System and method for comparison-based image quality assessment

• US 10,410,330 B2
• Filed: 11/11/2016
• Issued: 09/10/2019
• Est. Priority Date: 11/12/2015
• Status: Active Grant

• Current Assignee: University of Virginia Patent Foundation (University of Virginia)
• Original Assignee: University of Virginia Patent Foundation (University of Virginia)
• Inventors: Liang, Haoyi, Weller, Daniel Stuart
• Primary Examiner(s): Potts, Ryan P
• Application Number:
  US15/349,971
• Publication Number:
  US 20170140518A1
• Time in Patent Office: 1,033 Days
• Field of Search: None
• CPC Class Codes: G06T 2207/20056 Discrete and fast Fourier transform, [DFT, FFT]
• CPC Class Codes: G06T 2207/30168 Image quality inspection
• CPC Class Codes: G06T 5/00 Image enhancement or restoration
• CPC Class Codes: G06T 5/50 using two or more images, e.g. averaging or subtraction
• CPC Class Codes: G06T 7/0002 Inspection of images, e.g. flaw detection
• CPC Class Codes: H04L 67/10 in which an application is distributed across nodes in the network software deployment G06F8/60; multiprogramming arrangements G06F9/46

Abstract

An image quality assessment and restoration system may include a processor and a memory storing instructions to receive an input image, receive a predetermined number of parameter candidates, generate a reconstructed image from the input image for each parameter candidate, sort the reconstructed images by the overall comparative quality between them and determine the best reconstructed image, calculate the overall comparative qualities between the remaining reconstructed images and the best reconstructed image, eliminate any parameter candidates that are suboptimal based on the calculated overall comparative quality, iteratively generate and sort additional reconstructed images and eliminate suboptimal parameter candidates until each of the remaining parameter candidates is converged, and output the converged parameters for use in image restoration. The overall comparative quality may depend upon the local gradient-based structure information and/or the global texture quality information.

Claims

1. A system for automatically assessing and comparing quality of digital images, comprising:

a memory storing instructions; and

a processor configured to execute the instructions to;

receive an input image;

receive a plurality of base images;

select at least one key image from the plurality of base images based on a minimum mean squared difference of each base image as compared to a predetermined threshold;

calculate overall comparative qualities between pairs of the input image and the at least one key image, the overall comparative qualities representing a weighted combination of comparative quality indices;

sort the input image and at least one key image by the overall comparative qualities between the pairs and determine the best key image, the best key image being a key image sorted highest by the overall comparative qualities; and

output at least one of;

the best key image as sorted by overall comparative quality or the overall comparative qualities,

wherein overall comparative quality depends upon local gradient-based structure information, and the local gradient-based structure information depends upon differences between images being compared.

3. The system of claim 1 or 2, further comprising:

a plurality of processors; and

a plurality of memories,

wherein the plurality of processors and the plurality of memories are adapted to execute the instructions in a cloud computing environment.

4. The system of claim 1 or 2, further comprising:

a network interface device,

wherein the interface device receives, at least in part, the instructions.

5. The system of claim 4, wherein the interface device transmits, at least in part, the output.

6. The system of claim 1 or 2, wherein the memory includes a static memory and a main memory.

7. The system of claim 1 or 2, further comprising at least one of:

a display unit;

an alphanumeric input device; and

a user interface navigation device.

8. The system of claim 1 or 2, wherein the overall comparative quality further depends upon global texture quality information, and the global texture quality information depends upon distributions of natural scene statistics.

9. The system of claim 1, wherein the plurality of base images comprise a plurality of reconstructed images generated, at least in part, from partial Fourier transforms of the input image.

2. A system for automatically assessing and restoring quality of a digital image, comprising:

a memory storing instructions; and

a processor configured to execute the instructions to;

receive an input image;

receive a predetermined number of parameter candidates;

generate a reconstructed image from the input image for each parameter candidate;

sort the reconstructed images by overall comparative qualities between pairs of the reconstructed images in the sorting, the overall comparative qualities representing a weighted combination of comparative quality indices, and determine the best reconstructed image, the best reconstructed image being a reconstructed image sorted highest by the overall comparative qualities;

calculate the overall comparative qualities between each remaining reconstructed image and the best reconstructed image;

eliminate any parameter candidates that are suboptimal based on the calculated overall comparative quality;

iteratively generate and sort additional reconstructed images and eliminate suboptimal parameter candidates until each of the remaining parameter candidates is converged; and

output at least one of;

the converged parameter candidates for use in image restoration or one or more reconstructed images whose parameter candidates converged,

wherein overall comparative quality depends upon local gradient-based structure information, and the local gradient-based structure information depends upon differences between images being compared.

