IPC분류정보
국가/구분 |
United States(US) Patent
등록
|
국제특허분류(IPC7판) |
|
출원번호 |
US-0452960
(1999-12-02)
|
우선권정보 |
SG-9902417 (1999-05-21) |
발명자
/ 주소 |
- Lin, Weisi
- Tye, Bee June
- Ong, Ee Ping
|
출원인 / 주소 |
- Institute of Microelectronics
|
대리인 / 주소 |
Akin Gump Strauss Hauer & Feld LLP
|
인용정보 |
피인용 횟수 :
30 인용 특허 :
7 |
초록
▼
A method for use in a real-time video encoder based on two processing means to optimise the use of computing power of the processing means. This is achieved by dynamically load-balancing between two processing means. The load pattern is determined with measures for the progress of motion estimation
A method for use in a real-time video encoder based on two processing means to optimise the use of computing power of the processing means. This is achieved by dynamically load-balancing between two processing means. The load pattern is determined with measures for the progress of motion estimation in one processing means and the idle time in the other. Adaptive adjustment of load balancing is fulfilled via making a decision on whether the DCT/IDCT/Quantization/Dequantization sub-process is skipped for a macroblock. The invention is not constrained to particular motion estimation techniques.
대표청구항
▼
1. A method for balancing the workload between a first and second processing means for video encoding, the method comprising: measuring the load distribution dynamically between the first and second processing means, and adaptively adjusting the amount of computation performed by the second pro
1. A method for balancing the workload between a first and second processing means for video encoding, the method comprising: measuring the load distribution dynamically between the first and second processing means, and adaptively adjusting the amount of computation performed by the second processing means based on said load distribution, wherein the first processing means performs motion estimation and motion estimation associated functions and the second processing means performs discrete cosine transformation (DCT), inverse discrete cosine transformation (IDCT), quantization, dequantization and source coding functions, and wherein the first processing means performs functions different than the second processing means. 2. The method of claim 1 wherein the measurement of the load distribution is realized by the measurement of the amount of idle time of the second processing means and counting the number of motion-estimation-completed macroblocks awaiting DCT.3. The method of claim 2 wherein the measurement of the load distribution includes the measurement of the progress of motion estimation in the first processing means.4. The method of claim 1 wherein adaptively adjusting the amount of computation performed by the second processing means is performed by dynamically determining whether or not any of DCT, IDCT, quantization and dequantization should be performed.5. The method of claim 4 wherein the video to be coded contains frames divided into macroblocks, and wherein a parameter is determined for each macroblock which allows the second processing means to skip any of DCT, IDCT, quantization and dequantization for that macroblock.6. The method of claim 4 wherein the video to be coded contains frames divided into macroblocks, and wherein a parameter is determined for each frame which allows the second processing means to skip any of DCT, IDCT, quantization and dequantization for the macroblocks of that frame.7. The method of claim 5 wherein the measurement of the load distribution includes the determination of the idle time of the second processing means.8. The method of claim 7 wherein the measurement of the load distribution includes the measurement of the progress of motion estimation in the first processing means.9. A method for balancing the workload between a first processing means and a second processing means for video encoding of video frames divided into macroblocks, comprising: using the first processing means to perform motion estimation; using the second processing means to perform any of discrete cosine transformation (DCT), inverse discrete cosine transformation (IDCT), quantization and dequantization; determining the load distribution dynamically between the first and second processing means using a combination of the idle time of the second processing means and the progress of motion estimation in the first processing means; and determining for each macroblock whether any of DCT, IDCT, quantization and dequantization should be performed on that macroblock by the second processing means based on said load distribution. 10. The method of claim 9 wherein the determination for each macroblock as to whether DCT, IDCT, quantization and dequantization should be performed on that macroblock is effected by the use of a parameter which is dependent on the idle time of the second processing means and the progress of motion estimation in the first processing means.11. A method for balancing the workload between a first processing means and a second processing means for video encoding of video frames divided into macroblocks, comprising: using the first processing means to perform motion estimation; using the second processing means to perform discrete cosine transformation (DCT), inverse discrete cosine transformation (IDCT), quantization and dequantization; determining a load pattern dynamically between the first and second processing means using a combination of the idle time of the se cond processing means and the progress of motion estimation in the first processing means; and determining for each frame of the video whether DCT, IDCT, quantization and dequantization should be performed on the macroblocks of that frame by the second processing means based on said load pattern. 12. A method for balancing the workload between a first processing means and a second processing means, the first processing means performing motion estimation functions and the second processing means performing discrete cosine transformation (DCT), inverse discrete cosine transformation (IDCT), quantization, dequantization and source coding functions, the first processing means performing functions different than the second processing means, wherein the amount of computation in the second processing means is reduced when the first processing means progresses faster than the second processing means; otherwise the amount of computation in the second processing means is increased. 13. A method for balancing the workload between a first and second processing means for video encoding, the first processing means performing motion estimation functions and the second processing means performing discrete cosine transformation (DCT), inverse discrete cosine transformation (IDCT), quantization, dequantization and source coding functions, the first processing means performing functions different than the second processing means, the method comprising the steps of: measuring the load distribution dynamically between the first and second processing means, and adaptively adjusting the amount of computation performed by the second processing means based on said measured load distribution; wherein the amount of computation in the second processing means is reduced when the first processing means progresses faster than the second processing means; otherwise the amount of computation in the second processing means is increased.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.