Parallel image-clustering apparatus

Claim 2

A parallel image-clustering apparatus for classifying an image into a plurality of clusters each being formed by a set of picture elements having similar natures to each other, said apparatus comprising: a frame memory for storing therein coded picture elements; a sampling circuit for reading out picture element information at random from said frame memory; a plurality of processing units, the number of which is equal to at least the number of said clusters, said processing units each including: (a) a cluster parameter memory for storing therein parameters defining each nature of a plurality of clusters, (b) a likelihood operation circuit for calculating a likelihood based on said parameters and sampling data of image supplied from said sampling circuit, (c) a likelihood memory for storing therein data of a likelihood obtained by said likelihood operation circuit, (d) a reference data memory for storing therein plural reference data which is used for modifying cluster parameters supplied from an external circuit, the plural reference data allowing the processing units to modify plural cluster parameters in parallel, thereby reducing the number of time units required to perform image clustering, and (e) a reference data modification circuit for modifying data stored in said cluster parameter memory by utilizing said reference data; and a likelihood comparison circuit for comparing together a plurality of likelihoods each stored in said likelihood memory and selecting a cluster having the maximum likelihood, said likelihood comparison circuit inputting reference data to said reference data memory of a processing unit corresponding to a selected cluster.

Claim 3

A parallel image-clustering apparatus for classifying an image into a plurality of clusters each being formed by a set of picture elements having similar natures to each other, said apparatus comprising: a plurality of processing units, the number of which is equal to at least the number of divided images of one image on frame, each of said processing units including: (a) a sub frame memory for storing therein a set of coded picture elements of each of said divided images, (b) a sampling circuit for reading out picture element information at random from said sub frame memory, (c) a plurality of cluster parameter memories each for storing therein parameters defining each of said clusters, (d) a likelihood operation and comparison circuit for calculating respective likelihoods based on said parameters and sampling data of an image supplied from said sampling circuit with respect to all of said cluster parameter memories and comparing said likelihoods with each other to obtain a cluster having the maximum likelihood, (e) a modified data memory for storing therein modified data issued from said likelihood operation and comparison circuit, and (f) a reference data memory for storing therein plural reference data which is used for modifying cluster parameters supplied thereto from an external circuit, the plural reference data allowing the processing units to modify plural cluster parameters in parallel, thereby reducing the number of time units required to perform image clustering, and (g) a reference data modification circuit for modifying data stored in said cluster parameter memory by utilizing said reference data; and a transmission line for transmitting said modified data to said reference data memory of each of said processing units.

Claim 4

A parallel image-clustering apparatus for classifying an image into a plurality of clusters each being formed by a set of picture elements having similar natures to each other, said apparatus comprising: a plurality of processing units, the number of which is equal to at least the number of divided images of one image on frame, each of said processing units including: (a) a sub frame memory for storing therein a set of coded picture elements of each of said divided images, (b) a sampling circuit for reading out picture element information at random from said sub frame memory, (c) a plurality of cluster parameter memories each for storing therein parameters defining each of said clusters, (d) a likelihood operation and comparison circuit for calculating respective likelihoods based on said parameters and sampling data of an image supplied from said sampling circuit with respect to all of said cluster parameter memories and comparing said likelihoods with each other to obtain a cluster having the maximum likelihood, (e) a modified data memory for storing therein modified data issued from said likelihood operation and comparison circuit, and (f) a reference data memory for storing therein plural reference data which is used for modifying cluster parameters supplied thereto from an external circuit, the plural reference data allowing the processing units to modify plural cluster parameters in parallel, thereby reducing the number of time units required to perform image clustering, and (g) a reference data modification circuit for modifying data stored in said cluster parameter memory by utilizing said reference data; and a transmission line for transmitting said modified data to said reference data memory of each of said processing units.

Claim 5

A parallel image-clustering apparatus for classifying an image into a plurality of clusters each being formed by a set of picture elements having similar natures to each other, said apparatus comprising: a plurality of processing units, the number of which is equal to at least the number of divided images of one image on frame, each of said processing units including: (a) a sub frame memory for storing therein a set of coded picture elements of a classified image, (b) a sampling circuit for reading out picture element information at random from said sub frame memory, (c) a plurality of cluster parameter memories each for storing therein parameters defining each of said clusters, (d) a likelihood operation and comparison circuit for calculating respective likelihoods based on said parameters and sampling data of an image supplied from said sampling circuit with respect to all of said cluster parameter memories and comparing said likelihoods with each other to obtain a cluster having the maximum likelihood, (e) a modified data memory for storing therein modified data issued from said likelihood operation and comparison circuit, (f) a reference data memory for storing therein plural reference data which is used for modifying cluster parameters supplied thereto from an external circuit, the plural reference data allowing the processing units to modify plural cluster parameters in parallel, thereby reducing the number of time units required to perform image clustering, and (g) a reference data modification circuit for modifying data stored in said cluster parameter memory by utilizing said reference data; and a transmission line for transmitting said modified data to said reference data memory of predetermined one of said processing units, wherein data-modified in said reference data modification circuit of said predetermined one of the processing units is transmitted to said reference data memory of each of said processing units, and said processing units replace said cluster parameters with data of said reference data memory.

Claim 6

A parallel image-clustering apparatus for classifying an image into a plurality of clusters each being formed by a set of picture elements having similar natures to each other, said apparatus comprising: a plurality of processing units, the number of which is equal to at least the number of divided images of one image on frame, each of said processing units including: (a) a sub frame memory for storing therein a set of coded picture elements of a classified image, (b) a sampling circuit for reading out picture element information at random from said sub frame memory, (c) a plurality of cluster parameter memories each for storing therein parameters defining each of said clusters, (d) a likelihood operation and comparison circuit for calculating respective likelihoods based on said parameters and sampling data of an image supplied from said sampling circuit with respect to all of said cluster parameter memories and comparing said likelihoods with each other to obtain a cluster having the maximum likelihood, (e) a modified data memory for storing therein modified data issued from said likelihood operation and comparison circuit, (f) a reference data memory for storing therein plural reference data which is used for modifying cluster parameters supplied thereto from an external circuit, the plural reference data allowing the processing units to modify plural cluster parameters in parallel, thereby reducing the number of time units required to perform image clustering, and (g) a reference data modification circuit for modifying data stored in said cluster parameter memory by utilizing said reference data; and a transmission line for transmitting said modified data to said reference data memory of predetermined one of said processing units, wherein data-modified in said reference data modification circuit of said predetermined one of the processing units is transmitted to said reference data memory of each of said processing units, and said processing units replace said cluster parameters with data of said reference data memory.