10. The system of claim 2, wherein the reconstructed images are generated, at least in part, from partial Fourier transforms of the input image.

11. The system of claim 2, wherein the instructions to eliminate any parameter candidates that are suboptimal comprise instructions to:

calculate a quality prediction index for each of the parameter candidates;

determine whether each quality prediction index remains positive or negative over a predetermined number of iterations; and

eliminate any parameter candidates whose quality prediction index remains positive or negative over the predetermined number of iterations.

12. The system of claim 2, wherein the instructions to eliminate any parameter candidates that are suboptimal comprise instructions to:

calculate a quality prediction index for each of the parameter candidates;

calculate a global texture quality prediction for each of the parameter candidates;

determine whether each quality prediction index remains positive or negative over a predetermined number of iterations;

determine whether each global texture quality prediction remains positive or negative over the predetermined number of iterations; and

eliminate any parameter candidates whose quality prediction index and global texture quality prediction both remain positive or negative over the predetermined number of iterations.

13. The system of claim 2, wherein the weighted combination uses weights based on a smoothness index of total variances between patches of images being compared.

14. A method for automatically assessing and restoring quality of a digital image, comprising:

receiving an input image;

receiving a predetermined number of parameter candidates;

generating a reconstructed image from the input image for each parameter candidate;

sorting the reconstructed images by overall comparative qualities between pairs of the reconstructed images in the sorting, the overall comparative qualities representing a weighted combination of comparative quality indices, and determining the best reconstructed image, the best reconstructed image being a reconstructed image sorted highest by the overall comparative qualities;

calculating the overall comparative qualities between each remaining reconstructed image and the best reconstructed image;

eliminating any parameter candidates that are suboptimal based on the calculated overall comparative quality;

iteratively generating and sorting additional reconstructed images and eliminate suboptimal parameter candidates until each of the remaining parameter candidates is converged; and

outputting at least one of;

the converged parameter candidates for use in image restoration or one or more reconstructed images whose parameter candidates converged,

wherein overall comparative quality depends upon local gradient-based structure information, and the local gradient-based structure information depends upon differences between images being compared.

15. The method of claim 14, wherein the converged parameters are output using a communications network.

16. The method of claim 14, wherein the reconstructed images are generated, at least in part, from partial Fourier transforms of the input image.

17. The method of claim 14, wherein eliminating any parameter candidates that are suboptimal comprises:

calculating a quality prediction index for each of the parameter candidates;

determining whether each quality prediction index remains positive or negative over a predetermined number of iterations; and

eliminating any parameter candidates whose quality prediction index remains positive or negative over the predetermined number of iterations.

18. The method of claim 14, wherein eliminating any parameter candidates that are suboptimal comprises:

calculating a quality prediction index for each of the parameter candidates;

calculating a global texture quality prediction for each of the parameter candidates;

determining whether each quality prediction index remains positive or negative over a predetermined number of iterations;

determining whether each global texture quality prediction remains positive or negative over the predetermined number of iterations; and

eliminating any parameter candidates whose quality prediction index and global texture quality prediction both remain positive or negative over the predetermined number of iterations.

19. The method of claim 14, wherein determining the best reconstructed image comprises:

selecting a plurality of key images from the reconstructed images and the additional reconstructed images based on a minimum mean squared difference of each reconstructed image as compared to a predetermined threshold; and

selecting the best key image from the plurality of key images using a sorting algorithm, the best key image being the key image sorted highest by the overall comparative qualities,

wherein sorting the reconstructed images by overall comparative qualities comprises comparing overall comparative qualities between each reconstructed image and at least one of the plurality of key images.

20. The method of claim 14, wherein the overall comparative quality further depends upon global texture quality information, and the global texture quality information depends upon distributions of natural scene statistics between images being compared.