109137-appb-000019
其中，fnew(x,y)为重建的人脸图像，a1nor和a3nor为所述2个希尔伯特变换解析信号的归一化后的幅度值，
Figure PCTCN2016109137-appb-000020
和
Figure PCTCN2016109137-appb-000021
是解析信号的相位值。
可选地，在所述解析信号为四元数解析信号时，所述重建模块利用输入有所述归一化处理后得到的幅度值的解析信号重建人脸图像，得到的人脸图像如下：
Figure PCTCN2016109137-appb-000022
其中，fnew(x,y)为重建的人脸图像，aqnor为所述四元数解析信号的归一化后的幅度值。
在本发明实施例中，采用从人脸图像的解析图像中获取解析信号，其中，解析信号的幅度值携带有人脸图像中表征光照的信息；对解析信号中的幅度值进行归一化处理；对利用归一化处理后得到的幅度值重建的人脸图像进行图像识别的方式，在对采集的人脸图像进行识别之前，对解析新型号中携带有表征光照的信息的幅度值，将幅度值继续归一化处理，降低多个幅度值之间的幅度差，也就消除了较高亮度和较低亮度之间的亮度差，提高了亮度变换的鲁棒性，降低了光照对采集到的人脸图像的影响。然后利用归一化处理后的幅度值重建人脸图像，重建的人脸图像消除了采集的人脸图像中光照变化对人脸识别的影响，从而在利用重建的人脸图像在进行图像识别时，能够提高人脸识别的准确率，从而解决了现有技术人脸识别的识别率比较低的技术问题，达到了提高人脸识别的识别率的效果。
附图说明
此处所说明的附图用来提供对本发明的进一步理解，构成本申请的一部分，本发 明的示意性实施例及其说明用于解释本发明，并不构成对本发明的不当限定。在附图中：
图1是根据本发明实施例的图像识别的方法的流程图；
图2是根据本发明实施例的图像识别的装置的示意图；
图3是根据本发明一实施例的图像识别装置中获取单元包括第一获取模块的示意图；
图4是根据本发明另一实施例的图像识别装置中获取单元包括第二获取模块的示意图；
图5是根据本发明实施例的图像识别装置中识别单元包括多个模块的示意图。
具体实施方式
为了使本技术领域的人员更好地理解本发明方案，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分的实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都应当属于本发明保护的范围。
需要说明的是，本发明的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本发明的实施例能够以除了在这里图示或描述的那些以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、 产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
根据本发明实施例，提供了一种图像识别的方法实施例，需要说明的是，在附图的流程图示出的步骤可以在诸如一组计算机可执行指令的计算机系统中执行，并且，虽然在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤。
该图像识别的方法是在获得人脸图像之后，对获得的人脸图像进行预处理，将人脸图像中携带有光照信息的幅度值进行归一化处理，以消除亮度对人脸图像的影响。然后利用处理后的幅度值重建人脸图像，并对重建后的人脸图像进行识别，提高了人脸识别过程亮度变化的鲁棒性，从而解决了人脸识别过程中光照变化下使得对人脸识别的识别率比较低的技术问题，到达了提高人脸识别率的技术效果。
由于该图像识别的方法中对人脸图像的处理是应用在识别之前的，是对人脸图像的预处理，因此，其不受限于采用何种人脸识别算法，可以应用以PCA为人脸识别算法的人脸识别中，也可以应用以其他人脸识别算法的人脸识别中。
图1是根据本发明实施例的图像识别的方法的流程图，如图1所示，该方法包括如下步骤：
步骤S102，从人脸图像的解析图像中获取解析信号，其中，解析信号的幅度值携带有人脸图像中表征光照的信息。
