Method and system for determining a quality measure for an image using multi-level decomposition of images

• US 8,660,364 B2
• Filed: 07/09/2013
• Issued: 02/25/2014
• Est. Priority Date: 02/11/2009
• Status: Active Grant

• Current Assignee: Ecole de technologie superieure
• Original Assignee: Ecole de technologie superieure
• Inventors: Rezazadeh, Soroosh, Coulombe, Stephane
• Primary Examiner(s): Bayat, Ali
• Application Number:
  US13/938,181
• Publication Number:
  US 20130294703A1
• Time in Patent Office: 231 Days
• Field of Search: 382/199, 382/232, 382/233, 382/128, 382/203, 382/240, 382/243, 382/246, 382/260, 382/166, 375/240.18, 375/240.19, 375/240.8, 250/202, 250/363.06
• US Class Current: 382/199
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06T 2207/20064 Wavelet transform [DWT]
• CPC Class Codes: G06T 2207/30168 Image quality inspection
• CPC Class Codes: G06T 7/0002 Inspection of images, e.g. flaw detection

Abstract

Method and system for determining a measure of quality for images by using multi-level decomposition are presented. Multi-level decomposition of images is performed in the wavelet domain producing subbands at each level of decomposition. Aggregation of subbands is performed across multiple levels to produce an accurate measure of image quality. By aggregating only selected subbands the computational complexity of the method is greatly reduced.

Claims

1. A method for determining a measure of quality for an image Y, characterizing a similarity between the image Y and a reference image X having same resolution, the method comprising:

(a) applying an N level multiresolution decomposition, comprising levels 1, 2, . . . i, i+1, . . . N, to the image X and the image Y respectively, to produce;

for each level i, with i ranging from 1 to N−

1, intermediate subbands of the image X and the image Y for processing at level i+1; and

for the level N, approximation subbands containing respective main contents of the image X and the image Y, and respective detail subbands containing edges of the image X and the image Y;

(b) applying an image quality metric (IQM) to the approximation subband of the image X and the approximation subband of the image Y to produce an approximation quality measure characterizing similarity between the main content of the image X and the main content of the image Y;

(c) at at least one level i, with i ranging from levels 1 to N−

1, aggregating the intermediate subbands and the detail subbands, for the image X and the image Y respectively, to produce respective edge maps for the image X and the image Y, characterizing edges of the image X and the image Y respectively;

(d) applying the IQM to the edge maps of the image X and the image Y to produce an edge quality measure characterizing similarity between the edges of the image X and the image Y; and

(e) processing the approximation quality measure and the edge quality measure to determine the measure of quality.

2. The method of claim 1, wherein the step (c) further comprises:

selecting the intermediate subbands at said at least one level i, with i ranging from 1 to N−

1, and the detail subbands, for the image X and the image Y based on an accuracy to be achieved in determining the measure of quality; and

aggregating only selected intermediate subbands and selected detail subbands, for the image X and the image Y.

3. The method of claim 2, further comprising choosing the selected intermediate subbands at the level i and the selected detail subbands, having substantially same resolution, for aggregating.

4. The method of claim 1, applying weighting factors to the respective edge maps for the image X and the image Y, for aggregating.

5. The method of claim 2, further comprising selecting the intermediate subbands at the level i and the detail subbands for the image X and the image Y based on a number of the intermediate and the detailed subbands required for achieving the accuracy.

6. The method of claim 1, wherein the intermediate subbands at said at least one level i for the image X and the image Y include for each respective image:

level i-detail subbands and level i-wavelet packet (WP) subbands, the level i-WP subbands comprising level i-WP approximation subbands and level i-WP detail subbands.

7. The method of claim 1, further comprising aggregating one or more of the following:

horizontal intermediate subbands to produce a horizontal edge map;

vertical intermediate subbands to produce a vertical edge map;

or

diagonal intermediate subbands to produce a diagonal edge map.

