Method for finding a most likely matching of a target facial image in a data base of facial images
First Claim
1. A method of finding a most likely match for a target facial image in a data base of facial images comprising;
- creating a data base of facial images in digitized form stored in a digital storage apparatus,digitizing a target facial image from a video image into a digital memory apparatus,establishing a value system to provide a value within the system which is a function of quantization of facial feature measurements,selecting a set of facial features to be subjected to quantization on the digitized target facial image and on the digitized data base facial images,locating the set of facial features on the target image and on each of the data base images by placing a window in a determined position for a selected first facial feature and within said window locating a point or points defining the first facial feature location and locating other facial features directly or indirectly by determined relationship to said point or points defining said first facial feature.quantizing the set of facial features of each facial image in the data base,finding the value in the value system which responds to the quantization of each facial image in the data base,quantizing the target facial image,finding the value in the value system which responds to the quantization of the target image,comparing the value in the value system which responds to the quantization of the target facial image, to each value in the value system which responds to the quantization of each of the facial images in the data base,ordering the facial images in the data base according to closeness of the value in the value system for the data base facial image to that of the target facial image.
0 Assignments
0 Petitions
Accused Products
Abstract
A method of finding a most likely match for a target facial image within a data base of stored facial images comprising determining a score for each data base image as a function of closeness of a quantization of selected facial features between each data base image and the target image and ordering the data base for sequential processing according to the potential value score in descending order, sequentially processing each data base image starting from the highest potential value score by an image comparison process to establish a correlation score for each comparison, and applying one or more decision rules to each comparison to reach a decision.
283 Citations
25 Claims
-
1. A method of finding a most likely match for a target facial image in a data base of facial images comprising;
-
creating a data base of facial images in digitized form stored in a digital storage apparatus, digitizing a target facial image from a video image into a digital memory apparatus, establishing a value system to provide a value within the system which is a function of quantization of facial feature measurements, selecting a set of facial features to be subjected to quantization on the digitized target facial image and on the digitized data base facial images, locating the set of facial features on the target image and on each of the data base images by placing a window in a determined position for a selected first facial feature and within said window locating a point or points defining the first facial feature location and locating other facial features directly or indirectly by determined relationship to said point or points defining said first facial feature. quantizing the set of facial features of each facial image in the data base, finding the value in the value system which responds to the quantization of each facial image in the data base, quantizing the target facial image, finding the value in the value system which responds to the quantization of the target image, comparing the value in the value system which responds to the quantization of the target facial image, to each value in the value system which responds to the quantization of each of the facial images in the data base, ordering the facial images in the data base according to closeness of the value in the value system for the data base facial image to that of the target facial image. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11)
-
-
12. A method of finding a most likely match for a target facial image within a data base of stored facial images comprising;
-
digitizing each data base facial image from a video source, digitizing the target facial image from a video source, ordering the data base facial images according to closeness of value to a value of the target facial image based on similarity of measurements chosen from the group consisting of distances between selected facial features, and intensity ratio of selected facial features, choosing a correlation score as a threshold to select one or more data base facial images chosen from the group consisting of, (a) the only data base facial image which exceeds the threshold (b) all data base facial images which exceed the threshold defining at least one window area containing specified facial features to be applied to the target image and the data base image or images selected by the preceding step, intensity correlating the target facial image successively to each data base facial image starting from the data base facial image which is highest in the order to provide the correlation score said correlating being limited to pixel by pixel correlation within the defined window area. - View Dependent Claims (13, 14, 15)
-
-
16. A method of finding a most likely match for a target facial image within a data base of stored facial images comprising;
-
determining a potential value score for each data base image as a function of closeness of a quantization of at least one of selected facial features between each data base image and the target image, ordering the data base for sequential processing according to potential value score in a descending order, sequentially processing each data base image starting from the highest potential value score by an image comparison process to establish a correlation score for each comparison, said image comparison process comprising defining at least one window area containing specified facial features to be applied to the target image and the data base images and correlating the image portions within said window area, applying one or more decision rules to each comparison to reach a decision.
