Statistically derived method and system for decoding MPEG motion compensation and transform coded video data

• US 5,596,369 A
• Filed: 01/24/1995
• Issued: 01/21/1997
• Est. Priority Date: 01/24/1995
• Status: Expired due to Term

• Current Assignee: MediaTek, Inc.
• Original Assignee: LSI Logic Corporation (Broadcom, Inc.)
• Inventors: Chau, Kwok K.
• Primary Examiner(s): Britton, Howard W.
• Application Number:
  US08/377,160
• Time in Patent Office: 728 Days
• Field of Search: 348/402, 348/413, 348/416, 348/423
• US Class Current: 375/240.15
• 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 19/61 in combination with predictive coding

Abstract

A Motion Picture Experts Group (MPEG) video/audio data bitstream includes macroblocks of video data which can each include input Motion Compensation (M) data and input discrete cosine Transform Coded (I) data. A motion pipeline processes the input M data to produce processed M data, and a transform pipeline processes the input I data to produce processed I data. A controller controls the motion pipeline and the transform pipeline to concurrently process the input M data and the input I data respectively such that a length of time required for processing each macroblock is variable and is determined by the largest of a length of time required for the motion pipeline to process the input M data and a length of time required for the transform pipeline to process the input I data of the macroblock. The time required to process all of the macroblocks in a video picture is statistically less than if a fixed length of time were provided for processing each macroblock based on a worst case scenario, thereby reducing the memory bandwidth requirement of the system.

Claims

1. A decoding system for decoding a video data bitstream including macroblocks of data which can each include input Motion Compensation (M) data and input Transform Coded (I) data, comprising:

a motion pipeline for processing said input M data to produce processed M data;

a transform pipeline for processing said input I data to produce processed I data; and

a controller for controlling the motion pipeline and the transform pipeline to concurrently process said input M data and said input I data respectively such that a length of time required for processing each macroblock is variable and is determined by the largest of a length of time required for the motion pipeline to process said input M data and a length of time required for the transform pipeline to process said input I data of said macroblock.

2. A system as in claim 1, in which:

each macroblock further comprises a header;

the system further comprises a data memory for storing said processed M and I data; and

the controller further causes said processed M data of each macroblock to be merged and stored in the data memory with said processed I data of said macroblock concurrently with reading a header of a next macroblock.

3. A system as in claim 2, in which:

the controller further stores said input M data and said input I data from said bitstream in the data memory;

the motion pipeline reads said input M data from the data memory and stores said processed M data in the data memory;

the transform pipeline reads said input I data from the data memory and stores said processed I data in the data memory; and

the controller further schedules the motion pipeline, the transform pipeline and the data memory such that reading and writing of I data has priority over reading and writing of M data.

4. A system as in claim 2, in which:

the controller further stores said input M data and said input I data from said bitstream in the data memory;

the motion pipeline reads segments of said input M data from the data memory and stores segments of said processed M data in the data memory;

the transform pipeline reads segments of said input I data from the data memory and stores segments of said processed I data in the data memory; and

the controller further controls the motion pipeline to inhibit reading a next segment of said input M data while the transform pipeline is processing a segment of said input I data.

5. A system as in claim 2, in which said input M data is included in said header of each macroblock respectively.

6. A system as in claim 2, in which:

said bitstream is encoded in accordance with an MPEG specification; and

said input I data is coded using a Discrete Cosine Transform (DCT).

7. A system as in claim 2, in which:

the motion pipeline, transform pipeline and controller are formed on a single integrated circuit chip; and

the data memory is separate from and operatively connected to said chip.

8. A system as in claim 7, in which the controller comprises:

a microcontroller formed on said chip; and

a program memory that is formed on the chip and stores a program for execution by the microcontroller.

9. A system as in claim 2, further comprising:

a communications channel for receiving said bitstream;

in which the controller further extracts said input M data and said input I data from said bitstream in the channel and stores said input M data and said input I data in the data memory.

10. A system as in claim 9, in which:

said bitstream is further variable length encoded; and

the system further comprises a variable length decoder for variable length decoding said bitstream before the controller stores said input M data and said input I data in the data memory.

11. A method of decoding a video data bitstream including macroblocks of data which can each include input Motion Compensation (M) data and input Transform Coded (I) data, comprising the steps of:

(a) processing said input M data to produce processed M data;

(b) processing said input I data to produce processed I data; and

(c) concurrently controlling processing of said input M data and said input I data in steps (a) and (b) respectively such that a length of time required for processing each macroblock is variable and is determined by the largest of a length of time required for processing said input M data in step (a) and a length of time required for processing said input I data in step (b) of said macroblock.

12. A method as in claim 11, in which:

each macroblock further comprises a header; and

the method further comprises the step of;

(d) causing processed M data of each macroblock to be merged and stored in a data memory with processed I data from said macroblock concurrently with reading a header of a next macroblock.

13. A method as in claim 12, further comprising the step of:

(e) storing said input M data and said input I data from said bitstream in the data memory;

in which

step (a) comprises reading said input M data from the data memory and storing said processed M data in the data memory;

step (b) comprises reading said input I data from the data memory and storing said processed I data in the data memory; and

step (c) comprises scheduling the motion pipeline, the transform pipeline and the data memory such that reading and writing of I data has priority over reading and writing of M data.

14. A method as in claim 12, in which further comprising the step of:

(e) storing said input M data and said input I data from said bitstream in the data memory;

in which

step (a) comprises reading segments of said input M data from the data memory and storing segments of said processed M data in the data memory;

step (b) comprises reading segments of said input I data from the data memory and storing segments of said processed I data in the data memory; and

step (c) comprises inhibiting reading a next segment of said input M data while processing a segment of said input I data.

15. A method as in claim 12, in which said input M data is included in said header of each respective macroblock.

16. A method as in claim 12, in which:

said bitstream is encoded in accordance with an MPEG specification; and

said input I data is coded using a Discrete Cosine Transform (DCT).

17. A method as in claim 12, further comprising the steps of:

(e) extracting said input M data and said input I data from bitstream; and

(f) storing said input M data and said input I data in the data memory.

18. A method as in claim 17, in which:

said bitstream is further variable length encoded; and

the method further comprises the step of;

(g) variable length decoding said bitstream before storing said input M data and said input I data in the data memory.