步骤S104，对解析信号中的幅度值进行归一化处理。
步骤S106，对利用归一化处理后得到的幅度值重建的人脸图像进行图像识别。
在采集到人脸图像后，计算人脸图像的解析图像，并从解析图像中提取出解析信号，该解析信号用来表征采集的人脸图像。在对采集的人脸图像进行识别之前，对解析新型号中携带有表征光照的信息的幅度值，将幅度值继续归一化处理，降低多个幅度值之间的幅度差，也就消除了较高亮度和较低亮度之间的亮度差，提高了亮度变换的鲁棒性，降低了光照对采集到的人脸图像的影响。然后利用归一化处理后的幅度值重建人脸图像，重建的人脸图像消除了采集的人脸图像中光照变化对人脸识别的影响，从而在利用重建的人脸图像在进行图像识别时，能够提高人脸识别的准确率，从而解决了现有技术人脸识别的识别率比较低的技术问题，达到了提高人脸识别的识别率的效果。
可选地，对采集的人脸图像可以计算其2维解析图像，2维解析图像可以以下两种方式进行定义：即从人脸图像的解析图像中获取解析信号包括：获取人脸图像的2维解析图像，并从2维解析图像中获取2个希尔伯特变换解析信号；或者获取人脸图像的2维解析图像，并从2维解析图像中获取四元数解析信号。
(1)2个希尔伯特变换解析信号为
Figure PCTCN2016109137-appb-000023
Figure PCTCN2016109137-appb-000024
其中，f是人脸图像的2维解析图像，ψ1和ψ3为2维解析图像的希尔伯特变换的映射值，a1和a3为2个希尔伯特变换解析信号的幅度值，
Figure PCTCN2016109137-appb-000025
和
Figure PCTCN2016109137-appb-000026
是解析信号的相位值，i'是虚数单位，H{f}是2维解析图像f的希尔伯特变换，Hx{f}是2维解析图像f的x方向的希尔伯特变换，Hy{f}是2维解析图像f的y方向的希尔伯特变换，对解析信号中的幅度值进行归一化处理包括：对a1和a3进行归一化处理。
在上述2个希尔伯特变换解析信号中，a1和a3为2个希尔伯特变换解析信号的幅度值，对a1和a3进行归一化处理，使得2个希尔伯特变换解析信号的幅度值都变换成相对值，缩小了不同幅度值之间的幅度差，从而缩小了不同亮度之间的亮度差，也就提高了对亮度变化的鲁棒性，避免了采集的人脸图像在亮度变化下对有较大的变化。
(2)四元数解析信号为
Figure PCTCN2016109137-appb-000027
其中，f是人脸图像的2维解析图像，ψq为2维解析图像的希尔伯特变换的映射值，aq为四元数解析信号的幅度值，
Figure PCTCN2016109137-appb-000028
和
Figure PCTCN2016109137-appb-000029
是相位值，i、j和k是四元数的虚数单位，H{f}是2维解析图像f的希尔伯特变换，Hx{f}是2维解析图像f的x方向的希尔伯特变换，Hy{f}是2维解析图像f的y方向的希尔伯特变换，对解析信号中的幅度值进行归一化处理包括：对aq进行归一化处理。
在上述四元数解析信号中，aq为四元数解析信号的幅度值，对aq进行归一化处理，消除部分亮度变化的信息，而这部分亮度变化的信息是对人脸识别有负面影响、造成人脸识别不准确的信息，因此，利用归一化后的幅度值重建的人脸图像不再携带有这些消除的亮度变化的信息，从而提高了人脸识别的识别率和准确率。
需要说明的是，采集的人脸图像也可以采用其他形式的2维解析图像，并对2维解析图像的解析信号中的幅度值进行归一化处理，以进行人脸识别。
可选地，对利用归一化处理后得到的幅度值重建的人脸图像进行图像识别包括：将归一化处理后得到的幅度值输入到解析信号中；利用输入有归一化处理后得到的幅度值的解析信号重建人脸图像；对重建的人脸图像进行人脸识别。
将归一化处理后得到的幅度值输入到上述解析信号中，例如，a1和a3归一化后相 应得到a1nor和a3nor，aq归一化后相应得到aqnor。
对于2个希尔伯特变换解析信号，将a1nor和a3nor带入到上述2个希尔伯特变换解析信号中，以替换归一化之前的a1和a3，带入后得到重建的图像如下：
Figure PCTCN2016109137-appb-000030
