Method and system for image analysis based upon correlation relationships of sub-arrays of a descriptor array

Patent metadata

Field	Value
Publication number	US-9412037-B2
Application number	US-201214370133-A
Country	US
Kind code	B2
Filing date	Oct 12, 2012
Priority date	Jan 2, 2012
Publication date	Aug 9, 2016
Grant date	Aug 9, 2016

Abstract

A method for processing an image, including: identifying a group of keypoints in the image; for each keypoint, calculating a corresponding descriptor array including plural array elements, each array element storing values taken by a corresponding color gradient histogram of a respective sub-region of the image in the neighborhood of the keypoint; for each keypoint, subdividing the descriptor array in at least two sub-arrays each including a respective number of elements of the descriptor array, and generating a compressed descriptor array including a corresponding compressed sub-array for each of the at least two sub-arrays, each compressed sub-array obtained by compressing the corresponding sub-array by vector quantization using a respective codebook; exploiting the compressed descriptor arrays of the keypoints for image analysis. For each keypoint of the group, the subdividing is based on correlation relationships among color gradient histograms with values stored in the elements of the descriptor array of each keypoint.

First claim

The invention claimed is:

1. A method for processing an image, comprising:
identifying a group of keypoints in the image;
for each keypoint of the group:
a) calculating a corresponding descriptor array including a plurality of array elements, each array element storing values taken by a corresponding color gradient histogram of a respective sub-region of the image in the neighborhood of the keypoint;
b) arranging the color gradient histograms in a plurality of correlation families, each correlation family comprising a corresponding set of correlated color gradient histograms having a similar statistical behavior;
c) subdividing the descriptor array in at least two sub-arrays, each sub-array comprising a respective number of elements of the descriptor array;
d) generating a compressed descriptor array comprising a corresponding compressed sub-array for each of said at least two sub-arrays, each compressed sub-array being obtained by compressing the corresponding sub-array of said at least two sub-arrays by means of vector quantization using a respective codebook;
e) arranging the at least two sub-arrays into at least one group of sub-arrays; and
f) using a same codebook for compressing the at least two sub-arrays of the at least one group of sub-arrays; and

exploiting the compressed descriptor arrays of the keypoints of said group for analysing the image, wherein:
for each keypoint of said group, the subdivision of the descriptor array in at least two sub-arrays is carried out based on correlation relationships among the color gradient histograms whose values are stored in the elements of the descriptor array of said each keypoint, and
for each of said at least one group of sub-arrays, an array element in one sub-array of the group occupying a same position as an array element in another sub-array of the group correspond to color gradient histograms belonging to a same correlation family.

2. The method of claim 1 ,
wherein, for each keypoint of the group:
said keypoint is associated with sixteen respective sub-regions in the neighborhood thereof
the corresponding descriptor array includes sixteen array elements, each one corresponding to a respective one among the sixteen sub-regions in the neighborhood thereof; and
arranging the color gradient histograms in of the plurality of the correlation families includes including arranging the color gradient histograms in four correlation families each one including four correlated color gradient histograms.

3. The method of claim 2 , wherein, for each keypoint of the group:
said sixteen respective sub-regions are arranged according to a grid arrangement centered at said keypoint and including four rows and four columns, said arranging the color gradient histograms in the plurality of correlation families includes:
arranging the first and fourth sub-regions of the first row of the grid arrangement and the first and fourth sub-regions of the fourth row of the grid arrangement in a first correlation family;
arranging the second and third sub-regions of the first row of the grid arrangement and the second and third sub-regions of the fourth row of the grid arrangement in a second correlation family;
arranging the first and fourth sub-regions of the second row of the grid arrangement and the first and fourth sub-regions of the third row of the grid arrangement in a third correlation family; and
arranging the color gradient histograms in a plurality of correlation families includes arranging the color gradient histograms in four correlation families each one including four correlated color gradient histograms.

4. The method of claim 3 , further including, for each keypoint of the group:
subdividing the descriptor array in:
a first sub-array, comprising a first element corresponding to the first array element of the descriptor array, a second element corresponding to the second array element of the descriptor array, a third element corresponding to the sixth array element of the descriptor array and a fourth element corresponding to the fifth array element of the descriptor array;
a second sub-array, comprising a first element corresponding to the fourth array element of the descriptor array, a second element corresponding to the third array element of the descriptor array, a third element corresponding to the seventh array element of the descriptor array and a fourth element corresponding to the eighth array element of the descriptor array;
a third sub-array, comprising a first element corresponding to the sixteenth array element of the descriptor array, a second element corresponding to the fifteenth array element of the descriptor array, a third element corresponding to the eleventh array element of the descriptor array and a fourth element corresponding to the twelfth array element of the descriptor array; and
a fourth sub-array, comprising a first element corresponding to the thirteenth array element of the descriptor array, a second element corresponding to the fourteenth array element of the descriptor array, a third element corresponding to the tenth array element of the descriptor array and a fourth element corresponding to the ninth array element of the descriptor array, and

using the same codebook for compressing the first, second, third and fourth sub-arrays.

5. The method of claim 3 , further including, for each keypoint of the group:
subdividing the descriptor array in:
a first sub-array, comprising a first element corresponding to the first array element of the descriptor array, a second element corresponding to the second array element of the descriptor array, a third element corresponding to the third array element of the descriptor array, a fourth element corresponding to the fourth array element of the descriptor array, a fifth element corresponding to the fifth array element of the descriptor array, a sixth element corresponding to the sixth array element of the descriptor array, a seventh element corresponding to the seventh element of the descriptor array, and an eighth element corresponding to the eighth element of the descriptor array, and
a second sub-array, comprising a first element corresponding to the thirteenth array element of the descriptor array, a second element corresponding to the fourteenth array element of the descriptor array, a third element corresponding to the fifteenth array element of the descriptor array, a fourth element corresponding to the sixteenth array element of the descriptor array, a fifth element corresponding to the ninth array element of the descriptor array, a sixth element corresponding to the tenth array element of the descriptor array, a seventh element corresponding to the eleventh element of the descriptor array, and an eighth element corresponding to the twelfth element of the descriptor array, and

using the same codebook for compressing the first and the second sub-arrays.

6. The method of claim 3 , further including, for each keypoint of the group:
subdividing the descriptor array in:
a first sub-array, comprising a first element corresponding to the fifth array element of the descriptor array, a second element corresponding to the first array element of the descriptor array, and a third element corresponding to the second array element of the descriptor array;
a second sub-array, comprising a first element corresponding to the eight array element of the descriptor array, a second element corresponding to the fourth array element of the descriptor array, and a third element corresponding to the third array element of the descriptor array;
a third sub-array, comprising a first element corresponding to the ninth array element of the descriptor array, a second element corresponding to the thirteenth array element of the descriptor array, and a third element corr

Assignees

• Telecom Italia Spa

Inventors

• Balestri Massimo
• Francini Gianluca
• Lepsoy Skjalg

Classifications

• G06F2218/12

  Classification; Matching · CPC title

• G06F16/583 Primary

  using metadata automatically derived from the content · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06K9/6256

  Physics · mapped topic

• G06K9/4642

  Physics · mapped topic