Methods and systems for detecting regions in digital images

• US 7,792,359 B2
• Filed: 06/15/2006
• Issued: 09/07/2010
• Est. Priority Date: 03/02/2006
• 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: Matsuda, Toyohisa, Campbell, Richard John
• Primary Examiner(s): Ahmed; Samir A.
• Assistant Examiner(s): Li; Ruiping
• Application Number:
  US11/424,281
• Publication Number:
  US 20070206856A1
• Time in Patent Office: 1,545 Days
• Field of Search: 382/176, 382/118, 382/110, 378/98, 358/462
• US Class Current: 382/168
• CPC Class Codes: G06V 10/507 Summing image-intensity values; Histogram projection analysis
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables

Abstract

Embodiments of the present invention comprise systems, methods and devices for detection of image regions using a masking condition and an entropy measure.

Claims

1. A method for detecting a region in a digital image, said method comprising:

a) in a region-detection system, applying a masking condition to a digital image thereby identifying masked pixels that satisfy said masking condition;

b) in a feature extractor in said region-detection system, calculating an entropy feature associated with a pixel-of-interest in said digital image, wherein said calculating said entropy feature comprises;

i) quantizing, according to a first quantization, a plurality of pixel values associated with a plurality of pixels in a region surrounding said pixel-of-interest, wherein said plurality of pixels in said region surrounding said pixel-of-interest does not include said masked pixels;

ii) forming a first histogram of said first-quantization quantized pixel values;

iii) calculating a first entropy measure for said first histogram;

iv) quantizing, according to a second quantization, said plurality of pixel values associated with said plurality of pixels in said region surrounding said pixel-of-interest;

v) forming a second histogram of said second-quantization quantized pixel values; and

vi) calculating a second entropy measure for said second histogram; and

c) classifying a neighborhood of said digital image based on a combination of said first entropy measure and said second entropy measure and said pixel-of-interest location.

2. A method as described in claim 1, wherein said first quantization is related to the amount of noise associated with a uniformly colored area in said digital image.

3. A method as described in claim 1 wherein said first histogram bin size comprises a plurality of pixel values.

4. A method as described in claim 1, wherein said first histogram range is smaller than said dynamic range of said first-quantization quantized pixel values.

5. A method as described in claim 1, wherein said calculating said first entropy feature further comprises:

a) selecting a first lobe of said first histogram; and

b) calculating said first entropy measure using only said first lobe of said first histogram.

6. A method as described in claim 1 wherein said first histogram bin size is not the same as said second histogram bin size.

7. A method as described in claim 1 wherein said first histogram bins are not located at the same locations as said second histogram bins.

8. A method for detecting a region in a digital image, said method comprising:

a) in are region-detection system, applying a masking condition to a digital image thereby identifying masked pixels that satisfy said masking condition;

b) in a feature extractor in said region-detection system, calculating an entropy feature associated with a pixel-of-interest in said digital image, wherein said calculating said entropy feature comprises;

i) quantizing, according to a first quantization, a plurality of pixel values associated with a plurality of pixels in a region surrounding said pixel-of-interest, wherein said plurality of pixels in said region surrounding said pixel-of-interest does not include said masked pixels;

ii) forming a first histogram of said first-quantization quantized pixel values; and

iii) calculating an entropy measure for said first histogram, wherein said calculating said entropy measure for said first histogram comprises;

(1) selecting a first lobe of said first histogram; and

(2) calculating said entropy measure using only said first lobe of said first histogram; and

c) classifying a neighborhood of said digital image based on said entropy measure and said pixel-of-interest location.

9. A method as described in claim 8 wherein said selecting comprises identifying a lobe containing said pixel value of said pixel-of-interest.