其中，fnew(x,y)为重建的人脸图像，a1nor和a3nor为2个希尔伯特变换解析信号的归一化后的幅度值，
Figure PCTCN2016109137-appb-000031
和
Figure PCTCN2016109137-appb-000032
是解析信号的相位值。
对于四元数解析信号，将aqnor带入到上述四元数解析信号中，以替换归一化之前的aq，带入后得到重建的图像如下：
Figure PCTCN2016109137-appb-000033
其中，fnew(x,y)为重建的人脸图像，aqnor为四元数解析信号的归一化后的幅度值。
通过上述实施例，可以实现对人脸图像的2维解析图像中的幅度值进行归一化，消除部分亮度变化的信息，再利用归一化后的幅度值进行人脸图像的重建时，消除了这部分亮度信息所带来的降低人脸识别的识别率的问题，解决了人脸识别的识别率低的问题，达到了提高人脸识别的识别率的技术效果。
本发明实施例还提供了一种图像识别的装置，该图像识别的装置可以执行上述图像识别的方法，如图2所示，该图像识别的装置包括：获取单元10、处理单元20和识别单元30。
获取单元10用于从人脸图像的解析图像中获取解析信号，其中，解析信号的幅度值携带有人脸图像中表征光照的信息。
处理单元20用于对解析信号中的幅度值进行归一化处理。
识别单元30用于对利用归一化处理后得到的幅度值重建的人脸图像进行图像识别。
在采集到人脸图像后，计算人脸图像的解析图像，并从解析图像中提取出解析信号，该解析信号用来表征采集的人脸图像。在对采集的人脸图像进行识别之前，对解析新型号中携带有表征光照的信息的幅度值，将幅度值继续归一化处理，降低多个幅度值之间的幅度差，也就消除了较高亮度和较低亮度之间的亮度差，提高了亮度变换的鲁棒性，降低了光照对采集到的人脸图像的影响。然后利用归一化处理后的幅度值重建人脸图像，重建的人脸图像消除了采集的人脸图像中光照变化对人脸识别的影响，从而在利用重建的人脸图像在进行图像识别时，能够提高人脸识别的准确率，从而解决了现有技术人脸识别的识别率比较低的技术问题，达到了提高人脸识别的识别率的效果。
可选地，对采集的人脸图像可以计算其2维解析图像，2维解析图像可以以下两种方式进行定义：即如图3和图4所示，获取单元10包括：第一获取模块102，用于获取人脸图像的2维解析图像，并从2维解析图像中获取2个希尔伯特变换解析信号；或者第二获取模块104，用于获取人脸图像的2维解析图像，并从2维解析图像中获取四元数解析信号。
(1)2个希尔伯特变换解析信号为
Figure PCTCN2016109137-appb-000034
Figure PCTCN2016109137-appb-000035
其中，f是人脸图像的2维解析图像，ψ1和ψ3为2维解析图像的希尔伯特变换的映射值，a1和a3为2个希尔伯特变换解析信号的幅度值，
Figure PCTCN2016109137-appb-000036
和
Figure PCTCN2016109137-appb-000037
是解析信号的相位值， i'是虚数单位，H{f}是2维解析图像f的希尔伯特变换，Hx{f}是2维解析图像f的x方向的希尔伯特变换，Hy{f}是2维解析图像f的y方向的希尔伯特变换，处理单元还用于对a1和a3分别进行归一化处理。
在上述2个希尔伯特变换解析信号中，a1和a3为2个希尔伯特变换解析信号的幅度值，对a1和a3进行归一化处理，使得2个希尔伯特变换解析信号的幅度值都变换成相对于1的相对值，缩小较大的幅度值，增大较小的幅度值，缩小了不同幅度值之间的幅度差，从而缩小了不同亮度之间的亮度差，也就提高了对亮度变化的鲁棒性，避免了采集的人脸图像在亮度变化下对有较大的变化。
需要说明的是，此处的缩小和增大幅度值是相对于解析信号中的幅度值，以便于重建人脸图像。
(2)四元数解析信号为
Figure PCTCN2016109137-appb-000038
其中，f是人脸图像的2维解析图像，ψq为2维解析图像的希尔伯特变换的映射值，aq为四元数解析信号的幅度值，
Figure PCTCN2016109137-appb-000039
和
