# US 2008/0187186 A1 · Image processing apparatus, image processing method and computer program > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2008/0187186 A1 - Document ID: US20080187186A1 - Status: Published application - Kind: A1 - Application number: 12/013,050 - Filed: 2008-01-11 - Published: 2008-08-07 - Earliest claimed: 2007-02-02 - Assignees: Sony Corporation - Inventors: Haruo Togashi - Web page: https://patentgrep.com/patent/US20080187186A1 ## CPC - G06V 10/52: Scale-space analysis, e.g. wavelet analysis (multi-scale boundary representations G06V10/42) - G06F 18/214: Generating training patterns; Bootstrap methods, e.g. bagging or boosting - G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting - G06V 40/168: {Feature extraction; Face representation} - G06V 40/172: {Classification, e.g. identification} - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract An image processing apparatus for performing identification processing of an input face image, including: an image conversion section receiving input of the face image to be identified, executing image conversion on the input face image, and performing normalization processing into an image having a predetermined setting; a similarity calculation section calculating similarities between a feature quantity obtained from the converted input image in the image conversion section and a feature quantity of a registered image; and matching determination means determining whether the input face image matches the registered face image on the basis of the similarities calculated by the similarity calculation section, wherein the image conversion section executes the normalization processing by image conversion processing using a center of right and left eyes of the input face image as an origin. ## Claims ### Claim 1 An image processing apparatus for performing identification processing of an input face image, comprising: an image conversion section receiving input of the face image to be identified, executing image conversion on the input face image, and performing normalization processing into an image having a predetermined setting; a similarity calculation section calculating similarities between a feature quantity obtained from the converted input image in the image conversion section and a feature quantity of a registered image; and matching determination mean determining whether the input face image matches the registered face image on the basis of the similarities calculated by the similarity calculation section, wherein the image conversion section executes the normalization processing by image conversion processing using a center of right and left eyes of the input face image as an origin. ### Claim 2 (depends on 1) The image processing apparatus according to claim 1, wherein the image conversion section obtains a face image from a first memory storing the face image to be normalization processed, performs the normalization by image conversion and stores the face image after the normalization into a second memory, and includes a calculation section calculating a conversion parameter for calculating a corresponding point in the first memory to each pixel position in the second memory, a coordinate conversion section calculating the corresponding point in the first memory to each pixel position in the second memory on the basis of the conversion parameter calculated by the calculation section, and an interpolation processing section determining a pixel value corresponding to each pixel position in the second memory in accordance with corresponding point information calculated by the coordinate conversion section. ### Claim 3 (depends on 2) The image processing apparatus according to claim 2, wherein the calculation section calculates, as the conversion parameter, a conversion parameter defining at last any one of image contraction processing, image rotation processing, or image translation processing to be performed when the face image stored in the first memory is converted into the face image stored in the second memory, and the coordinate conversion section applies the conversion parameter defining at least any one of the image contraction processing, the image rotation processing, or the image translation processing to perform processing to calculate a corresponding point in the first memory to each pixel position in the second memory. ### Claim 4 (depends on 3) The image processing apparatus according to claim 3, wherein the coordinate conversion section calculates coordinates (s1x, s1y) in the first memory, which is an image storage memory before the normalization, using coordinates (s2x, s2y) in the second memory by expressions as follows: s1x=(c/P)×((s2x)×cos θ+s2y×sin θ)+(org—s1x_frc) s1y=(c/P)×((−s2x)×sin θ+s2y×cos θ)+(org—s1y_frc) where s1x is an x-coordinate in the first memory coordinate system, s1y is a y-coordinate in the first memory coordinate system, s2x is an x-coordinate in the second memory coordinate system, s2y is a y-coordinate in the second memory coordinate system, and an image contraction processing parameter, an image rotation parameter, image translation processing parameters, which are input from the calculation section, are (c/P), θ, (org_s1x_frc) and (org_s1y_frc), respectively. ### Claim 5 (depends on 4) The image processing apparatus according to claim 4, wherein the translation processing parameters (org_s1x_frc, org_s1y_frc) are information indicating a position corresponding to an origin of the second memory in the first memory coordinate system. ### Claim 6 (depends on 2) The image processing apparatus according to claim 2, wherein the interpolation processing section performs pixel-value interpolation processing to which a bi-linear method or a bi-cubic method is applied using the corresponding point information calculated by the coordinate conversion section. ### Claim 7 (depends on 1) The image processing apparatus according to claim 1, wherein the similarity calculation section includes Gabor filter application means performing convolution operation of the Gabor filter coefficients and a pixel while changing a frequency and an angle of a response function of a Gabor filter spatially expressed by a Gabor function using a Gaussian function for a window and a sine function or a cosine function as a basis for a frequency response for each feature-quantity extraction position of the converted input image converted by the image conversion section and obtaining a Gabor jet including a number of scalar values corresponding to a filter type, and similarity operation means calculating similarities between the Gabor jets of the input face image and the Gabor jets of the registered face image for each feature-quantity extraction position, and obtaining a similarity vector including a set of similarities in a plurality of feature-quantity extraction positions, wherein the matching determination mean determines whether the input face image matches the registered face image on the basis of the similarity vector. ### Claim 8 A method of processing an image for performing identification processing of an input face image in an image processing apparatus, the method comprising the steps of: converting an