System and method for detecting and ranking images in order of usefulness based on vignette score

• US 7,492,921 B2
• Filed: 01/10/2005
• Issued: 02/17/2009
• Est. Priority Date: 01/10/2005
• Status: Expired due to Fees

• Current Assignee: Fuji Xerox Company Limited (Fujifilm Holdings Corporation)
• Original Assignee: Fuji Xerox Company Limited (Fujifilm Holdings Corporation)
• Inventors: Foote, Jonathan T.
• Primary Examiner(s): Dang; Duy M
• Application Number:
  US11/032,576
• Publication Number:
  US 20060153456A1
• Time in Patent Office: 1,499 Days
• Field of Search: 382/100, 382/132, 382/173, 382/224
• US Class Current: 382/100
• CPC Class Codes: G06F 16/583 using metadata automatically derived from the content

Abstract

A system and method for detecting useful images and for ranking images in order of usefulness based on a vignette score describing how closely each one resembles a “vignette,” or a central object or image surrounded by a featureless or deemphasized background. Several methods for determining an image'"'"'s vignette score are disclosed as examples. Variance ratio analysis entails calculation of the ratio of variance between the edge region of the image and the entire image. Statistical model analysis entails developing a statistical classifier capable of determining a statistical model of each image class based on pre-entered training data. Spatial frequency analysis involves estimating the energy at different spatial frequencies in the central and edge regions and in the image as a whole. A vignette score is calculated as the ratio of mid-frequency energies in the edge region to the mid-frequency energies of the entire image.

Claims

1. A method for detecting useful images and for ranking images in order of usefulness, comprising the step of:

a) computing a vignette score for each image in a given set of images, wherein a vignette score describes how closely an image resembles a central object surrounded by a featureless or deemphasized background; and

b) ranking the images with numerical ranks according to each image'"'"'s vignette score.

2. The method of claim 1, wherein the step of computing a vignette score utilizes variance ratio analysis.

3. The method of claim 1, wherein the step of computing a vignette score for each image in a given set of images comprises computing a vignette score for each still image in a given set of still images.

4. The method of claim 1, wherein the step of computing a vignette score for each image in a given set of images comprises computing a vignette score for each video frame in a given set of video frames.

5. The method of claim 4, wherein the method is applied to video frames comprised in a video database.

6. The method of claim 1, wherein the step of computing a vignette score for each image in a given set of images comprises computing a vignette score for each motion picture frame in a given set of motion picture frames.

7. The method of claim 1, comprising the further step of:

c) selecting one or more images as one or more most representative images based on each image'"'"'s numerical rank.

8. The method of claim 7, wherein the step of computing a vignette score for each image in a given set of images comprises computing a vignette score for each video frame in a given set of video frames.

9. The method of claim 7, wherein the step of computing a vignette score for each image in a given set of images comprises computing a vignette score for each motion picture frame in a given set of motion picture frames.

10. The method of claim 1, wherein the step of computing a vignette score for each image in a given set of images comprises computing a vignette score for each image obtained in an Internet search in a given set of images, wherein each image in the set of images is one of a) obtained in an Internet search;

and b) selected from a collection of stock images.

11. The method of claim 1, wherein the method is applied to images comprised in one of a media organizer and a video database.

12. The method of claim 1, wherein the vignette score for each image is computed by determining the ratio of variance between the edge region of the image and the entire image.

13. The method of claim 1, wherein the vignette score for each image is computed by calculating a weighted ratio of variance of the image from more central regions to outlying regions, wherein the weighting is done based on the distance from the boundary.

14. The method of claim 1, wherein the vignette score for each image is computed by using a statistical classifier capable of determining a statistical model of each image class based on pre-entered training data consisting of images defined as vignettes and images defined as non-vignettes.

15. The method of claim 14, wherein the statistical classifier is a Gaussian mixture model.

16. The method of claim 14, wherein the statistical classifier is a Support Vector Machine.

17. The method of claim 1, wherein the vignette score for each image is computed by performing a spatial frequency analysis and calculating the ratio of mid-frequency energies in an edge region to the mid-frequencies of the entire image.

18. The method of claim 17, wherein the spatial frequency analysis consists of direct computation using Joint Photographic Experts Group (JPEG) coefficients.

19. The method of claim 17, wherein the spatial frequency analysis performed is a linear discrete transform.

20. The method of claim 19, wherein the linear discrete transform is one of a discrete cosine transform (DCT), a discrete Fourier transform, a Hadamard transform, and a wavelet transform.

21. A method for detecting useful images and for ranking images in order of usefulness, comprising the steps of:

a) computing a vignette score for each image in a given set of images, wherein a vignette score describes how closely an image resembles a central object surrounded by a featureless or deemphasized background;

b) ranking the images with numerical ranks according to each image'"'"'s vignette score; and

c) selecting one or more images as one or more most representative images based on each image'"'"'s numerical rank.

22. The method of claim 21, comprising the further step of:

d) creating an icon comprising the one or more most representative images.

23. A system for detecting useful images and for ranking images in order of usefulness, comprising;

a) one or more processors capable of computing a vignette score for each image in a given set of images, wherein the vignette score describes how closely an image resembles a central object surrounded by a featureless or deemphasized background and is determined using a method selected from the group consisting of variance ratio analysis, Guassian mixture model analysis, linear classifiers, support vector machines, discrete cosine transform, discrete fourier transform, Hadamard transform and wavelet transform, the one or more processors being further capable of ranking the images numerically according to each image'"'"'s vignette score; and

b) a machine readable medium including operations stored thereon that when processed by the one or more processors cause a system to perform the steps of computing the vignette score for each image in the given set of images and ranking the images numerically according to each image'"'"'s vignette score.

24. A method for detecting useful images and for ranking images in order of usefulness, comprising the steps of:

c) allowing a user to select a set of one or more images to be ranked;

d) computing a vignette score for each image in the set of images, wherein a vignette score describes how closely an image resembles a central object surrounded by a featureless or deemphasized background; and

e) ranking the images with numerical ranks according to each image'"'"'s vignette score.

25. A machine readable medium including operations stored thereon that when processed by one or more processors cause a system to perform the steps of:

a) computing a vignette score for each image in a given set of images, wherein a vignette score describes how closely an image resembles a central object surrounded by a featureless or deemphasized background; and

b) ranking the images with numerical ranks according to each image'"'"'s vignette score.

26. A system apparatus for detecting useful images for ranking images in order of usefulness, comprising:

means for computing a vignette score for each image in a given set of images, wherein a vignette score describes how closely an image resembles a central object surrounded by a featureless or deemphasized background; and

means for ranking the images numerically according to each image'"'"'s vignette score.