Automatic image cropping

• US 5,978,519 A
• Filed: 08/06/1996
• Issued: 11/02/1999
• Est. Priority Date: 08/06/1996
• Status: Expired due to Term

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Rao, Ramana L., Eschbach, Reiner, Venable, Dennis L., Bollman, James E.
• Primary Examiner(s): COUSO, YON JUNG
• Application Number:
  US08/692,559
• Time in Patent Office: 1,183 Days
• Field of Search: 382/282, 382/283, 382/171, 382/172, 382/270, 382/272, 358/538
• US Class Current: 382/282
• CPC Class Codes: G06T 7/70 Determining position or orientation of objects or cameras camera calibration G06T7/80

Abstract

The present invention describes a method for automatic cropping of images containing regions where intensity levels are uniform and other regions where intensity levels vary considerably. An image to be automatically cropped is scaled down to a grid and divided into non-overlapping blocks. The mean and variance of intensity levels are calculated for each block. Based on the distribution of variances in the blocks, a threshold is selected for the variance. All blocks with a variance higher than this threshold variance are selected as regions of interest. The regions of interest are then cropped to a bounding rectangle.

Claims

1. A method for automatically cropping an image, comprising:

scaling down said image to a grid having four sides;

dividing said grid into a plurality of non-overlapping blocks;

calculating a mean intensity level for each of said blocks;

calculating a variance of an intensity level for each of said blocks;

creating a variance profile for said blocks;

calculating a threshold variance based on said variance profile;

selecting said blocks having said variance higher than said threshold variance as regions of interest;

cropping said regions of interest to a bounding rectangle;

calculating a profile of edge strengths for said blocks;

calculating a threshold edge strength from said profile; and

selecting said blocks having an edge strength higher than said threshold edge strength and not selected as regions of interest.

2. The automatic image cropping method of claim 1, wherein said intensity level is a luminance intensity.

3. The automatic image cropping method of claim 1, wherein said intensity level is a red, green, blue intensity.

4. The automatic image cropping method of claim 1, wherein said mean is calculated by the formula ##EQU3## wherein g_i is the intensity level of the i^th pixel in the block, and N is a height and a width of each block in pixels.

5. The automatic image cropping method of claim 4, wherein said variance is calculated by the formula ##EQU4##

6. The automatic image cropping method of claim 1, wherein said threshold voltage variance is selected as a point on a curved portion of a variance profile.

7. The automatic image cropping method of claim 6, wherein said threshold variance is adjusted by a selectivity factor.

8. The automatic image cropping method of claim 7, wherein said threshold variance is reduced to select more regions of interest.

9. The automatic image cropping method of claim 7, wherein said selectivity factor is about forty percent.

10. The automatic image cropping method of claim 1, wherein said grid is 256 by 256 pixels.

11. The automatic image cropping method of claim 1, wherein said non-overlapping blocks are 4 by 4 pixels.

12. The automatic image cropping method of claim 1, wherein said bounding rectangle is defined by a first selected block along each of the four sides of the grid.

13. The automatic image cropping method of claim 1, further comprising subjecting an autocropped image to a post-processing image operation.

14. The automatic image cropping method of claim 1, further comprising scaling said bounding rectangle to a larger or smaller dimension having a border.

15. The automatic image cropping method of claim 14, wherein said border is a default border of about 1% of said dimension.

16. The automatic image cropping method of claim 1, further comprising marking unselected blocks inside selected regions for further post-processing image operations.

17. The automatic image cropping method of claim 1, further comprising removing details corresponding to noise in the background of the cropped image to provide a better bounding rectangle.

18. An apparatus for automatically cropping an image, comprising:

means for scaling down said image to a grid having four sides;

means for dividing said grid into a plurality of non-overlapping blocks;

means for calculating a mean intensity level for each said blocks;

means for calculating a variance of an intensity level for each of said blocks;

means for creating a variance profile for said blocks;

means for calculating a threshold variance based on said variance profile;

means for selecting said blocks having said variance higher than said threshold variance as regions of interest;

means for cropping said regions of interest to a bounding rectangle;

means for calculating a profile of edge strengths for said blocks;

means for calculating a threshold edge strength from said profile; and

means for selecting said blocks having an edge strength higher than said threshold edge strength and not selected as regions of interest.

19. The apparatus of claim 18, further comprising an input means for acquiring said image and an output means for storing said autocropped image.

20. The apparatus of claim 19, wherein said input means is selected from the group consisting of a digital scanner and a digital camera.

21. The apparatus of claim 20, wherein said output means is selected from the group consisting of a printer, a LCD display, a CRT display, a magnetic media, and a photographic printer.