Data compression device and data compression method

Patent metadata

Field	Value
Publication number	US-9735803-B2
Application number	US-201415119155-A
Country	US
Kind code	B2
Filing date	Mar 5, 2014
Priority date	Mar 5, 2014
Publication date	Aug 15, 2017
Grant date	Aug 15, 2017

Abstract

A dimensional compression unit 13 multiplies a feature quantity extracted from an image by a check matrix of an error correcting code which consists of binary elements each of which is 0 or 1, to compress the image.

First claim

The invention claimed is:

1. A data compression device that compresses data used for performing a search, recognition or prediction, the data compression device comprising:
a processor; and
memory having stored therein program instructions which cause the processor to
extract features of the data in the form of an n dimensional feature quantity vector x;
perform dimensional compression on the feature quantity vector x by multiplying the feature quantity vector x by a check matrix H of a random code, which consists of binary elements 0 or 1, to generate a compressed m dimensional feature quantity vector y, where n>m; and
reduce a quantization size of each element y k of the compressed feature quantity vector y, wherein said dimensional compression adds values, in said feature quantity vector x corresponding to 1 elements of said check matrix on a row by row basis, to generate the compressed feature quantity vector y.

2. A data compression device that compresses data used for performing a search, recognition or prediction, the data compression device comprising:
a processor; and
memory having stored therein program instructions which cause the processor to
extract features of the data in the form of an n dimensional feature quantity vector x; and
perform dimensional compression on the feature quantity vector x by multiplying the feature quantity vector x by a check matrix H which consists of ternary elements 0, 1 or −1 to generate a compressed m dimensional feature quantity vector y, where n>m,

wherein said dimensional compression adds or subtracts values, in said feature quantity vector x corresponding to 1 or −1 elements of said check matrix on a row by row basis, to generate the compressed feature quantity vector.

3. A data compression method of compressing data used for performing a search, recognition or prediction, the method comprising:
extracting, using a processor, features of the data in the form of an n dimensional feature quantity vector x;
performing, using a processor, dimensional compression on the feature quantity vector x by multiplying the feature quantity vector x by a check matrix which consists of binary elements 0 or 1 to generate a compressed m dimensional feature quantity vector y, where n>m; and
reducing, processor, a quantization size of each element y k of the compressed feature quantity vector y,
wherein said dimensional compression adds or subtracts values, in said feature quantity vector x corresponding to 1 or −1 elements of said check matrix on a row by row basis, to generate the compressed feature quantity vector.

Assignees

• Mitsubishi Electric Corp

Inventors

• Matsumoto Wataru
• Yamazaki Takashi

Classifications

• H03M7/3082 Primary

  Vector coding (for television signals, see H04N19/94) · CPC title

• H03M7/30 Primary

  Compression (speech analysis-synthesis for redundancy reduction G10L19/00; for image communication H04N); Expansion; Suppression of unnecessary data, e.g. redundancy reduction · CPC title

• G06V10/462

  Salient features, e.g. scale invariant feature transforms [SIFT] · CPC title

• H03M13/1102

  Codes on graphs and decoding on graphs, e.g. low-density parity check [LDPC] codes · CPC title

• G06K9/4671

  Physics · mapped topic