Pixel perturbation for image quality measurement

Patent metadata

Field	Value
Publication number	US-9842387-B2
Application number	US-201514743633-A
Country	US
Kind code	B2
Filing date	Jun 18, 2015
Priority date	Jun 18, 2015
Publication date	Dec 12, 2017
Grant date	Dec 12, 2017

Abstract

Techniques disclosed herein provide an approach using pixel perturbation to measure image quality. In one embodiment, a pixel perturbation engine perturbs pixels in an image with a reference image for measuring quality of the image after transmission. By perturbing least significant bits, a composite image may be generated in which the reference image is hidden in the original image. The perturbations in the composite image may then be recovered after the composite image is transmitted to a remote device and used to determine image quality based on preservation of the perturbations. In another embodiment, image(s) perturbed with reference image(s) at increasingly higher order bit positions may be transmitted, and quality of the transmitted reference image determined, until the determined quality exceeds a threshold, with the perturbed bit position at which the determined quality exceeds the threshold being indicative of the quality of the image(s) transmitted.

First claim

What is claimed is:

1. A computer-implemented method for measuring image quality, comprising:
receiving a composite image, wherein bits of one or more pixels of a first image are perturbed to generate the composite image in which a plurality of reference images are hidden in the first image via perturbations, and wherein a respective one of a plurality of bit positions is perturbed to hide each of the reference images;
recovering, from the composite image, the perturbations; and
determining a quality of the first image in the received composite image based on at least a lowest of the plurality of perturbed bit positions for which quality of a corresponding one of the hidden reference images meets a predefined threshold value, wherein quality of each of the hidden reference images is determined based on at least a comparison of recovered perturbations representing the hidden reference image and a corresponding stored reference image.

2. The method of claim 1 , wherein the hidden reference images include multiple distinct copies of the same reference image.

3. The method of claim 1 , wherein the comparison of the recovered perturbations representing the hidden reference image and the corresponding stored reference image is performed using at least one of structural similarity (SSIM), peak signal-to-noise ratio (PSNR), and mean squared error (MSE).

4. The method of claim 1 , wherein the perturbations are performed on tiles having a predefined tile size in the first image.

5. The method of claim 1 , wherein one or more color channels of the first image are perturbed.

6. The method of claim 1 , wherein the bits of the one or more pixels of the first image are perturbed to generate the composite image in which a regular pattern of the same reference image is hidden in the first image.

7. The method of claim 1 , wherein the first image is an RGB image of a virtual desktop user interface.

8. The method of claim 1 ,
wherein the first image is a YUV image converted from an RGB image; and
wherein the perturbations are recovered from the composite image which is a YUV image and used in determining the quality of the first image prior to a conversion of the composite image back to an RGB image.

9. The method of claim 1 , wherein the composite image is generated to include a black and white pattern hidden in the first image.

10. A non-transitory computer-readable storage medium containing a program which, when executed by one or more processors, performs operations for measuring image quality, the operations comprising:
receiving a composite image, wherein the composite image is generated by perturbing bits of one or more pixels in a corresponding user-perceptible image to hide in the user-perceptible image a corresponding reference image via perturbations, and wherein a first bit position is perturbed in the user-perceptible image;
recovering, from the composite image, the perturbations;
determining a quality of the hidden reference image based on at least a comparison of the recovered perturbations representing the hidden reference image and a stored copy of the reference image;
if the determined quality of the hidden reference image does not satisfy a predefined threshold value, repeating the receiving, recovering, and determining steps for one or more additional composite images in which one or more successively higher bit positions are perturbed in one or more corresponding user-perceptible images, until a determined hidden reference image quality satisfies the predefined threshold value; and
determining an indication of image quality based on at least a perturbed bit position for which the determined hidden reference image quality satisfies the predefined threshold value.

11. The computer-readable storage medium of claim 10 , wherein the composite images each include a hidden copy of the same reference image.

12. The computer-readable storage medium of claim 10 , wherein the comparison of the recovered perturbations representing the hidden reference image and the stored copy of the reference image is performed using at least one of structural similarity (SSIM), peak signal-to-noise ratio (PSNR), and mean squared error (MSE).

13. The computer-readable storage medium of claim 10 , wherein the perturbations are performed on tiles having a predefined tile size.

14. The computer-readable storage medium of claim 10 , wherein one or more color channels are perturbed.

15. The computer-readable storage medium of claim 10 , wherein each of the composite images includes a regular pattern of the same reference image hidden in the corresponding user-perceptible image.

16. The computer-readable storage medium of claim 10 , wherein each of the user-perceptible images is an RGB image of a virtual desktop user interface.

17. The computer-readable storage medium of claim 10 ,
wherein each of the user-perceptible images is a YUV image converted from a corresponding RGB image; and
wherein, for each of the composite images, perturbations are recovered from the composite image which is a YUV image and used in determining image quality prior to a conversion of the composite image back to an RGB image.

18. A system, comprising:
a processor; and
a memory, wherein the memory includes a program executable in the processor to perform operations for measuring image quality, the operations comprising:
receiving a composite image, wherein bits of one or more pixels of a first image are perturbed to generate the composite image in which a plurality of reference images are hidden in the first image via perturbations, and wherein a respective one of a plurality of bit positions is perturbed to hide each of the reference images,
recovering, from the composite image, the perturbations, and
determining a quality of the first image in the received composite image based on at least a lowest of the plurality of perturbed bit positions for which quality of a corresponding one of the hidden reference images meets a predefined threshold value, wherein quality of each of the hidden reference images is determined based on at least a comparison of recovered perturbations representing the hidden reference image and a corresponding stored reference image.

19. A system, comprising:
a processor; and
a memory, wherein the memory includes a program executable in the processor to perform operations for measuring image quality, the operations comprising:
receiving a composite image, wherein the composite image is generated by perturbing bits of one or more pixels in a corresponding user-perceptible image to hide in the user-perceptible image a corresponding reference image via perturbations, and wherein a first bit position is perturbed in the user-perceptible image,
recovering, from the composite image, the perturbations,
determining a quality of the hidden reference image based on at least a comparison of the recovered perturbations representing the hidden reference image and a stored copy of the reference image,
if the determined quality of the hidden reference image does not satisfy a predefined threshold value, repeating the receiving, recovering, and determining steps for one or more additional composite images in which one or more successively higher bit positions are perturbed in one or more corresponding user-perceptible images, until a determined hidden reference image quality satisfies the predefined threshold value, and
determining an indication of image quality based on at least a perturbed bit position for which the determined hidden reference image quality satis

Assignees

• Vmware Inc

Inventors

• Kurkure Uday
• Sivaraman Hari
• Farkas Peter
• He Xingze

Classifications

• G06T2207/10024

  Color image · CPC title

• G06T7/0002 Primary

  Inspection of images, e.g. flaw detection · CPC title

• G06T2207/30168

  Image quality inspection · CPC title

• G06T2207/20221

  Image fusion; Image merging · CPC title