Figure PCTCN2016109137-appb-000040
是相位值，i、j和k是四元数的虚数单位，H{f}是2维解析图像f的希尔伯特变换，Hx{f}是2维解析图像f的x方向的希尔伯特变换，Hy{f}是2维解析图像f的y方向的希尔伯特变换，处理单元还用于对aq进行归一化处理。
在上述四元数解析信号中，aq为四元数解析信号的幅度值，对aq进行归一化处理，消除部分亮度变化的信息，而这部分亮度变化的信息是对人脸识别有负面影响、造成人脸识别不准确的信息，因此，利用归一化后的幅度值重建的人脸图像不再携带有这些消除的亮度变化的信息，从而提高了人脸识别的识别率和准确率。
需要说明的是，采集的人脸图像也可以采用其他形式的2维解析图像，并对2维解析图像的解析信号中的幅度值进行归一化处理，以进行人脸识别。
可选地，如图5所示，识别单元30包括：输入模块302，用于将归一化处理后得到的幅度值输入到解析信号中；重建模块304，用于利用输入有归一化处理后得到的幅度值的解析信号重建人脸图像；识别模块306，用于对重建的人脸图像进行人脸识别。
将归一化处理后得到的幅度值输入到上述解析信号中，例如，a1和a3归一化后相应得到a1nor和a3nor，aq归一化后相应得到aqnor。
对于2个希尔伯特变换解析信号，将a1nor和a3nor带入到上述2个希尔伯特变换解析信号中，以替换归一化之前的a1和a3，带入后得到重建的图像如下：
Figure PCTCN2016109137-appb-000041
其中，fnew(x,y)为重建的人脸图像，a1nor和a3nor为2个希尔伯特变换解析信号的归一化后的幅度值，
Figure PCTCN2016109137-appb-000042
和
Figure PCTCN2016109137-appb-000043
是解析信号的相位值。
对于四元数解析信号，将aqnor带入到上述四元数解析信号中，以替换归一化之前的aq，带入后得到重建的图像如下：
Figure PCTCN2016109137-appb-000044
其中，fnew(x,y)为重建的人脸图像，aqnor为四元数解析信号的归一化后的幅度值。
通过上述实施例，可以实现对人脸图像的2维解析图像中的幅度值进行归一化，消除部分亮度变化的信息，减少亮度干扰，再利用归一化后的幅度值进行人脸图像的重建时，消除了这部分亮度信息所带来的降低人脸识别的识别率的问题，解决了人脸 识别的识别率低的问题，达到了提高人脸识别的识别率的技术效果。
上述本发明实施例序号仅仅为了描述，不代表实施例的优劣。
在本发明的上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述的部分，可以参见其他实施例的相关描述。
在本申请所提供的几个实施例中，应该理解到，所揭露的技术内容，可通过其它的方式实现。其中，以上所描述的装置实施例仅仅是示意性的，例如所述单元的划分，可以为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，单元或模块的间接耦合或通信连接，可以是电性或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本发明各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本发明的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一 台计算机设备(可为个人计算机、服务器或者网络设备等)执行本发明各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、移动硬盘、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述仅是本发明的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本发明原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也应视为本发明的保护范围。

