Information processing apparatus and information processing method

• US 20060083443A1
• Filed: 10/18/2005
• Published: 04/20/2006
• Est. Priority Date: 10/19/2004
• Status: Active Grant

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Tojo, Hiroshi
• Granted Patent: US 7,734,095 B2
• Time in Patent Office: Days
• US Class Current: 382/305
• 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 characterized by comprising:

holding means for holding a plurality of sets of media and at least two feature amounts of the media in correspondence with each other;

read means for reading out the at least two feature amounts from said holding means;

input means for inputting a medium as a query;

feature amount acquisition means for acquiring at least two feature amounts of the medium as the query;

calculation means for calculating similarities between corresponding feature amounts of the at least two feature amounts read out by said read means, and the at least two feature amounts acquired by said feature amount acquisition means;

weight calculation means for calculating weight values respectively for the similarities for respective feature amounts calculated by said calculation means on the basis of statistical amounts of the at least two feature amounts acquired by said feature amount acquisition means;

similarity calculation means for calculating similarities between media corresponding to the at least two feature amounts read out by said read means, and the medium as the query, on the basis of results obtained by weighting the similarities for respective feature amounts calculated by said calculation means by the weight values calculated by said weight calculation means; and

output means for outputting similar candidates for the medium as the query of the media held by said holding means on the basis of the similarities calculated by said similarity calculation means.

3. The apparatus according to claim 1, characterized in that the statistical amount used in said weight calculation means is a variance of the feature amount.

2. An information processing apparatus characterized by comprising:

holding means for holding a plurality of sets of images and at least two feature amounts of the images in correspondence with each other;

read means for reading out the at least two feature amounts from said holding means;

input means for inputting an image as a query;

feature amount acquisition means for acquiring at least two feature amounts of the image as the query;

calculation means for calculating similarities between corresponding feature amounts of the at least two feature amounts read out by said read means, and the at least two feature amounts acquired by said feature amount acquisition means;

weight calculation means for calculating weight values respectively for the similarities for respective feature amounts calculated by said calculation means on the basis of statistical amounts of the at least two feature amounts acquired by said feature amount acquisition means;

similarity calculation means for calculating similarities between images corresponding to the at least two feature amounts read out by said read means, and the image as the query, on the basis of results obtained by weighting the similarities for respective feature amounts calculated by said calculation means by the weight values calculated by said weight calculation means; and

output means for outputting similar candidates for the image as the query of the images held by said holding means on the basis of the similarities calculated by said similarity calculation means.

4. The apparatus according to claim 2, characterized in that the statistical amount used in said weight calculation means is a variance of the feature amount.

5. The apparatus according to claim 4, characterized in that when a value of the variance of the feature amount acquired by said feature amount acquisition means is larger than a predetermined threshold, said weight calculation means sets a maximum weight value, and when the value of the variance is smaller than the predetermined threshold, said weight calculation means sets a smaller weight value with decreasing variance value.

6. The apparatus according to claim 4, characterized in that letting Var be the variance of the feature amount acquired by said feature amount acquisition means, Var_th be a predetermined threshold for the variance, W be a weight value for a similarity of the feature amount, and W_max be a maximum value predetermined for the weight value, said weight calculation means calculates the weight value for the similarity of the feature amount by calculating:

Wg=W_max×

(1−

Var/Var_—

th)

if Var≦

Var_th, and

W=W_max

if Var>

Var_th.

7. The apparatus according to claim 4, characterized in that said weight calculation means calculates the variance of the feature amount acquired by said feature amount acquisition means, and determines to set a maximum weight value when a value of the variance is larger than a predetermined threshold, and to set zero weight value when the value of the variance is smaller than the predetermined threshold.

8. An information processing apparatus characterized by comprising:

holding means for holding a plurality of sets of images and at least two feature amounts of the images in correspondence with each other;

total information storage means for storing information that totals a statistical amount of the feature amounts held by said holding means;

read means for reading out the at least two feature amounts from said holding means;

input means for inputting an image as a query;

feature amount acquisition means for acquiring at least two feature amounts of the image as the query;

calculation means for calculating similarities between corresponding feature amounts of the at least two feature amounts read out by said read means, and the at least two feature amounts acquired by said feature amount acquisition means;

weight calculation means for calculating weight values respectively for the similarities for respective feature amounts calculated by said calculation means on the basis of statistical amounts of the at least two feature amounts acquired by said feature amount acquisition means and the total information stored in said total information storage means;

similarity calculation means for calculating similarities between images corresponding to the at least two feature amounts read out by said read means, and the image as the query, on the basis of results obtained by weighting the similarities for respective feature amounts calculated by said calculation means by the weight values calculated by said weight calculation means; and

output means for outputting similar candidates for the image as the query of the images held by said holding means on the basis of the similarities calculated by said similarity calculation means.

