Hierarchical method and system for pattern recognition and edge detection
First Claim
1. A method to recognize model instances in an input signal, the method comprising the steps of:
- storing in a long-term memory a hierarchical collection of models, wherein;
the instances of each model are represented by an ensemble of patterns;
a non primitive pattern, defined on a domain and representative of an instance of a non-primitive model from said hierarchical collection has a fragmentation comprising a clique of local patterns; and
whereineach local pattern, defined on a sub-domain of said domain, is representative of a local model from said hierarchical collection and provides an approximate representation of said non-primitive pattern within said sub-domain;
generating, by a processing means, a plurality of descriptors wherein each of said descriptors specifies an estimated pattern representative of an instance of a model detected in said input signal;
storing said descriptors in a short-term memory;
extracting from said short-term memory, by said processing means, a clique of descriptors which specifies a clique of estimated patterns;
selecting a model to be recognized from said hierarchical collection; and
generating, by said processing means, and by using said clique of descriptors as auxiliary descriptors, a new descriptor that specifies a new pattern, wherein said new pattern has a fragmentation given by said clique of estimated patterns and is representative of an instance of said model to be recognized.
0 Assignments
0 Petitions
Accused Products
Abstract
A method and a system for pattern recognition utilizes an ensemble of reference patterns to represent the possible instances of the models to be recognized; constructs a hierarchy of estimators to simplify and enhance the recognition of the models of interest; approximates complex reference patterns with linear compositions of simpler patterns; fragments complex patterns into local patterns so that interference between the local patterns is sufficiently small for linearization methods to be applicable; constructs estimators during an offline stage to offload calculations from the online signal processing stage; designs model estimators based on optimization principles to enhance performance and to provide performance metrics for the estimated model instances; generates a hierarchy of reference descriptors during the offline stage, which are used for the design and construction of the model estimators. Specific examples are provided for the recognition of image features such as edges and junctions.
27 Citations
20 Claims
-
1. A method to recognize model instances in an input signal, the method comprising the steps of:
-
storing in a long-term memory a hierarchical collection of models, wherein; the instances of each model are represented by an ensemble of patterns; a non primitive pattern, defined on a domain and representative of an instance of a non-primitive model from said hierarchical collection has a fragmentation comprising a clique of local patterns; and
whereineach local pattern, defined on a sub-domain of said domain, is representative of a local model from said hierarchical collection and provides an approximate representation of said non-primitive pattern within said sub-domain; generating, by a processing means, a plurality of descriptors wherein each of said descriptors specifies an estimated pattern representative of an instance of a model detected in said input signal; storing said descriptors in a short-term memory; extracting from said short-term memory, by said processing means, a clique of descriptors which specifies a clique of estimated patterns; selecting a model to be recognized from said hierarchical collection; and generating, by said processing means, and by using said clique of descriptors as auxiliary descriptors, a new descriptor that specifies a new pattern, wherein said new pattern has a fragmentation given by said clique of estimated patterns and is representative of an instance of said model to be recognized. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17)
-
-
18. An apparatus to recognize model instances in an input signal, comprising:
-
means for storing a hierarchical collection of models and a plurality of descriptors, wherein; the instances of each model are represented by an ensemble of patterns; a non-primitive pattern, defined on a domain and representative of an instance of a non-primitive model from said hierarchical collection has a fragmentation comprising a clique of local patterns; and
whereineach local pattern, defined on a sub-domain of said domain, is representative of a local model from said hierarchical collection and provides an approximate representation of said non-primitive pattern within said sub-domain; and said apparatus further comprising processing means adapted to; generate said plurality of descriptors, wherein each of said descriptors specifies an estimated pattern representative of an instance of a model detected in said input signal; extract from said storing means a clique of descriptors which specifies a clique of estimated patterns; select a model to be recognized from said hierarchical collection; and generate, by using said clique of descriptors as auxiliary descriptors, a new descriptor that specifies a new pattern, wherein said new pattern has a fragmentation given by said clique of estimated patterns and is representative of an instance of said model to be recognized. - View Dependent Claims (19)
-
-
20. A computer readable medium for use in an apparatus to recognize model instances, the computer readable medium containing:
-
encoded data representing a hierarchical collection of models, wherein; the instances of each model are represented by an ensemble of patterns; a non primitive pattern, defined on a domain and representative of an instance of a non-primitive model from said hierarchical collection has a fragmentation comprising a clique of local patterns; and
whereineach local pattern, defined on a sub-domain of said domain, is representative of a local model from said hierarchical collection and provides an approximate representation of said non-primitive pattern within said sub-domain; the computer readable medium farther containing instructions to perform a plurality of steps comprising; generating a plurality of descriptors wherein each of said descriptors specifies an estimated pattern representative of an instance of a model detected in said input signal; storing said descriptors in a short-term memory; extracting from said short-term memory a clique of descriptors which specifies a clique of estimated patterns; selecting a model to be recognized from said hierarchical collection; and generating, by using said clique of descriptors as auxiliary descriptors, a new descriptor that specifies a new pattern, wherein said new pattern has a fragmentation given by said clique of estimated patterns and is representative of an instance of said model to be recognized.
-
Specification