Information processing apparatus and information processing method

• US 7,734,095 B2
• Filed: 10/18/2005
• Issued: 06/08/2010
• Est. Priority Date: 10/19/2004
• Status: Expired due to Fees

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Tojo, Hiroshi
• Primary Examiner(s): Bali; Vikkram
• Assistant Examiner(s): ENTEZARI, MICHELLE M
• Application Number:
  US11/251,789
• Publication Number:
  US 20060083443A1
• Time in Patent Office: 1,694 Days
• Field of Search: 382/305, 382/190
• US Class Current: 382/190
• CPC Class Codes: G06V 10/50 by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis

Abstract

Distances between a query image and registered images are calculated for respective feature amounts, and weight values for the calculated distances are calculated on the basis of statistical amount of the feature amounts of the query image (S507). Similarities between the registered images and query image are calculated on the basis of the results obtained by weighting the distances for respective feature amounts by the calculated weight values (S508). Images similar to the query image are displayed as a list on the basis of the calculated similarities (S510).

Claims

1. An information processing apparatus comprising:

holding means for holding a plurality of images;

input means for inputting a query image;

feature amount acquisition means for dividing the query image into a plurality of blocks, and calculating plural kinds of feature amounts including a luminance feature amount and an edge feature amount for each of the divided blocks;

variance calculation means for calculating a variation of all feature amounts belonging to an identical kind, for each of the plural kinds;

similarity calculations means for calculating a similarity between the query image and an image of interest held in said holding means for each of the plural kinds;

weight calculation means for calculating a weight value for each of the plural kinds, wherein the weight value is larger for a kind with a large variation and is smaller for a kind with a small variation;

weighting means for weighting the similarity calculated for each of the plural kinds by using the weight value calculated for each of the plural kinds, respectively; and

retrieving means for retrieving a similar candidate, from said holding means, by using the weighted similarities.

2. The apparatus according to claim 1, wherein when a variation is larger than a predetermined threshold, said weighting means weights by using a maximum weight value, and when a variation is smaller than the predetermined threshold, said weighting means weights by using a weight value that decreases with decreasing variation.

3. The apparatus according to claim 1, wherein letting Var be the variation, Var_th be a predetermined threshold for the variation, W be a weight value, and W_max be a maximum value predetermined for the weight value, said weighting means weights by using the weight value obtained by calculating:

W=W_max ×

(Var/Var_th)

if Var≦

Var_th, and

W=W_max

if Var>

Var_th.

4. An information processing method to be executed by an information processing apparatus which comprises holding means for holding a plurality of images, the method comprising:

an input step of inputting a query image;

a feature amount acquisition step of dividing the query image into a plurality of blocks, and calculating plural kinds of feature amounts including a luminance feature amount and an edge feature amount for each of the divided blocks;

a variance calculation step of calculating a variation of all feature amounts belonging to an identical kind, for each of the plural kinds;

a similarity calculation step of calculating a similarity between the query image and an image of interest held in said holding means for each of the plural kinds;

weight calculation step of calculating a weight value for each of the plural kinds, wherein the weight value is larger for a kind with a large variation and is smaller for a kind with a small variation;

a weighting step of weighting the similarity calculated for each of the plural kinds by using the weight value calculated for each of the plural kinds, respectively; and

a retrieving step of retrieving a similar candidate, from said holding means, by using the weighted similarities.

5. A non-transitory computer readable storage medium storing a computer executable program, the computer-executable program being executable by a computer so as to cause the computer to execute an information processing method of claim 4.