System and method for uncued discrimination of bated features in image

A method for uncued discrimination of a bated spatial pattern in an image electronically rendered as a frame of image data points, comprising the steps of:

• establishing a general dictionary of predefined multi-dimensional reference functions;
• executing candidate detection in a processor to pre-screen the frame of image data points to identify therein at least one candidate grouping of image data points collectively indicative of a predetermined spatial pattern;
• and,
• executing sparse approximation in a processor to compute for each said candidate grouping an optimal reference function defined with respect to said general dictionary, said sparse approximation including projecting said optimal reference function of each said candidate grouping onto a corresponding portion of the image data frame defined thereby, each projected portion being extracted from the image data frame to leave a residual image data frame;
• wherein during iterative execution of said candidate detection and sparse approximation, the residual image data frame of each iteration provides the image data frame for the immediately succeeding iteration.

A method for uncued discrimination of a bated structural feature in a frame of spatially referenced data points, comprising the steps of:

• establishing a general dictionary of predefined multi-dimensional reference functions;
• executing candidate detection in a processor to pre-screen the frame of spatially referenced data points to identify therein at least one candidate grouping of data points collectively indicative of a predetermined structural feature;
• and,
• executing sparse approximation in a processor to compute for each said candidate grouping an optimal reference function defined with respect to said general dictionary, said sparse approximation including projecting said optimal reference function of each said candidate grouping onto a corresponding portion of the data frame defined thereby, each projected portion being extracted from the data frame to leave a residual data frame;
• wherein during iterative execution of said candidate detection and sparse approximation, the residual data frame of each iteration provides the data frame for the immediately succeeding iteration.