Claims (14)

  1. 一种图像识别的方法，其特征在于，包括：
    从人脸图像的解析图像中获取解析信号，其中，所述解析信号的幅度值携带有所述人脸图像中表征光照的信息；
    对所述解析信号中的所述幅度值进行归一化处理；
    对利用归一化处理后得到的幅度值重建的人脸图像进行图像识别。
  2. 根据权利要求1所述的方法，其特征在于，从人脸图像的解析图像中获取解析信号包括：
    获取所述人脸图像的2维解析图像，并从所述2维解析图像中获取2个希尔伯特变换解析信号；或者
    获取所述人脸图像的2维解析图像，并从所述2维解析图像中获取四元数解析信号。
  3. 根据权利要求2所述的方法，其特征在于，所述2个希尔伯特变换解析信号为
    Figure PCTCN2016109137-appb-100001
    Figure PCTCN2016109137-appb-100002
    其中，f是所述人脸图像的2维解析图像，ψ1和ψ3为所述2维解析图像的希尔伯特变换的映射值，a1和a3为所述2个希尔伯特变换解析信号的幅度值，
    Figure PCTCN2016109137-appb-100003
    和
    Figure PCTCN2016109137-appb-100004
    是解析信号的相位值，i'是虚数单位，H{f}是所述2维解析图像f的希尔伯特变换，Hx{f}是所述2维解析图像f的x方向的希尔伯特变换，Hy{f}是所述 2维解析图像f的y方向的希尔伯特变换，
    对所述解析信号中的所述幅度值进行归一化处理包括：对所述a1和所述a3进行归一化处理。
  4. 根据权利要求2所述的方法，其特征在于，所述四元数解析信号为
    Figure PCTCN2016109137-appb-100005
    其中，f是所述人脸图像的2维解析图像，ψq为所述2维解析图像的希尔伯特变换的映射值，aq为四元数解析信号的幅度值，
    Figure PCTCN2016109137-appb-100006
    和
    Figure PCTCN2016109137-appb-100007
    是相位值，i、j和k是四元数的虚数单位，H{f}是所述2维解析图像f的希尔伯特变换，Hx{f}是所述2维解析图像f的x方向的希尔伯特变换，Hy{f}是所述2维解析图像f的y方向的希尔伯特变换，
    对所述解析信号中的所述幅度值进行归一化处理包括：对所述aq进行归一化处理。
  5. 根据权利要求1所述的方法，其特征在于，对利用归一化处理后得到的幅度值重建的人脸图像进行图像识别包括：
    将所述归一化处理后得到的幅度值输入到所述解析信号中；
    利用输入有所述归一化处理后得到的幅度值的解析信号重建人脸图像；
    对重建的人脸图像进行人脸识别。
  6. 根据权利要求5所述的方法，其特征在于，在所述解析信号为2个希尔伯特变换 解析信号时，利用输入有所述归一化处理后得到的幅度值的解析信号重建人脸图像，得到的人脸图像如下：
    Figure PCTCN2016109137-appb-100008
    其中，fnew(x,y)为重建的人脸图像，a1nor和a3nor为所述2个希尔伯特变换解析信号的归一化后的幅度值，
    Figure PCTCN2016109137-appb-100009
    和
    Figure PCTCN2016109137-appb-100010
    是解析信号的相位值。
  7. 根据权利要求5所述的方法，其特征在于，在所述解析信号为四元数解析信号时，利用输入有所述归一化处理后得到的幅度值的解析信号重建人脸图像，得到的人脸图像如下：
    Figure PCTCN2016109137-appb-100011
    其中，fnew(x,y)为重建的人脸图像，aqnor为所述四元数解析信号的归一化后的幅度值。
  8. 一种图像识别的装置，其特征在于，包括：
    获取单元，用于从人脸图像的解析图像中获取解析信号，其中，所述解析信号的幅度值携带有所述人脸图像中表征光照的信息；
    处理单元，用于对所述解析信号中的所述幅度值进行归一化处理；
