Low power, high performance transform coprocessor for video compression

• US 7,756,351 B2
• Filed: 10/28/2004
• Issued: 07/13/2010
• Est. Priority Date: 12/19/2003
• Status: Active Grant

• Current Assignee: STMicroelectronics Incorporated (STMicroelectronics NV)
• Original Assignee: STMicroelectronics Incorporated (STMicroelectronics NV)
• Inventors: Dang, Philip P.
• Primary Examiner(s): Dang; Duy M
• Application Number:
  US10/975,638
• Publication Number:
  US 20050141776A1
• Time in Patent Office: 2,084 Days
• Field of Search: 382232-236, 382239-250, 708/401, 708/402, 375/240.05
• US Class Current: 382/250
• CPC Class Codes: H04N 19/42 characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation H04N19/635 takes precedence
• CPC Class Codes: H04N 19/436 using parallelised computational arrangements
• CPC Class Codes: H04N 19/60 using transform coding

Abstract

First and second integer transform matrices can be used to approximate the discrete cosine transform. An input matrix of data is multiplied by a first transform matrix of integers to produce an intermediate matrix of data. The intermediate matrix is multiplied by a second transform matrix of integers to produce a transform result matrix of data. The multiplications by the first and second transform matrices can be pipelined to increase throughput. A plurality of transform data paths can also be provided in parallel to increase throughput.

Claims

1. An apparatus for calculating integer transforms of input matrices of data, comprising:

an input for receiving input matrices of data;

a matrix multiplication apparatus coupled to said input;

said matrix multiplication apparatus including a first matrix multiplier for multiplying a first input matrix of data by a first transform matrix of integers to produce a first intermediate matrix of data;

said matrix multiplication apparatus including a second matrix multiplier coupled to said first matrix multiplier for multiplying the first intermediate matrix by a second transform matrix of integers to produce a first transform result matrix of data; and

a further said matrix multiplication apparatus coupled to said input for operation in parallel with said first-mentioned matrix multiplication apparatus, wherein said matrix multiplication apparatus includes a third matrix multiplier coupled to said second matrix multiplier for multiplying a transpose of the first transform result matrix by the first transform matrix to produce a second intermediate matrix of data and a fourth matrix multiplier coupled to said third matrix multiplier for multiplying the second intermediate matrix by the second transform matrix to produce a two-dimensional result matrix.

2. The apparatus of claim 1, wherein said first matrix multiplier is further for multiplying a second input matrix of data by the first transform matrix to produce a second intermediate matrix of data while said second matrix multiplier is simultaneously multiplying the first intermediate matrix by the second transform matrix.

3. The apparatus of claim 1, wherein said second matrix multiplier is for producing a second said transform result matrix of data, said third matrix multiplier is further for multiplying a transpose of said second transform result matrix by the first transform matrix to produce a third intermediate matrix of data while said fourth matrix multiplier is simultaneously multiplying the second intermediate matrix by the second transform matrix, and wherein said first matrix multiplier is for multiplying a second input matrix of data by the first transform matrix to produce a fourth intermediate matrix of data while said second matrix multiplier is simultaneously multiplying the first intermediate matrix by the second transform matrix.

4. The apparatus of claim 3, including a further said matrix multiplication apparatus coupled to said input for operation in parallel with said first-mentioned matrix multiplication apparatus.

5. The apparatus of claim 1, including a further said matrix multiplication apparatus coupled to said input for operation in parallel with said first-mentioned matrix multiplication apparatus.

