System and method for image upsampling using natural image statistics of first and second derivatives

• US 8,369,653 B1
• Filed: 02/27/2009
• Issued: 02/05/2013
• Est. Priority Date: 02/27/2009
• Status: Active Grant

• Current Assignee: Adobe Inc.
• Original Assignee: Adobe Systems Incorporated (Adobe Inc.)
• Inventors: Cohen, Scott D.
• Primary Examiner(s): Park, Chan S
• Assistant Examiner(s): PERLMAN, DAVID S
• Application Number:
  US12/395,316
• Time in Patent Office: 1,439 Days
• Field of Search: None
• US Class Current: 382/299
• CPC Class Codes: G06T 3/40 Scaling of whole images or parts thereof, e.g. expanding or contracting

Abstract

Systems and methods for upsampling input images may evaluate potential upsampling solutions with respect to an objective function that is dependent on a sparse derivative prior on second derivative(s) of the potential upsampling solutions to identify an acceptable higher-resolution output image. The objective function may also be dependent on fidelity term(s) and/or sparse derivative prior(s) on first derivative(s) of potential upsampling solutions. The methods may include applying the iteratively re-weighted least squares procedure in minimizing the objective function and generating improved candidate solutions from an initial solution. The identified solution may be stored as a higher-resolution version of the input image in memory, and made available to subsequent operations in an image editing application or other graphics application. The methods may produce sharp results that are also smooth along edges. The methods may be implemented as program instructions stored on computer-readable storage media, executable by a CPU and/or GPU.

Claims

1. A system, comprising:

one or more processors; and

a memory coupled to the one or more processors and storing program instructions executable by the one or more processors to implement;

accessing data representing an input image to be upsampled;

for each of a plurality of potential upsampling solutions;

evaluating the potential upsampling solution with respect to an objective function, wherein the objective function is dependent on a sparse derivative prior in which sparse derivative prior probabilities for one or more second derivatives of the potential upsampling solution are multiplied by sparse derivative prior probabilities for one or more first derivatives of the potential upsampling solution;

identifying one of the plurality of potential upsampling solutions as an acceptable solution, dependent on said evaluating;

storing the identified solution as a higher-resolution version of the input image.

2. The system of claim 1, wherein the objective function is dependent on a sparse derivative prior on two or more second derivatives of the potential upsampling solution, and wherein one of the two or more second derivatives of the potential upsampling solution comprises a derivative in a given direction on a result of a first derivative of the potential upsampling solution in a different direction.

3. The system of claim 1, wherein the objective function is dependent on a sparse derivative prior on a derivative of the potential upsampling solution in a direction other than horizontal or vertical.

4. The system of claim 1, wherein the program instructions are further executable by the one or more processors to implement an interface through which one or more of an upsampling factor, a target output image resolution, a target relative resolution, a number of derivatives, an order of a derivative, a direction of a derivative, or a number of solution iterations is configurable by a user.

5. The system of claim 1, wherein the objective function is further dependent on a fidelity term configured to bias said evaluating toward potential upsampling solutions that closely resemble the input image.

6. The system of claim 5, wherein the fidelity term is dependent on a blur kernel.

7. The system of claim 6, wherein the program instructions are further executable by the one or more processors to implement an interface through which a shape or spread of the blur kernel is configurable by a user.

8. The system of claim 6, wherein the blur kernel comprises a box filter convolved with a disk.

9. The system of claim 1, wherein said identifying one of the plurality of potential upsampling solutions as an acceptable solution comprises minimizing the objective function using an iteratively re-weighted least squares procedure.

10. The system of claim 1, wherein the objective function models the sparse derivative prior as a zero mean Generalized Gaussian Distribution.

11. The system of claim 1, wherein the program instructions are further executable by the one or more processors to implement determining an upsampling factor by which resolution of the input image is to be increased.

12. The system of claim 1, wherein said evaluating, said identifying, and said storing are performed on each of two or more image planes of the input image.

13. The system of claim 1, wherein said evaluating, said identifying, and said storing are performed on a portion of the input image that is less than the entire input image.

14. The system of claim 13, wherein the portion of the input image comprises pixels near an edge of an image element in the input image.

15. The system of claim 1,

wherein the program instructions are further executable by the one or more processors to implement:

converting the input image to a different image representation than that of the input image; and

wherein said evaluating, said identifying, and said storing are performed on the different image representation of the input image.

16. The system of claim 1, wherein the one or more processors comprise at least one of:

a general-purpose central processing unit (CPU) or a graphics processing unit (GPU).

