Fingerprinting and recognition of data

A method for determining if a second image is a derivative copy of a first image comprising the steps of:

• a. generating a first set of selected features for the first image using a fingerprinting algorithm which generates at least one resolution specific image feature set, each of said resolution specific image feature sets corresponding to a specific resolution variant of the image being fingerprinted using said fingerprinting algorithm;
• b. generating a second set of selected features for the second image using said fingerprinting algorithm;
• c. creating at least one image feature set pair from said first set of selected features and said second set of selected features, where each image feature set pair consists of at least one resolution specific image feature set from said first set of selected features and at least one resolution specific feature set from said second set of selected features;
• and
• d. starting with an initial low resolution image feature set pair, iteratively:
• i. determining a confidence value for said image feature set pair;
• ii. outputting the last matched pair if said confidence value is greater than a predetermined confidence threshold and there are no more image feature set pairs to process;
• iii. outputting the last matched pair if said confidence value is less than a predetermined confidence and the highest resolution of any image features are greater than a predetermined resolution threshold;
• and
• iv. reporting a mismatch if said confidence value is less than a predetermined confidence and the highest resolution of any image feature is not greater than a predetermined resolution threshold.

An image fingerprinter for generating a fingerprint from an image comprising:

• a. a gradient identifier for converting said image to a gradient image;
• b. an image scaler for resealing said gradient image into a pyramid of various resolution gradient images, said pyramid having a multitude of pyramid levels;
• c. a feature detector for detecting features in each said pyramid of various resolution gradient images;
• d. a feature selector for selecting at least one selected feature point from each of said gradient images;
• e. a feature comparator for comparing said selected features from said gradient images at the same pyramid level for matches;
• f. a feature weight assigner for assigning a weight to each of said features determined by the quantity of matches each of said selected features has;
• and
• g. a feature sorter for sorting said selected features by their assigned weights;
• wherein, said fingerprint includes said sorted features and said assigned weights.

An image matching apparatus for determining whether a second image is a derivative copy of a first image comprising:

• a. an image fingerprinter for generating a first fingerprint from said first image using a first threshold and a second fingerprint from said second image using a second threshold, said fingerprints including sorted features with assigned weights, each of said assigned weights indicating the importance of each of said sorted features, each of said sorted features associated with a specific resolution version of the image fingerprinted;
• and
• b. a comparator for determining matches from said fingerprints further comprising:
• i. an image pairer for creating a multitude of sorted feature pairs, each pair containing at least one sorted feature from said first fingerprint and at least one sorted feature from said second fingerprint;
• ii. a pair comparator for analyzing said features in said pairs for matches;
• iii. match list for holding said matches;
• iv. a triplet creator for creating triplets from said matches in said match list;
• v. a triple area calculator for calculating a first image area and a second image area for each said triplet;
• vi. a triple list for storing said triplets, said first image areas, and said second image areas;
• vii. a line detector for detecting lines in the space defined by said first image areas, and said second image areas;
• viii. a line strength calculator which calculates a line strength for each of said detected lines determined by the number of points on each of said lines;
• and
• ix. a confidence calculator for determining an overall match confidence determined by said line strengths;
• wherein said apparatus reports a match if said confidence value is greater than a confidence threshold.