6. The apparatus of claim 1, wherein said data is H.264 video data.

7. A video data processing system, comprising:

a video data processor including an input for receiving input matrices of data, and a transform apparatus coupled to said input for calculating integer transforms of said input matrices of data, said transform apparatus including a first matrix multiplier for multiplying a first input matrix of data by a first transform matrix of integers to produce a first intermediate matrix of data, and said transform apparatus including a second matrix multiplier coupled to said first matrix multiplier for multiplying the first intermediate matrix by a second transform matrix of integers to produce a first transform result matrix of data, wherein said transform apparatus includes a third matrix multiplier coupled to said second matrix multiplier for multiplying a transpose of the first transform result matrix by the first transform matrix to produce a second intermediate matrix of data, and said matrix multiplication apparatus including a fourth matrix multiplier coupled to said third matrix multiplier for multiplying the second intermediate matrix by the second transform matrix to produce a two-dimensional result matrix; and

an interface coupled to said video data processor for permitting a user to communicate with said video data processor.

8. The system of claim 7, including a further data processor coupled to said video data processor for controlling operation of said video data processor.

9. The system of claim 7, provided as one of a high definition television system and a digital cinema system.

10. The system of claim 7, wherein said data is H.264 video data.

11. A method of calculating integer transforms of input matrices of data, comprising:

storing a first input matrix of data and a first transform matrix of integers into a memory;

multiplying the first input matrix of data by the first transform matrix of integers to produce a first intermediate matrix of data, wherein the multiplication of the first input matrix of data by the first transform matrix of integers is preformed by at least one processor;

multiplying the first intermediate matrix by a second transform matrix of integers to produce a first transform result matrix of data; and

storing the first transform result matrix of data into the memory, wherein the method includes multiplying a third input matrix of data by the first transform matrix to produce a second intermediate matrix of data while multiplying the first input matrix by the first transform matrix, and multiplying the second intermediate matrix by the second transform matrix to produce a second transform result matrix of data while multiplying the first intermediate matrix by the second transform matrix.

12. The method of claim 11, including multiplying a second input matrix of data by the first transform matrix to produce a second intermediate matrix of data while simultaneously multiplying the first intermediate matrix by the second transform matrix.

13. The method of claim 12, wherein the second input matrix follows the first input matrix in a first received sequence of input matrices of data.

14. The method of claim 13, including multiplying the third input matrix of data by the first transform matrix to produce a third intermediate matrix of data while simultaneously multiplying the first input matrix by the first transform matrix, multiplying the third intermediate matrix by the second transform matrix to produce a second transform result matrix of data while simultaneously multiplying the first intermediate matrix by the second transform matrix, and multiplying a fourth input matrix of video data by the first transform matrix to produce a fourth intermediate matrix of data while simultaneously multiplying the third intermediate matrix by the second transform matrix and also while simultaneously multiplying the second input matrix by the first transform matrix and also while simultaneously multiplying the first intermediate matrix by the second transform matrix.

15. The method of claim 14, wherein the fourth input matrix follows the third input matrix in a second received sequence of input matrices of data provided in parallel with the first sequence of input matrices of data.

16. The method of claim 11, including multiplying a transpose of the first transform result matrix by the first transform matrix to produce a second intermediate matrix of data, and multiplying the second intermediate matrix by the second transform matrix to produce a two-dimensional transform result matrix.

17. The method of claim 16, including performing said matrix multiplying steps with respect to a second input matrix of data to produce a second transform result matrix of data, multiplying a transpose of the second transform result matrix by the first transform matrix to produce a third intermediate matrix of data while simultaneously multiplying the second intermediate matrix by the second transform matrix, and said performing step including multiplying the second input matrix by the first transform matrix to produce a fourth intermediate matrix of data while simultaneously multiplying the first intermediate matrix by the second transform matrix.

18. The method of claim 11, wherein each of said matrix multiplying steps includes a plurality of arithmetic operations consisting of (a) adding eight pairs of addends to produce eight respectively corresponding sums, and (b) multiplying two pairs of factors to produce two respectively corresponding products.

19. The method of claim 18, wherein said step of multiplying two pairs of factors includes performing a shift operation on a digital representation of one of the factors of each of said two pairs of factors.

20. The method of claim 11, wherein all of said matrices are 4×

4 matrices.

21. The method of claim 11, wherein said data is H.264 video data.