Image recognition method, image recognition device, and recording medium

Patent metadata

Field	Value
Publication number	US-2016259995-A1
Application number	US-201615049149-A
Country	US
Kind code	A1
Filing date	Feb 22, 2016
Priority date	Mar 6, 2015
Publication date	Sep 8, 2016
Grant date	—

Abstract

An image recognition method includes: receiving an image; acquiring processing result information including values of processing results of convolution processing at positions of a plurality of pixels that constitute the image by performing the convolution processing on the image by using different convolution filters; determining 1 feature for each of the positions of the plurality of pixels on the basis of the values of the processing results of the convolution processing at the positions of the plurality of pixels included in the processing result information and outputting the determined feature for each of the positions of the plurality of pixels; performing recognition processing on the basis of the determined feature for each of the positions of the plurality of pixels; and outputting recognition processing result information obtained by performing the recognition processing.

First claim

What is claimed is:

1 . An image recognition method performed by a computer of an image recognition device, comprising:
receiving an image; acquiring processing result information including values of processing results of convolution processing at positions of a plurality of pixels that constitute the image by performing the convolution processing on the image by using different convolution filters; determining 1 feature for each of the positions of the plurality of pixels on the basis of the values of the processing results of the convolution processing at the positions of the plurality of pixels included in the processing result information and outputting the determined feature for each of the positions of the plurality of pixels; performing recognition processing on the basis of the determined feature for each of the positions of the plurality of pixels; and outputting recognition processing result information obtained by performing the recognition processing.

2 . The image recognition method according to claim 1 , wherein
the convolution processing using the different convolution filters is convolution processing that is performed on the image by using a plurality of convolution filters that are different in resolution or scale parameter.

3 . The image recognition method according to claim 1 , wherein
the convolution processing using the different convolution filters includes first convolution processing that is performed on the image by using a convolution filter having first resolution and a second convolution processing that is performed on the image by using a convolution filter having second resolution higher than the first resolution.

4 . The image recognition method according to claim 1 , wherein
the convolution processing using the different convolution filters is convolution processing that is performed on the image by using convolution filters that are different in color to be processed.

5 . The image recognition method according to claim 1 , wherein
in the outputting the determined feature for each of the positions of the plurality of pixels, 1 feature for the position is determined by selecting, as the feature at the position, a maximum value among the values of the plurality of processing results at the position, and the determined feature for the position is output.

6 . The image recognition method according to claim 1 , wherein
in the outputting the determined feature for each of the positions of the plurality of pixels, 1 feature for the position is determined by calculating a median or an average of the values of the plurality of processing results at the position and determining the calculated value as the feature at the position, and the determined feature at the position is output.

7 . The image recognition method according to claim 1 , wherein
in the outputting the determined feature for each of the positions of the plurality of pixels, subsampling processing for determining, for each region including a plurality of adjacent pixels, any of the features at positions of the plurality of pixels included in the region as a representative feature representing the region; and the recognition processing is performed on the basis of the representative feature determined in the subsampling processing.

8 . The image recognition method according to claim 7 , wherein
in the subsampling processing, a feature whose value is largest among the features at the positions of the plurality of pixels included in the region is determined as the representative feature.

9 . The image recognition method according to claim 1 , wherein
at least one of the receiving the image, the acquiring the processing result information, the determining the 1 feature for each of the positions of the plurality of pixels and the outputting the feature for each of the positions of the plurality of pixels, the performing the recognition processing and outputting the recognition processing result information is performed by a processor provided in the computer of the image recognition device.

10 . An image recognition device comprising:
an image input unit that receives an image; a convolution processor that acquires processing result information including values of processing results of convolution processing at positions of a plurality of pixels that constitute the image by performing the convolution processing on the image by using different convolution filters, determines 1 feature for each of the positions of the plurality of pixels on the basis of the values of the processing results of the convolution processing at the positions of the plurality of pixels included in the processing result information, outputs the determined feature for each of the positions of the plurality of pixels; a recognition processor that performs recognition processing on the basis of the determined feature for each of the positions of the plurality of pixels that is output by the convolution processor; and a recognition result output unit that outputs recognition processing result information obtained by performing the recognition processing in the recognition processor.

11 . The image recognition device according to claim 10 , wherein
at least one of the image input unit, the convolution processor, the recognition processor, and the recognition result output unit includes a processor.

12 . A non-transitory computer-readable recording medium storing a program for causing a computer of an image recognition device to:
receive an image; acquire processing result information including values of processing results of convolution processing at positions of a plurality of pixels that constitute the image by performing the convolution processing on the image by using different convolution filters; determine 1 feature for each of the positions of the plurality of pixels on the basis of the values of the processing results of the convolution processing at the positions of the plurality of pixels included in the processing result information and output the determined feature for each of the positions of the plurality of pixels; perform recognition processing on the basis of the determined feature for each of the positions of the plurality of pixels; and output recognition processing result information obtained by performing the recognition processing.

Assignees

• Panasonic Ip Man Co Ltd

Inventors

• Ishii Yasunori
• Tsukizawa Sotaro
• Hagawa Reiko

Classifications

• G06V10/82 Primary

  using neural networks · CPC title

• G06F18/24

  Classification techniques · CPC title

• G06F18/24143

  Distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN] · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06F18/251 Primary

  of input or preprocessed data · CPC title