Methods and systems for identifying regions of substantially uniform color in a digital image

• US 8,437,054 B2
• Filed: 06/15/2006
• Issued: 05/07/2013
• Est. Priority Date: 06/15/2006
• Status: Expired due to Fees

• Current Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Original Assignee: Sharp Laboratories of America Incorporated (Hon Hai Precision Industry Co., Ltd.)
• Inventors: Campbell, Richard John, Chen, Lawrence Shao-hsien, Matsuda, Toyohisa
• Primary Examiner(s): SARPONG, AKWASI
• Application Number:
  US11/424,290
• Publication Number:
  US 20070291120A1
• Time in Patent Office: 2,518 Days
• Field of Search: 358/522, 358/515, 358/1.9, 358/462.02, 358/538, 358/539, 358/453, 382164-167, 382173-180, 382/118, 382/149, 348/616, 348/587, 345/605
• US Class Current: 358/522
• CPC Class Codes: G06T 2207/10008 from scanner, fax or copier
• CPC Class Codes: G06T 2207/20152 Watershed segmentation
• CPC Class Codes: G06T 2207/30176 Document
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/155 involving morphological operators
• CPC Class Codes: G06T 7/90 Determination of colour characteristics
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/15 Cutting or merging image elements, e.g. region growing, watershed or clustering-based techniques
• CPC Class Codes: G06V 30/18105 related to colour

Abstract

Embodiments of the present invention comprise methods and systems for identifying uniformly colored regions in a digital image.

Claims

1. A method for identifying a region of substantially uniform color in a digital image, said method comprising:

in a computing system, determining a first plurality of select pixels surrounding an image-content edge in a digital image, wherein said determining a first plurality of select pixels comprises filtering said digital image, based on image classification information, to determine a plurality of text pixels located in a non-pictorial region of said digital image and, wherein said image classification information is a pictorial measure and an identification of a text candidate and said filtering comprises removing a text candidate when said text candidate is in a pictorial region of said digital image, wherein said text candidate comprises a pixel that is a portion of text;

for said first plurality of select pixels, determining, in said computing system, a frequency-of-occurrence of a plurality of color values;

in said computing system, identifying at least one color value associated with a peak frequency-of-occurrence; and

in said computing system, associating said at least one color value with an image label, thereby producing an association between said at least one color value and said image label.

2. The method of claim 1 further comprising generating a look-up-table wherein said look-up-table comprises said association between said at least one color value and said image label, wherein said image label is associated with an image region.

3. The method of claim 1 wherein said determining a frequency-of-occurrence further comprises generating a histogram.

4. The method of claim 3 further comprising smoothing said histogram.

5. The method of claim 1 wherein said first plurality of select pixels does not include any pixel that is substantially distant from a text edge in said digital image.

6. The method of claim 1 wherein said image-content edge is a text edge.

7. The method of claim 1 wherein said first plurality of select pixels is selected by generating a selection mask.

8. The method of claim 7 wherein said generating a selection mask comprises:

dilating said plurality of text pixels thereby producing a text dilation; and

removing said plurality of text pixels from said text dilation.

9. The method of claim 1 wherein said image label indicates a local background region.

10. The method of claim 1 wherein said image label indicates a page background region.

11. The method of claim 1 further comprising:

identifying at least one color value associated with a frequency-of-occurrence below a significance threshold; and

associating said at least one color value associated with said frequency-of-occurrence below said significance threshold with a reserved image label, thereby producing an association between said at least one color value associated with said frequency-of-occurrence below said significance threshold and said reserved image label.

12. A system for identifying a region of substantially uniform color in a digital image, said system comprising:

a pixel selector for selecting a first plurality of pixels in a digital image, wherein said first plurality of pixels surround an image-content edge in said digital image;

a filter for combining a pictorial measure and a plurality of text candidates to determine a plurality of text pixels located in a non-pictorial region of said digital image and, wherein said filter removes a first text candidate from said plurality of text candidates when said first text candidate appears in a pictorial region of said digital image, wherein said first text candidate comprises a pixel that is a portion of text;

a counter for counting a frequency-of-occurrence of a plurality of color values in said first plurality of pixels;

an identifier for identifying at least one color value associated with a peak frequency-of-occurrence; and

an associator for associating said at least one color value with an image label, thereby producing an association between said at least one color value and said image label.

13. The system of claim 12 further comprising a generator for generating a look-up-table wherein said look-up-table comprises said association between said at least one color value and said image label, wherein said image label is associated with an image region.

14. The system of claim 12, wherein said pixel selector comprises a pixel excluder that excludes any pixel in said first plurality of pixels that is substantially distant from a text edge in said digital image.

15. The system of claim 12 wherein said image-content edge is a text edge.

16. The system of claim 12 wherein said plurality of color values comprises color values in a luminance-chrominance color space.

17. The system of claim 12 wherein said pixel selector comprises a selection-mask generator for generating a selection mask.

18. The system of claim 17 wherein said selection-mask generator comprises:

a dilator for dilating said plurality of text pixels thereby producing text dilations; and

an excluder for excluding said plurality of text pixels from said text dilations.

19. The system of claim 12 wherein said image label indicates a local background region.

20. The system of claim 12 further comprising:

a low frequency-of-occurrence identifier for identifying at least one color value associated with a low frequency-of-occurrence; and

a reserved label associator for associating said at least one color value associated with a low frequency-of-occurrence with a reserved image label, thereby producing an association between said at least one color value associated with a low frequency-of-occurrence and said reserved image label.

21. A method for identifying a region of substantially uniform color in a digital image, said method comprising:

in a computing system, detecting a text pixel in a non-pictorial region in a digital image comprising pixels, wherein said detecting a text pixel comprises filtering said digital image based on image classification information, wherein said image classification information is a pictorial measure and an identification of a text candidate and said filtering comprises removing a text candidate when said text candidate is in a pictorial region of said digital image, wherein said text candidate comprises a pixel that is a portion of text;

in said computing system, generating a selection mask comprising selection-mask pixels, wherein said selection-mask pixels comprise a first pixel in a region substantially local to said text pixel, and wherein said selection-mask pixels do not include said text pixel;

in said computing system, generating a histogram of a plurality of color values, wherein said determining comprises counting said selection-mask pixels;

in said computing system, smoothing said histogram;

in said computing system, identifying at least one color value associated with a peak in said histogram;

in said computing system, associating said at least one color value with an image label, thereby producing an association between said at least one color value and said image label; and

in said computing system, generating a look-up-table wherein said look-up-table comprises said association between said at least one color value and said image label.

22. The method of claim 21 wherein said generating a selection mask comprises:

identifying a text edge in a non-pictorial region of said digital image;

dilating said text edge thereby producing a text dilation; and

removing said text edge from said text dilation.