10. A system for detecting a region in a digital image, said system comprising:

a) a processor for applying a masking condition to a digital image thereby identifying masked pixels that satisfy said masking condition;

b) a first quantizer for quantizing, according to a first quantization, a plurality of pixel values associated with a plurality of pixels in a region surrounding a pixel-of-interest, wherein said plurality of pixels in said region surrounding said pixel-of-interest does not include said masked pixels;

c) a second quantizer for quantizing, according to a second quantization, said plurality of pixel values associated with said plurality of pixels in said region surrounding said pixel-of-interest;

d) an entropy-feature calculator for calculating an entropy feature associated with said pixel-of-interest in said digital image, said entropy-feature calculator comprising;

i) a first histogram generator for forming a first histogram of said first-quantization quantized pixel values;

ii) a first entropy calculator for calculating a first entropy measure for said first histogram;

iii) a second histogram generator for forming a second histogram of said second-quantization quantized pixel values;

iv) a second entropy calculator for calculating a second entropy measure for said second histogram; and

v) a combiner for combining said first entropy measure and said second entropy measure to form a combined entropy measure; and

e) a classifier for classifying a neighborhood of said digital image based on said combined entropy measure and said pixel-of-interest location.

11. A system as described in claim 10, wherein said first quantization is related to the amount of noise associated with a uniformly colored area in said digital image.

12. A system as described in claim 10 wherein said first histogram bin size comprises a plurality of pixel values.

13. A system as described in claim 10, wherein said first histogram range is smaller than said dynamic range of said first-quantization quantized pixel values.

14. A system as described in claim 10, wherein said first entropy calculator further comprises:

a) a selector for selecting a first lobe of said first histogram; and

b) a lobe-based calculator for calculating said entropy measure using only said first lobe of said first histogram.

15. A system as described in claim 10 wherein said first histogram bin size is not the same as said second histogram bin size.

16. A system as described in claim 10 wherein said first histogram bins are not located at the same locations as said second histogram bins.

17. A system for detecting a region in a digital image, said system comprising:

a) a processor for applying a masking condition to a digital image thereby identifying masked pixels that satisfy said masking condition;

b) a first quantizer for quantizing, according to a first quantization, a plurality of pixel values associated with a plurality of pixels in a region surrounding a pixel-of-interest, wherein said plurality of pixels in said region surrounding said pixel-of-interest does not include said masked pixels;

c) an entropy-feature calculator for calculating an entropy feature associated with said pixel-of-interest in said digital image, said entropy-feature calculator comprising;

i) a first histogram generator for forming a first histogram of said first-quantization quantized pixel values; and

ii) a first entropy calculator for calculating an entropy measure for said first histogram, wherein said first entropy calculator comprises;

(1) a selector for selecting a first lobe of said first histogram; and

(2) a lobe-based calculator for calculating said entropy measure using only said first lobe of said first histogram; and

d) a classifier for classifying a neighborhood of said digital image based on said entropy measure and said pixel of interest.

18. A system as described in claim 17 wherein said selector comprises an identifier for identifying a lobe containing said pixel value of said pixel-of-interest.

19. A method for detecting a region in a digital image, said method comprising:

a) in a region-detection system, applying a masking condition to a digital image thereby identifying masked pixels that satisfy said masking condition;

b) in a feature extractor in said region-detection system, calculating an entropy feature associated with a pixel-of-interest in said digital image, wherein said calculating said entropy feature comprises;

i) forming a plurality of histograms, wherein each histogram in said plurality of histograms is associated with a different quantization, of pixel values for a plurality of pixels in a region surrounding said pixel-of-interest, wherein said masked pixels do not accumulate in said plurality of histograms;

ii) selecting a first histogram from said plurality of histograms;

iii) calculating an first entropy measure for said first histogram;

iv) selecting a second histogram from said plurality of histograms; and

v) calculating a second entropy measure for said second histogram; and

c) classifying a neighborhood of said digital image based on a combination of said first entropy measure and said second entropy measure and said pixel-of-interest location.

20. The method of claim 19 wherein said selecting comprises comparing the lobes of said plurality of histograms.