Blur detection system

• US 20040120598A1
• Filed: 12/18/2002
• Published: 06/24/2004
• Est. Priority Date: 12/18/2002
• Status: Active Grant

• Current Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Original Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Inventors: Feng, Xiao-Fan
• Granted Patent: US 7,181,082 B2
• Time in Patent Office: Days
• US Class Current: 382/263
• CPC Class Codes: G06T 2207/20008 Globally adaptive
• CPC Class Codes: G06T 2207/20021 Dividing image into blocks, subimages or windows
• CPC Class Codes: G06T 2207/20052 Discrete cosine transform [DCT]
• CPC Class Codes: G06T 5/10 using non-spatial domain filtering
• CPC Class Codes: G06T 5/73 Deblurring; Sharpening

Abstract

A system of estimating and reducing image blur in digital images.

Claims

1. A method of detecting and correcting blur in a digital image stored in a digital file as a sequence of DCT coefficients arranged in a plurality of blocks representing a grid of frequency components, each said block representing a portion of said image, said method comprising:

(a) reading a plurality of DCT coefficients from said sequence of DCT coefficients, said plurality of DCT coefficients comprising at least two DCT coefficients from each of said plurality of blocks;

(b) calculating from said plurality of DCT coefficients a first blur indicator that quantifies the amount of blur in the block having the least amount of blur among all of said plurality of blocks;

(c) calculating from said plurality of DCT coefficients a second blur indicator that quantifies the amount of blur throughout said digital image; and

(d) selectively applying a filter to said digital image based upon the value of said first blur indicator and said second blur indicator.

2. The method of claim 1 where the filter applied to said digital image is one of a plurality of differing types of filters.

3. The method of claim 1 including the step of selecting a representative frequency component in said grid and where said first blur indicator is the maximum valued DCT coefficient at said frequency component over all of said blocks in said image.

4. The method of claim 1 including the step of selecting two representative frequency components in said grid, calculating the standard deviation of said DCT coefficients over said image at said two representative frequency components, and where said second blur indicator is the ratio of said standard deviations.

5. The method of claim 1 including the steps of:

(a) dividing said grid of frequency components into a plurality of regions of increasing frequency;

(b) summarizing the DCT coefficients in each of said plurality of regions for each of said blocks;

(c) calculating said first blur indicator and said second blur indicator from said summary.

6. The method of claim 5 where said step of summarizing the DCT coefficients in each of said plurality of regions comprises obtaining an average value of the DCT coefficients in each of said plurality of regions.

7. The method of claim 6 where said first blur indicator is the maximum average value at a selected one of said plurality of regions.

8. The method of claim 6 including the step of obtaining the standard deviation of DCT coefficients at each of a selected two of said plurality of regions.

9. The method of claim 8 where said second blur indicator is the ration of said standard deviations.

10. A method of detecting and correcting blur in a digital image stored in a digital file as a sequence of DCT coefficients arranged in a plurality of blocks representing a grid of horizontal and vertical frequency components, each said block representing a portion of said image, said method comprising:

(a) obtaining a two-dimensional histogram of said DCT coefficients;

(b) summarizing said two-dimensional histogram with a one-dimensional histogram;

(c) calculating at least one blur indicator from said one-dimensional histogram; and

(d) selectively applying a filter to said digital image based upon said at least one blur indicator.

11. The method of claim 10 where said one-dimensional histogram is obtained by dividing said two-dimensional histogram into a plurality of regions of generally increasing frequency and averaging the DCT coefficients within each region for each of said blocks.

12. The method of claim 11 where said blur indicator is the maximum average DCT coefficient within a selective one of said regions.

13. The method of claim 11 where said blur indicator is the ratio of a average DCT values of a selective two of said regions.

14. The method of claim 11 including calculating a first blur indicator and a second blur indicator.

15. The method of claim 11 including a first blur indicator and a second blur indicator.

16. The method of claim 15 where said first blur indicator is the maximum average DCT coefficient within a selective one of said regions and said second blur indicator is the ratio of a average DCT values of a selective two of said regions.

17. The method of claim 16 where said filter is selected from a plurality of different types of filters based upon said first blur indicator and said second blur indicator.