17. A non-transitory computer-readable storage medium, storing program instructions computer-executable to implement:

accessing data representing an input image to be upsampled;

for each of a plurality of potential upsampling solutions;

evaluating the potential upsampling solution with respect to an objective function, wherein the objective function is dependent on a sparse derivative prior in which sparse derivative prior probabilities for one or more second derivatives of the potential upsampling solution are multiplied by sparse derivative prior probabilities for one or more first derivatives of the potential upsampling solution;

identifying one of the plurality of potential upsampling solutions as an acceptable solution, dependent on said evaluating;

storing the identified solution as a higher-resolution version of the input image.

18. The storage medium of claim 17, wherein the objective function is dependent on a sparse derivative prior on two or more second derivatives of the potential upsampling solution, and wherein one of the two or more second derivatives of the potential upsampling solution comprises a derivative in a given direction on a result of a first derivative of the potential upsampling solution in a different direction.

19. The storage medium of claim 17, wherein the objective function is dependent on a sparse derivative prior on a derivative of the potential upsampling solution in a direction other than horizontal or vertical.

20. The storage medium of claim 17, wherein the program instructions are further computer-executable to implement an interface through which one or more of an upsampling factor, a target output image resolution, target relative resolution, a number of derivatives, an order of a derivative, a direction of a derivative, or a number of solution iterations is configurable by a user.

21. The storage medium of claim 17, wherein the objective function is further dependent on a fidelity term configured to bias said evaluating toward potential upsampling solutions that closely resemble the input image.

22. The storage medium of claim 17, wherein said identifying one of the plurality of potential upsampling solutions as an acceptable solution comprises minimizing the objective function using an iteratively re-weighted least squares procedure.

23. The storage medium of claim 17, wherein the objective function models the sparse derivative prior as a zero mean Generalized Gaussian Distribution.

24. The storage medium of claim 17, wherein said evaluating, said identifying, and said storing are performed on each of two or more image planes of the input image.

25. The storage medium of claim 17, wherein said evaluating, said identifying, and said storing are performed on a portion of the input image that is less than the entire input image.

26. The storage medium of claim 17,

wherein the program instructions are further computer-executable to implement:

converting the input image to a different image representation than that of the input image; and

wherein said evaluating, said identifying, and said storing are performed on the different image representation of the input image.

27. A method, comprising

performing, by a computer:

accessing data representing an input image to be upsampled;

for each of a plurality of potential upsampling solutions;

evaluating the potential upsampling solution with respect to an objective function, wherein the objective function is dependent on a sparse derivative prior in which sparse derivative prior probabilities for one or more second derivatives of the potential upsampling solution are multiplied by sparse derivative prior probabilities for one or more first derivatives of the potential upsampling solution;

identifying one of the plurality of potential upsampling solutions as an acceptable solution, dependent on said evaluating; and

storing the identified solution as a higher-resolution version of the input image.

28. The method of claim 27, wherein the objective function is dependent on a sparse derivative prior on a derivative of the potential upsampling solution in a direction other than horizontal or vertical.

29. The method of claim 27, wherein the objective function is further dependent on a fidelity term configured to bias said evaluating toward potential upsampling solutions that closely resemble the input image.

30. The method of claim 27, wherein said identifying one of the plurality of potential upsampling solutions as an acceptable solution comprises minimizing the objective function using an iteratively re-weighted least squares procedure.

31. The method of claim 27, wherein the objective function models the sparse derivative prior as a zero mean Generalized Gaussian Distribution.

32. The method of claim 27, wherein said evaluating, said identifying, and said storing are performed on each of two or more image planes of the input image.

33. The method of claim 27, wherein said evaluating, said identifying, and said storing are performed on a portion of the input image that is less than the entire input image.

34. The method of claim 27, further comprising using the computer to perform:

converting the input image to a different image representation than that of the input image;

wherein said evaluating, said identifying, and said storing are performed on the different image representation of the input image.

35. A computer-implemented method, comprising:

executing instructions on a specific apparatus so that binary digital electronic signals representing an input image to be upsampled are accessed in memory;

executing instructions on said specific apparatus so that binary digital electronic signals representing each of a plurality of potential upsampling solutions are evaluated with respect to an objective function, wherein the objective function is dependent on a sparse derivative prior in which sparse derivative prior probabilities for one or more second derivatives of the potential upsampling solution are multiplied by sparse derivative prior probabilities for one or more first derivatives of the potential upsampling solution;

executing instructions on said specific apparatus so that binary digital electronic signals representing one of the plurality of potential upsampling solutions is identified as an acceptable solution, dependent on said evaluating; and

storing the binary digital electronic signals representing the identified solution as a higher-resolution version of the input image in a memory location of said specific apparatus for later use.