9. The apparatus according to claim 8, characterized in that the statistical amount of the feature amount is a variance, and the total information is the number of feature amounts each having the variance smaller than a predetermined threshold, and

said weight calculation means calculates a variance of the feature amount acquired by said feature amount acquisition means and determines a weight value in accordance with a value of the variance and the number of feature amounts each having the variance smaller than the predetermined threshold.

10. The apparatus according to claim 8, characterized in that said weight calculation means calculates a variance of the feature amount acquired by said feature amount acquisition means, and said weight calculation means determines to set a maximum weight value when a value of the variance is larger than a predetermined threshold, and to set zero weight value when the number of feature amounts each having the variance smaller than the predetermined threshold is small or to set a smaller weight value with decreasing variance when the number of feature amounts each having the variance smaller than the predetermined threshold is large, when the value of the variance is smaller than the predetermined threshold.

11. An information processing method to be executed by an information processing apparatus which comprises holding means for holding a plurality of sets of media and at least two feature amounts of the media in correspondence with each other, characterized by comprising:

a read step of reading out the at least two feature amounts from said holding means;

an input step of inputting a medium as a query;

a feature amount acquisition step of acquiring at least two feature amounts of the medium as the query;

a calculation step of calculating similarities between corresponding feature amounts of the at least two feature amounts read out in the read step, and the at least two feature amounts acquired in the feature amount acquisition step;

a weight calculation step of calculating weight values respectively for the similarities for respective feature amounts calculated in the calculation step on the basis of statistical amounts of the at least two feature amounts acquired in the feature amount acquisition step;

a similarity calculation step of calculating similarities between media corresponding to the at least two feature amounts read out in the read step, and the medium as the query, on the basis of results obtained by weighting the similarities for respective feature amounts calculated in the calculation step by the weight values calculated in the weight calculation step; and

an output step of outputting similar candidates for the medium as the query of the media held by said holding means on the basis of the similarities calculated in the similarity calculation step.

14. A program characterized by making a computer execute an information processing method of claim 11.

15. A computer-readable storage medium characterized by storing a program of claim 14.

12. An information processing method to be executed by an information processing apparatus which comprises holding means for holding a plurality of sets of images and at least two feature amounts of the images in correspondence with each other, characterized by comprising:

a read step of reading out the at least two feature amounts from said holding means;

an input step of inputting an image as a query;

a feature amount acquisition step of acquiring at least two feature amounts of the image as the query;

a calculation step of calculating similarities between corresponding feature amounts of the at least two feature amounts read out in the read step, and the at least two feature amounts acquired in the feature amount acquisition step;

a weight calculation step of calculating weight values respectively for the similarities for respective feature amounts calculated in the calculation step on the basis of statistical amounts of the at least two feature amounts acquired in the feature amount acquisition step;

a similarity calculation step of calculating similarities between images corresponding to the at least two feature amounts read out in the read step, and the image as the query, on the basis of results obtained by weighting the similarities for respective feature amounts calculated in the calculation step by the weight values calculated in the weight calculation step; and

an output step of outputting similar candidates for the image as the query of the images held by said holding means on the basis of the similarities calculated in the similarity calculation step.

13. An information processing method to be executed by an information processing apparatus which comprises holding means for holding a plurality of sets of images and at least two feature amounts of the images in correspondence with each other, characterized by comprising:

a total information storage step of storing information that totals a statistical amount of the feature amounts held by said holding means;

a read step of reading out the at least two feature amounts from said holding means;

an input step of inputting an image as a query;

a feature amount acquisition step of acquiring at least two feature amounts of the image as the query;

a calculation step of calculating similarities between corresponding feature amounts of the at least two feature amounts read out in the read step, and the at least two feature amounts acquired in the feature amount acquisition step;

a weight calculation step of calculating weight values respectively for the similarities for respective feature amounts calculated in the calculation step on the basis of statistical amounts of the at least two feature amounts acquired in the feature amount acquisition step and the total information stored in the total information storage step;

a similarity calculation step of calculating similarities between images corresponding to the at least two feature amounts read out in the read step, and the image as the query, on the basis of results obtained by weighting the similarities for respective feature amounts calculated in the calculation step by the weight values calculated in the weight calculation step; and

an output step of outputting similar candidates for the image as the query of the images held by said holding means on the basis of the similarities calculated in the similarity calculation step.