image by an image conversion section receiving input of the face image to be identified, executing image conversion on the input face image, and performing normalization processing into an image having a predetermined setting; calculating similarities, by a similarity calculation section, between a feature quantity obtained from the converted input image in the step of converting an image and a feature quantity of a registered image; and matching determining by a matching determination section determining whether the input face image matches the registered face image on the basis of the similarities calculated by the step of calculating similarities, wherein the step of converting an image includes executing the normalization processing by image conversion processing using a center of right and left eyes of the input face image as an origin. ### Claim 9 (depends on 8) The method of processing an image according to claim 8, wherein the step of converting an image includes obtaining a face image from a first memory storing the face image to be normalization processed, performing the normalization by image conversion, and storing the face image after the normalization into a second memory, and includes calculating a conversion parameter by a calculation section for calculating a corresponding point in the first memory to each pixel position in the second memory, converting coordinates by a coordinate conversion section calculating a corresponding point in the first memory to each pixel position in the second memory on the basis of the conversion parameter calculated by the step of calculating, and interpolation processing by an interpolation processing section determining a pixel value corresponding to each pixel position in the second memory in accordance with the corresponding point information calculated by the step of converting coordinates. ### Claim 10 (depends on 9) The method of processing an image according to claim 9, wherein the step of calculating calculates, as the conversion parameter, a conversion parameter defining at last any one of image contraction processing, image rotation processing, or image translation processing to be performed when the face image stored in the first memory is converted into the face image stored in the second memory, and the step of converting coordinates applies a conversion parameter defining at least any one of the image contraction processing, the image rotation processing, or the image translation processing to perform processing to calculate a corresponding point in the first memory to each pixel position in the second memory. ### Claim 11 (depends on 10) The method of processing an image according to claim 10, wherein the step of converting coordinates calculates coordinates (s1x, s1y) in the first memory, which is an image storage memory before normalization, using coordinates (s2x, s2y) in the second memory by expressions as follows: s1x=(c/P)×((s2x)×cos θ+s2y×sin θ)+(org—s1x_frc) s1y=(c/P)×((−s2x)×sin θ+s2y×cos θ)+(org—s1y_frc) where s1x is an x-coordinate in the first memory coordinate system, s1y is a y-coordinate in the first memory coordinate system, s2x is an x-coordinate in the second memory coordinate system, s2y is a y-coordinate in the second memory coordinate system, and an image contraction processing parameter, an image rotation parameter, image translation processing parameters, which are input from the calculation section, are (c/P), θ, (org_s1x_frc) and (org_s1y_frc), respectively. ### Claim 12 (depends on 11) The method of processing an image according to claim 11, wherein the translation processing parameters (org_s1x_frc, org_s1y_frc) are information indicating a position corresponding to an origin of the second memory in the first memory coordinate system. ### Claim 13 (depends on 9) The method of processing an image according to claim 9, wherein the step of interpolation processing performs pixel-value interpolation processing to which a bi-linear method or a bi-cubic method is applied using the corresponding point information calculated by the coordinate conversion section. ### Claim 14 (depends on 8) The method of processing an image according to claim 8, wherein the step of calculating similarities includes Gabor filter application means performing convolution operation of the Gabor filter coefficients and the pixel while changing a frequency and an angle of a response function of a Gabor filter spatially expressed by a Gabor function using a Gaussian function for a window and a sine function or a cosine function as a basis for a frequency response for each feature-quantity extraction position of the converted input image converted by the image conversion section and obtaining a Gabor jet including a number of scalar values corresponding to a filter type and, similarity operation means calculating similarities between the Gabor jets of the input face image and the Gabor jets of the registered face image, and obtaining a similarity vector including a set of similarities in a plurality of feature-quantity extraction positions, wherein the step of matching determining performs matching determination processing on whether the input face image matches the registered face image on the basis of the similarity vector. ### Claim 15 A computer program for causing an image processing apparatus to perform identification processing of an input face image, the computer program comprising the steps of: converting an image by an image conversion section receiving input of the face image to be identified, executing image conversion on the input face image, and performing normalization processing in an image having a predetermined setting; calculating similarities, by a similarity calculation section, between a feature quantity obtained from the converted input image in the step of converting an image and a feature quantity of a registered image; and matching determining by a matching determination section determining whether the input face image matches the registered face image on the basis of the similarities calculated by the step of calculating similarities, wherein the step of converting an image includes executing the normalization processing by image conversion processing using a center of right and left eyes of the input face image as an origin. ### Claim 16 An image processing apparatus for performing identification processing of an input face image, comprising: an image conversion section receiving input of the face image to be identified, executing image conversion on the input face image, and performing normalization processing into an image having a predetermined setting; a similarity calculation section calculating similarities between a feature quantity obtained from the converted input image in the image conversion section and a feature quantity of a registered image; and a matching determination mechanism determining whether the input face image matches the registered face image on the basis of the similarities calculated by the similarity calculation section, wherein the image conversion section executes the normalization processing by image conversion processing using a center of right and left eyes of the input face image as an origin.