Feature-based detection and context discriminate classification for digital images
First Claim
1. A method of detecting and classifying targets in a digital image, comprising the steps of:
- generating a feature set for each of a plurality of overlapping windowed portions of said image, each feature in said feature set defined by a value indicative of a mathematical measure of a corresponding one of said plurality of overlapping windowed portions;
forming a weighted sum for each of said plurality of overlapping windowed portions using said feature set corresponding thereto;
normalizing each feature in said feature set and said weighted sum for each of said plurality of overlapping windowed portions across said plurality of overlapping windowed portions, wherein a context matrix is defined by a normalized feature set and a normalized weighted sum for each of said plurality of overlapping windowed portions;
forming a score using said context matrix for each of said plurality of overlapping windowed portions;
normalizing said score for each of said plurality of overlapping windowed portions across said plurality of overlapping windowed portions, wherein a normalized score is defined for each of said plurality of overlapping windowed portions;
comparing a threshold criteria to a maximum score defined as the maximum of said normalized weighted sum and said normalized score for each of said plurality of overlapping windowed portions, wherein each of said plurality of overlapping windowed portions having said maximum score satisfying said threshold criteria is classified as a possible target window and wherein said maximum score is indicative of a target classification;
assigning each said possible target window to a group based on location of said possible target window in said image and said maximum score associated with said possible target window;
forming a group score for each said group using said maximum score associated with each said possible target window in said group; and
comparing each said group score to a group threshold criteria, wherein each said group having its corresponding said group score satisfying said group threshold criteria is classified as a target and wherein said group score is indicative of a target classification.
1 Assignment
0 Petitions
Accused Products
Abstract
Detection and classification of targets in a digital image is accomplished by evaluating windowed portions of the image. A feature set is generated for each of a plurality of overlapping windowed portions, a weighted sum is formed for each portion based upon its feature set, and a context matrix is defined for each window. A score is formed from each context matrix and is normalized for each window. A threshold criteria is compared to a maximum score for each window. Each window having its maximum score satisfy the threshold criteria is classified as a possible target window and is assigned to a group based on location of the possible target window and its maximum score. A group score is formed for each group and compared to a group threshold criteria. Each group having its corresponding group score satisfying the group threshold criteria is classified as a target.
-
Citations
17 Claims
-
1. A method of detecting and classifying targets in a digital image, comprising the steps of:
-
generating a feature set for each of a plurality of overlapping windowed portions of said image, each feature in said feature set defined by a value indicative of a mathematical measure of a corresponding one of said plurality of overlapping windowed portions; forming a weighted sum for each of said plurality of overlapping windowed portions using said feature set corresponding thereto; normalizing each feature in said feature set and said weighted sum for each of said plurality of overlapping windowed portions across said plurality of overlapping windowed portions, wherein a context matrix is defined by a normalized feature set and a normalized weighted sum for each of said plurality of overlapping windowed portions; forming a score using said context matrix for each of said plurality of overlapping windowed portions; normalizing said score for each of said plurality of overlapping windowed portions across said plurality of overlapping windowed portions, wherein a normalized score is defined for each of said plurality of overlapping windowed portions; comparing a threshold criteria to a maximum score defined as the maximum of said normalized weighted sum and said normalized score for each of said plurality of overlapping windowed portions, wherein each of said plurality of overlapping windowed portions having said maximum score satisfying said threshold criteria is classified as a possible target window and wherein said maximum score is indicative of a target classification; assigning each said possible target window to a group based on location of said possible target window in said image and said maximum score associated with said possible target window; forming a group score for each said group using said maximum score associated with each said possible target window in said group; and comparing each said group score to a group threshold criteria, wherein each said group having its corresponding said group score satisfying said group threshold criteria is classified as a target and wherein said group score is indicative of a target classification. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8)
-
-
9. A system for detecting and classifying targets in a digital image, comprising:
-
means for generating a feature set for each of a plurality of overlapping windowed portions of said image, each feature in said feature set being defined by a value indicative of a mathematical measure of a corresponding one of said plurality of overlapping windowed portions; a processor for i) forming a weighted sum for each of said plurality of overlapping windowed portions using said feature set corresponding thereto, ii) normalizing each feature in said feature set and said weighted sum for each of said plurality of overlapping windowed portions across said plurality of overlapping windowed portions, wherein a context matrix is defined by a normalized feature set and a normalized weighted sum for each of said plurality of overlapping windowed portions, iii) forming a score using said context matrix for each of said plurality of overlapping windowed portions, iv) normalizing said score for each of said plurality of overlapping windowed portions across said plurality of overlapping windowed portions, wherein a normalized score is defined for each of said plurality of overlapping windowed portions, v) comparing a threshold criteria to a maximum score defined as the maximum of said normalized weighted sum and said normalized score for each of said plurality of overlapping windowed portions, wherein each of said plurality of overlapping windowed portions having said maximum score satisfying said threshold criteria is classified as a possible target window and wherein said maximum score is indicative of a target classification, vi) assigning each said possible target window to a group based on location of said possible target window in said image and said maximum score associated with said possible target window, vii) forming a group score for each said group using said maximum score associated with each said possible target window in said group, and viii) comparing each said group score to a group threshold criteria, wherein each said group having its corresponding said group score satisfying said group threshold criteria is classified as a target and wherein said group score is indicative of a target classification; and at least one output device coupled to said processor for providing an indication that said group score satisfies said group threshold criteria. - View Dependent Claims (10, 11, 12, 13, 14, 15, 16, 17)
-
Specification