Apparatus and method for frame rate conversion

• US 8,446,524 B2
• Filed: 06/19/2011
• Issued: 05/21/2013
• Est. Priority Date: 06/21/2010
• Status: Active Grant

• Current Assignee: Realtek Semiconductor Corp.
• Original Assignee: Realtek Semiconductor Corp.
• Inventors: Chen, Yung-Chin, Gong, Yun, Pao, I-Ming
• Primary Examiner(s): Natnael, Paulos M
• Application Number:
  US13/163,717
• Publication Number:
  US 20110310295A1
• Time in Patent Office: 702 Days
• Field of Search: 348/441, 348/451, 348/452, 348699-701, 348/458, 348/459, 375/240.16, 375/240.12, 375/240.24, 382/236, 382/300
• US Class Current: 348/441
• CPC Class Codes: H04N 19/577 Motion compensation with bidirectional frame interpolation, i.e. using B-pictures

Abstract

The disclosure is regarding an apparatus and a method for frame rate conversion. The frame rate conversion apparatus, for receiving at least two input frames to generate at lease one output frame, comprises a motion estimation module for generating a candidate list comprising regional motion vectors and temporal motion vectors from the at least two input frames and choosing a winner motion vector from the candidate list, a motion compensation module for generating the at lease one output frame according to the winner motion vector, a frame feature extraction module for generating fade-in/out information or a logo detection information from the at least two input frames, and an adaptive MEMC control module for deciding a weighting of the winner motion vector according to the fade-in/out information or the logo detection information.

Claims

1. A frame rate conversion method, comprising:

receiving at least two input frames comprising a reference frame and a target frame;

generating a plurality of motion vectors according to the at least two input frames;

generating a candidate list comprising regional motion vectors and temporal motion vectors from the plurality of motion vectors;

generating a winner candidate from the candidate list; and

generating at least one output frame according to the winner candidate;

wherein the step of generating the winner candidate from the candidate list comprises;

choosing a winner motion vector considering its temporal and spatial neighbors'"'"' motion vectors as the winner candidate.

2. The frame rate conversion method of claim 1, wherein the step of generating the candidate list comprising regional motion vectors and temporal motion vectors from the plurality of motion vectors comprises:

generating the candidate list comprising local regional motion vectors; and

generating the candidate list comprising MvFlow candidates.

3. The frame rate conversion method of claim 1, wherein the step of generating the winner candidate from the candidate list further comprises:

choosing the winner motion vector with the smallest SAD as the winner candidate from the candidate list.

4. The frame rate conversion method of claim 1, wherein the step of generating the winner candidate from the candidate list further comprises:

choosing the winner motion vector with the total smallest SAD, spatial bias, and temporal bias as the winner candidate from the candidate list.

5. The frame rate conversion method of claim 1, wherein the step of generating the winner candidate from the candidate list further comprises:

generating the winner candidate from the candidate list by bi-directional search scheme.

6. The frame rate conversion method of claim 1, further comprising the following step:

performing a post ME process to correct the winner motion vector, wherein the post ME process comprises at least one of MV jump correction, cover detection, and outlier filtering.

7. The frame rate conversion method of claim 1, further comprising the following step:

performing a cover detection by using SAD information according to bi-directional search scheme to detect potential cover and uncover areas in the at least two input frames and generating cover detection information.

8. The frame rate conversion method of claim 7, further comprising the following step:

performing a filtering based on the cover detection information to avoid discontinuity on block boundaries in the at least one output frame.

9. The frame rate conversion method of claim 1, further comprising the following step:

detecting a logo block according to whether a block is moving between the reference frame and the target frame, the time the block has been standing still, or the edge information of the block.

10. The frame rate conversion method of claim 1, further comprising the following step:

performing a frame reliability detection to decide how the winner candidate is weighted to generate the at least one output frame.

11. The frame rate conversion method of claim 1, wherein the step of generating the at least one output frame according to the winner candidate comprises:

controlling a interpolated pixel value according to SADs of the winner candidate and its neighboring motion vectors.

