Region-based method for iterative regularization image enhancement, and associated region-based apparatus and associated processing circuit

1. A region-based method for iterative regularization image enhancement, the region-based method being applied to a processing circuit, the region-based method comprising:

for a region of a plurality of regions of an image, performing at least one gradient calculation of an objective function of a latest representative function of the region and updating the latest representative function according to the gradient calculation; and

for the region of the plurality of regions of the image, when at least one predetermined convergence criterion is not satisfied, iteratively updating the latest representative function according to at least one gradient calculation of the same objective function of the latest representative function until the at least one predetermined convergence criterion is satisfied, in order to enhance the image with aid of iteratively updating the latest representative function.

2. The region-based method of claim 1 , further comprising:

for the region of the plurality of regions of the image, when the predetermined convergence criterion is satisfied, performing postprocessing on the latest representative function of the region to generate an iteratively regularized partial image corresponding to the region.

3. The region-based method of claim 1 , wherein the step of performing the gradient calculation of the objective function of the latest representative function of the region and updating the latest representative function according to the gradient calculation further comprises:

performing the gradient calculation of the objective function of the latest representative function of the region by calculating a derivative of the objective function of the latest representative function; and

updating the latest representative function according to the derivative of the objective function of the latest representative function.

4. The region-based method of claim 3 , wherein the step of performing the gradient calculation of the objective function of the latest representative function of the region and updating the latest representative function according to the gradient calculation further comprises:

updating the latest representative function with an increment/decrement whose absolute value is proportional to a function of the derivative of the objective function of the latest representative function.

5. The region-based method of claim 1 , wherein the objective function is a linear combination of a data fidelity function and one or more knowledge-guided regularization functions.

6. The region-based method of claim 5 , further comprising:

calculating the objective function by summing up the data fidelity function and the one or more prior knowledge-guided regularization functions weighted by at least one Lagrange multiplier.

7. The region-based method of claim 1 , wherein the plurality of regions of the image have their own sizes, respectively.

8. The region-based method of claim 1 , wherein the regions represent a plurality of overlapped partial images of the image; and the method further comprises:

utilizing the latest representative function of the region to generate an updated version of a specific portion of the region.

9. The region-based method of claim 1 , wherein the regions represent a plurality of non-overlapped partial images of the image; and the method further comprises:

utilizing the latest representative function of the region to generate an updated version of a whole region.

10. The region-based method of claim 1 , further comprising:

for another region of the plurality of regions of the image, performing at least one gradient calculation of another objective function of a latest representative function of the other region and updating the latest representative function of the other region according to the gradient calculation of the other objective function; and

for the other region of the plurality of regions of the image, when at least one predetermined convergence criterion utilized for the other region is not satisfied, iteratively updating the latest representative function of the other region according to at least one gradient calculation of the other objective function of the latest representative function of the other region until the at least one predetermined convergence criterion utilized for the other region is satisfied, in order to enhance the image with aid of iteratively updating the latest representative function of the other region, wherein the at least one predetermined convergence criterion utilized for the other region is not necessarily equivalent to the at least one predetermined convergence criterion utilized for the region.

11. The region-based method of claim 1 , further comprising:

for the plurality of regions of the image, utilizing respective one or more predetermined convergence criteria that are not necessarily equivalent to the at least one predetermined convergence criterion.

12. A region-based apparatus for iterative regularization image enhancement, the region-based apparatus being within a processing circuit, the region-based apparatus comprising:

a region engine arranged to perform iterative regularization image enhancement;

wherein for a region of a plurality of regions of an image, the region engine is arranged to perform at least one gradient calculation of an objective function of a latest representative function of the region and update the latest representative function according to the gradient calculation, and, when at least one predetermined convergence criterion is not satisfied, iteratively update the latest representative function according to at least one gradient calculation of the same objective function of the latest representative function until the at least one predetermined convergence criterion is satisfied, in order to enhance the image with aid of iteratively updating the latest representative function.

13. The region-based apparatus of claim 12 , wherein for each region of a plurality of regions of an image, the region engine is arranged to perform postprocessing on the latest representative function of the region to generate an iteratively regularized partial image corresponding to the region when the predetermined convergence criterion is satisfied.

14. The region-based apparatus of claim 12 , wherein the region engine is arranged to perform the gradient calculation of the objective function of the latest representative function of the region by calculating a derivative of the objective function of the latest representative function, and update the latest representative function according to the derivative of the objective function of the latest representative function.

15. The region-based apparatus of claim 14 , wherein the region engine is arranged to update the latest representative function with an increment/decrement whose absolute value is proportional to a function of the derivative of the objective function of the latest representative function.

16. The region-based apparatus of claim 12 , wherein the objective function is a linear combination of a data fidelity function and one or more knowledge-guided regularization functions.

17. The region-based apparatus of claim 16 , wherein the region engine is arranged to calculate the objective function by summing up the data fidelity function and the one or more prior knowledge-guided regularization functions weighted by at least one Lagrange multiplier.

18. The region-based apparatus of claim 12 , wherein the plurality of regions of the image have respective their own sizes, respectively.

19. The region-based apparatus of claim 12 , wherein the regions represent a plurality of overlapped partial images of the image; and the region engine is arranged to utilize the latest representative function of the region to generate an updated version of a specific portion of the region.

20. The region-based apparatus of claim 12 , wherein the regions represent a plurality of non-overlapped partial images of the image; and the region engine is arranged to utilize the latest representative function of the region to generate an updated version of a whole of the region.

21. The region-based apparatus of claim 12 , wherein for another region of the plurality of regions of the image, the region engine is arranged to perform at least one gradient calculation of another objective function of a latest representative function of the other region and update the latest representative function of the other region according to the gradient calculation of the other objective function, and, when at least one predetermined convergence criterion utilized for the other region is not satisfied, iteratively update the latest representative function of the other region according to at least one gradient calculation of the other objective function of the latest representative function of the other region until the at least one predetermined convergence criterion utilized for the other region is satisfied, in order to enhance the image with aid of iteratively updating the latest representative function of the other region, wherein the at least one predetermined convergence criterion utilized for the other region is not necessarily equivalent to the at least one predetermined convergence criterion utilized for the region.

22. The region-based apparatus of claim 12 , wherein for the plurality of regions of the image, the region engine is arranged to utilize respective predetermined convergence criteria that are not necessarily equivalent to the at least one predetermined convergence criterion.

23. The region-based apparatus of claim 12 , wherein the region engine comprises:

a first region buffer arranged to store data before an iteration; and

a second region buffer arranged to store data after the same iteration;

wherein the first region buffer and the second region buffer exchange their roles for a next iteration.

24. The region-based apparatus of claim 12 , wherein the region engine comprises:

a region line buffer for line-buffering within the region, wherein the region line buffer comprises one or more line buffers.

25. The region-based apparatus of claim 12 , wherein the region engine is composed of multiple region engines and the multiple region engines process data in parallel.