-
-
17. A method of finding a most likely match for a target facial image within a data base of stored facial images comprising;
-
measuring preselected facial features for each data base image; measuring the same preselected facial features from the target image; establishing a measuring system to provide a value within the measuring system for the feature measurements of the data base images and the target image; ordering the data base images according to nearness of the value to that of the target image in descending order; selecting a portion of the ordered data base images for further processing; defining at least one window area containing specified facial features to be applied to the target image and the data base image or images selected by the preceding step; sequentially correlating the target image to each data base image according to the order said correlating being limited to the said window area; scoring each correlation to define a correlation score within a correlation measuring system for closeness of correlation; comparing a correlation to a selected threshold; reporting as a conclusion according to the following rules; (i) identity of a single data base image which passes the threshold, or; (ii) no data base image passes the threshold; (iii) identity of more than one data base image which passes the threshold. - View Dependent Claims (18, 19)
-
-
20. A method of finding a most likely match for a target facial image with a facial image in a data base of digitally stored facial images;
-
selecting a set of feature measurements to each image in the data base to extract feature measurements for each image; extracting the feature measurements of each data base image by placing a window in a determined position for a selected first facial feature and within said window locating a point or points defining the feature location and locating other facial features by direct or indirect relationship to said first facial feature. quantizing the feature measurements of each image in the data base according to a predetermined quantization scheme having a range; digitizing the target image; selecting the same set of feature measurements to the target image to extract feature measurements for the target image; extracting the feature measurements of the target image by placing a window in determined position for a selected first facial feature and within said window locating a point or points defining the feature location and locating other facial features by direct or indirect relationship to said first facial feature; quantizing the feature measurements of the target image according to the same quantization scheme; comparing the quantization of each image in the data base to the quantization of the target; identifying an order for the data base images wherein the quantization closest to that of the target image is highest and descending sequentially; correlating the target image to data base images starting with the highest in the order and descending sequentially; reporting the degree of correlation for each data base image. - View Dependent Claims (21, 22, 23)
-
-
24. A method of finding a match for a target facial image within a stored data base of known facial images;
-
extracting feature data elements for each facial image in the stored data base; extracting matching feature data elements for the target facial image; converting each of said feature data elements to a quantization which defines for each face a point in n-dimensional space, n being the number of feature data elements; establishing a score for each data base facial image face, defined as the P-norm distance in N-dimensional space between that data base facial image and the target image; establishing a potential value score for each image in the data base relative to the target face defined as the P-norm distance in the N-dimensional space; identifying an order of potential value scores for the data base images as the lowest to higher P-norm distances; comparing the one or more of data base images to the target image sequentially commencing with the data base image having the highest potential value score; said comparing comprising defining at least one window area containing specified facial features to be applied to the target image and the data base image or images and correlating said target image sequentially with said data base image or images said correlating being limited to the defined window area.
-
-
25. A method of finding a most likely match for a target facial image within a data base of stored facial images;
-
establishing a set of facial feature measurements for application to the target facial image and to facial images in the data base; intensity enhancing at least selected portions of said target facial image and the data base facial images; determining a portion for each image to be used in locating the set of facial features to be measured; finding the set of facial features to be measured in the target facial image and in the data base images by scanning each of the images in the determined portion of the image and locating within each scanned portion a predetermined degree of intensity variation detail as being the location of a preselected facial feature; determining the set of facial feature measurements from the locations of the preselected facial feature; applying the facial feature measurements to the target facial image to create a numerical quantization representing the target facial image of n quantities for the target face where n is 1 or more; applying the same feature measurements to the facial images in the data base to create a numerical quantization of n quantities; defining nt for the target facial image as a point in n dimensional space having n coordinates; defining nr for each facial image in the data base as a point in n dimensional space having n coordinates; ordering the images in the data base to create a list in order of their P-norm distances from the lowest P-norm distance to progressively higher P-norm distances; video image adjusting the data base image having the lowest P-norm distance for dimensional and lighting intensity equalization to the target image; comparing the equalized data base image to the target image; correlating the target image digital data sequentially to each of the data base images digital data beginning with the highest ordered image; and reporting the degree of correlation as a correlation score.
-
Specification