12. A frame rate conversion apparatus, for receiving at least two input frames to generate at least one output frame, comprising:

a motion estimation module, for generating a candidate list comprising regional motion vectors and temporal motion vectors from the at least two input frames and choosing a winner motion vector from the candidate list; and

a motion compensation module, for generating the at least one output frame according to the winner motion vector;

wherein the winner motion vector is generated according to its temporal and spatial neighbors'"'"' motion vectors.

13. The frame rate conversion apparatus of claim 12, wherein the regional motion vectors are local regional motion vectors and the temporal motion vectors are MvFlow candidates.

14. The frame rate conversion apparatus of claim 12, further comprising:

a frame feature extraction module, comprising a fade-in/out detection unit for generating fade-in/out information from the at least two input frames; and

an adaptive MEMC control module, for deciding a weighting of the winner motion vector according to the fade-in/out information;

wherein the at least one output frame is generated according the weighting.

15. The frame rate conversion apparatus of claim 12, further comprising:

a frame feature extraction module, comprising a logo detection unit for generating a logo detection information from the at least two input frames;

wherein the at least one output frame is generated according to the logo information.

16. The frame rate conversion apparatus of claim 12, wherein the winner motion vector is generated according to a bi-directional search scheme.

17. The frame rate conversion apparatus of claim 12, wherein the winner motion vector is with the smallest SAD in the candidate list.

18. The frame rate conversion apparatus of claim 12, wherein the motion estimation module further comprises a post ME unit to perform a post ME process to correct the winner motion vector, and wherein the post ME process comprises at least one of MV jump correction, cover detection, and outlier filtering.

19. A frame rate conversion method, comprising:

receiving at least two input frames comprising a reference frame and a target frame;

generating a plurality of motion vectors according to the at least two input frames;

generating a candidate list comprising regional motion vectors and temporal motion vectors from the plurality of motion vectors;

generating a winner candidate from the candidate list; and

generating at least one output frame according to the winner candidate;

wherein the step of generating the candidate list comprising regional motion vectors and temporal motion vectors from the plurality of motion vectors comprises;

generating the candidate list comprising local regional motion vectors; and

generating the candidate list comprising MvFlow candidatese.

20. A frame rate conversion method, comprising:

receiving at least two input frames comprising a reference frame and a target frame;

generating a plurality of motion vectors according to the at least two input frames;

generating a candidate list comprising regional motion vectors and temporal motion vectors from the plurality of motion vectors;

generating a winner candidate from the candidate list; and

generating at least one output frame according to the winner candidate;

wherein the step of generating the winner candidate from the candidate list comprises;

choosing a winner motion vector with the total smallest SAD, spatial bias, and temporal bias as the winner candidate from the candidate list.

21. A frame rate conversion method, comprising:

receiving at least two input frames comprising a reference frame and a target frame;

performing a cover detection by using SAD information according to bi-directional search scheme to detect potential cover and uncover areas in the at least two input frames and generating cover detection information;

generating a plurality of motion vectors according to the at least two input frames;

generating a candidate list comprising regional motion vectors and temporal motion vectors from the plurality of motion vectors;

generating a winner candidate from the candidate list; and

generating at least one output frame according to the winner candidate.

22. The frame rate conversion method of claim 21, further comprising the following step:

performing a filtering based on the cover detection information to avoid discontinuity on block boundaries in the at least one output frame.

23. A frame rate conversion method, comprising:

receiving at least two input frames comprising a reference frame and a target frame;

generating a plurality of motion vectors according to the at least two input frames;

generating a candidate list comprising regional motion vectors and temporal motion vectors from the plurality of motion vectors;

generating a winner candidate from the candidate list;

generating at least one output frame according to the winner candidate; and

detecting a logo block according to whether a block is moving between the reference frame and the target frame, the time the block has been standing still, or the edge information of the block.

24. A frame rate conversion method, comprising:

receiving at least two input frames comprising a reference frame and a target frame;

generating a plurality of motion vectors according to the at least two input frames;

generating a candidate list comprising regional motion vectors and temporal motion vectors from the plurality of motion vectors;

generating a winner candidate from the candidate list;

generating at least one output frame according to the winner candidate; and

wherein the step of generating the at least one output frame according to the winner candidate comprises;

controlling a interpolated pixel value according to SADs of the winner candidate and its neighboring motion vectors.