    识别单元，用于对利用归一化处理后得到的幅度值重建的人脸图像进行图像识别。
  9. 根据权利要求8所述的装置，其特征在于，所述获取单元包括：
    第一获取模块，用于获取所述人脸图像的2维解析图像，并从所述2维解析 图像中获取2个希尔伯特变换解析信号；或者
    第二获取模块，用于获取所述人脸图像的2维解析图像，并从所述2维解析图像中获取四元数解析信号。
  10. 根据权利要求9所述的装置，其特征在于，所述2个希尔伯特变换解析信号为
    Figure PCTCN2016109137-appb-100012
    Figure PCTCN2016109137-appb-100013
    其中，f是所述人脸图像的2维解析图像，ψ1和ψ3为所述2维解析图像的希尔伯特变换的映射值，a1和a3为所述2个希尔伯特变换解析信号的幅度值，
    Figure PCTCN2016109137-appb-100014
    和
    Figure PCTCN2016109137-appb-100015
    是解析信号的相位值，i'是虚数单位，H{f}是所述2维解析图像f的希尔伯特变换，Hx{f}是所述2维解析图像f的x方向的希尔伯特变换，Hy{f}是所述2维解析图像f的y方向的希尔伯特变换，
    所述处理单元还用于对所述a1和所述a3分别进行归一化处理。
  11. 根据权利要求9所述的装置，其特征在于，所述四元数解析信号为
    Figure PCTCN2016109137-appb-100016
    其中，f是所述人脸图像的2维解析图像，ψq为所述2维解析图像的希尔伯特变换的映射值，aq为四元数解析信号的幅度值，
    Figure PCTCN2016109137-appb-100017
    和
    Figure PCTCN2016109137-appb-100018
    是相位值，i、j和k是四元数的虚数单位，H{f}是所述2维解析图像f的希尔伯特变换，Hx{f}是所述2维解析图像f的x方向的希尔伯特变换，Hy{f}是所述2维解析图像f的y 方向的希尔伯特变换，
    所述处理单元还用于对所述aq进行归一化处理。
  12. 根据权利要求8所述的装置，其特征在于，所述识别单元包括：
    输入模块，用于将所述归一化处理后得到的幅度值输入到所述解析信号中；
    重建模块，用于利用输入有所述归一化处理后得到的幅度值的解析信号重建人脸图像；
    识别模块，用于对重建的人脸图像进行人脸识别。
  13. 根据权利要求12所述的装置，其特征在于，在所述解析信号为2个希尔伯特变换解析信号时，所述重建模块利用输入有所述归一化处理后得到的幅度值的解析信号重建人脸图像，得到的人脸图像如下：
    Figure PCTCN2016109137-appb-100019
    其中，fnew(x,y)为重建的人脸图像，a1nor和a3nor为所述2个希尔伯特变换解析信号的归一化后的幅度值，
    Figure PCTCN2016109137-appb-100020
    和
    Figure PCTCN2016109137-appb-100021
    是解析信号的相位值。
  14. 根据权利要求12所述的装置，其特征在于，在所述解析信号为四元数解析信号时，所述重建模块利用输入有所述归一化处理后得到的幅度值的解析信号重建人脸图像，得到的人脸图像如下：
    Figure PCTCN2016109137-appb-100022
    其中，fnew(x,y)为重建的人脸图像，aqnor为所述四元数解析信号的归一化后的幅度值。
PCT/CN2016/109137 2015-12-31 2016-12-09 图像识别的方法和装置 WO2017114133A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201511033082.0 2015-12-31
CN201511033082.0A CN106934335B (zh) 2015-12-31 2015-12-31 图像识别的方法和装置