8. The method of claim 1, wherein the step (a) comprises applying an N level discrete wavelet transform (DWT).

9. The method of claim 1, wherein the steps (b) and (d) comprise applying the IQM, which is one of the following:

a structural similarity (SSIM) IQM;

an Absolute Difference (AD) IQM;

a peak-signal-to-noise ratio (PSNR) IQM;

or

a Visual Information Fidelity (VIF) IQM.

10. The method of claim 1, further comprising determining N as a function of a minimum size of an approximation subband, which produces a substantially peak response for human visual system.

11. A system for determining a measure of quality for an image Y, characterizing a similarity between the image Y and a reference image X having same resolution, the system comprising:

a processor; and

a non-transitory computer readable storage medium having computer readable instructions stored thereon for execution by the processor, causing the processor to;

(a) apply an N level multiresolution decomposition, comprising levels 1, 2, . . . i, i+1, . . . N, to the image X and the image Y respectively, to produce;

for each level i, with i ranging from 1 to N−

1, intermediate subbands of the image X and the image Y for processing at level i+1; and

for the level N, approximation subbands containing respective main contents of the image X and the image Y, and respective detail subbands containing edges of the image X and the image Y;

(b) apply an image quality metric (IQM) to the approximation subband of the image X and the approximation subband of the image Y to produce an approximation quality measure characterizing similarity between the main content of the image X and the main content of the image Y;

(c) at at least one level i, with i ranging from levels 1 to N−

1, aggregate the intermediate subbands and the detail subbands, for the image X and the image Y respectively, to produce respective edge maps for the image X and the image Y, characterizing edges of the image X and the image Y respectively;

(d) apply the IQM to the edge maps of the image X and the image Y to produce an edge quality measure characterizing similarity between the edges of the image X and the image Y; and

(e) process the approximation quality measure and the edge quality measure to determine the measure of quality.

12. The system of claim 11, wherein the computer readable instructions further cause the processor to:

select the intermediate subbands at said at least one level i, with i ranging from 1 to N−

1, and the detail subbands, for the image X and the image Y based on an accuracy to be achieved in determining the measure of quality; and

aggregate only selected intermediate subbands and selected detail subbands, for the image X and the image Y.

13. The system of claim 12, wherein the computer readable instructions further cause the processor to choose the selected intermediate subbands at the level i and the selected detail subbands, having substantially same resolution, for aggregating.

14. The system of claim 12, wherein the computer readable instructions further cause the processor to apply weighting factors to the respective edge maps for the image X and the image Y, for aggregating.

15. The system of claim 12, wherein the computer readable instructions further cause the processor to select the intermediate subbands at the level i and the detail subbands for the image X and the image Y based on a number of the intermediate and the detailed subbands required for achieving the accuracy.

16. The system of claim 11, wherein the intermediate subbands at said at least one level i for the image X and the image Y include for each respective image:

level i-detail subbands and level i-wavelet packet (WP) subbands, the level i-WP subbands comprising level i-WP approximation subbands and level i-WP detail subbands.

17. The method of claim 11, wherein the computer readable instructions further cause the processor to aggregate one or more of the following:

horizontal intermediate subbands to produce a horizontal edge map;

vertical intermediate subbands to produce a vertical edge map;

or

diagonal intermediate subbands to produce a diagonal edge map.

18. The system of claim 11, wherein the N level multiresolution decomposition is an N level discrete wavelet transform (DWT).

19. The system of claim 11, wherein the computer readable instructions further cause the processor to apply the IQM, which is one of the following:

a structural similarity (SSIM) IQM;

an Absolute Difference (AD) IQM;

a peak-signal-to-noise ratio (PSNR) IQM;

or

a Visual Information Fidelity (VIF) IQM.

20. The system of claim 11, wherein the computer readable instructions further cause the processor to determine N as a function of a minimum size of an approximation subband, which produces a substantially peak response for human visual system.