Publications (1)

Publication Number Publication Date
WO2017114133A1 true WO2017114133A1 (zh) 2017-07-06

Family

ID=59224459

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/109137 WO2017114133A1 (zh) 2015-12-31 2016-12-09 图像识别的方法和装置

Country Status (2)

Country Link
CN (1) CN106934335B (zh)
WO (1) WO2017114133A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110163940B (zh) * 2018-05-24 2023-04-18 腾讯科技（深圳）有限公司 超声图像的显示方法和装置、存储介质及电子装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101271521A (zh) * 2008-05-13 2008-09-24 清华大学 基于各向异性双树复小波包变换的人脸识别方法
US20090141947A1 (en) * 2007-11-29 2009-06-04 Volodymyr Kyyko Method and system of person identification by facial image
CN103400114A (zh) * 2013-07-18 2013-11-20 上海交通大学 针对人脸识别的光照归一化处理系统
CN103500339A (zh) * 2013-09-11 2014-01-08 北京工业大学 一种联合单尺度Retinex算法和归一化结构描述子的光照人脸识别方法
CN103646244A (zh) * 2013-12-16 2014-03-19 北京天诚盛业科技有限公司 人脸特征的提取、认证方法及装置

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4071695A (en) * 1976-08-12 1978-01-31 Bell Telephone Laboratories, Incorporated Speech signal amplitude equalizer
US8503800B2 (en) * 2007-03-05 2013-08-06 DigitalOptics Corporation Europe Limited Illumination detection using classifier chains
CN100361157C (zh) * 2005-03-31 2008-01-09 上海交通大学 多分辨率的四元小波相位匹配方法
CN101089874B (zh) * 2006-06-12 2010-08-18 华为技术有限公司 一种远程人脸图像的身份识别方法
CN103065299B (zh) * 2012-12-22 2016-06-15 深圳先进技术研究院 超声图像边缘提取方法和装置
CN103295010B (zh) * 2013-05-30 2016-06-29 西安理工大学 一种处理人脸图像的光照归一化方法
CN103778434A (zh) * 2014-01-16 2014-05-07 重庆邮电大学 一种基于多分辨率多阈值局部二值模式的人脸识别方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090141947A1 (en) * 2007-11-29 2009-06-04 Volodymyr Kyyko Method and system of person identification by facial image
CN101271521A (zh) * 2008-05-13 2008-09-24 清华大学 基于各向异性双树复小波包变换的人脸识别方法
CN103400114A (zh) * 2013-07-18 2013-11-20 上海交通大学 针对人脸识别的光照归一化处理系统
CN103500339A (zh) * 2013-09-11 2014-01-08 北京工业大学 一种联合单尺度Retinex算法和归一化结构描述子的光照人脸识别方法
CN103646244A (zh) * 2013-12-16 2014-03-19 北京天诚盛业科技有限公司 人脸特征的提取、认证方法及装置

Also Published As

Publication number Publication date
CN106934335A (zh) 2017-07-07
CN106934335B (zh) 2021-02-02

Similar Documents

Publication Publication Date Title
An et al. Face image super-resolution using 2D CCA
Raghavendra et al. Novel image fusion scheme based on dependency measure for robust multispectral palmprint recognition
CN110353675B (zh) 基于图片生成的脑电信号情感识别方法及装置
Hou et al. Saliency detection: A spectral residual approach
Faraji et al. Face recognition under varying illuminations using logarithmic fractal dimension-based complete eight local directional patterns
Wang et al. Living-skin classification via remote-PPG
CN109993068B (zh) 一种基于心率和面部特征的非接触式的人类情感识别方法
Llano et al. Optimized robust multi-sensor scheme for simultaneous video and image iris recognition
CN108388889B (zh) 用于分析人脸图像的方法和装置
Huang et al. A novel iris segmentation using radial-suppression edge detection
Kantarcı et al. Thermal to visible face recognition using deep autoencoders
Zhao et al. Applying contrast-limited adaptive histogram equalization and integral projection for facial feature enhancement and detection
Wu et al. VP-NIQE: An opinion-unaware visual perception natural image quality evaluator
US11244456B2 (en) System and method for image segmentation and digital analysis for clinical trial scoring in skin disease
WO2017092272A1 (zh) 人脸识别方法和装置
CN106940904B (zh) 基于人脸识别和语音识别的考勤系统
CN111814682A (zh) 人脸活体检测方法及装置
WO2017114133A1 (zh) 图像识别的方法和装置
US10839251B2 (en) Method and system for implementing image authentication for authenticating persons or items
CN112861588A (zh) 一种活体检测的方法、装置
Wu et al. Biomedical video denoising using supervised manifold learning
Jian et al. Towards reliable object representation via sparse directional patches and spatial center cues
Ouloul et al. Improvement of age estimation using an efficient wrinkles descriptor
CN112507877A (zh) 一种在视频部分信息缺失情况下检测心率的系统与方法
Le-Tien et al. Combined Zernike moment and multiscale Analysis for tamper detection in digital images

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16880923

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS (EPO FORM 1205A DATED 19.11.2018)

122 Ep: pct application non-entry in european phase

Ref document number: 16880923

Country of ref document: EP

Kind code of ref document: A1