Matsushita Shrinks Image from MPEG Streams
nikkeibp.asiabiztech.com
Matsushita Electric Industrial Co, Ltd has developed Realtime Moving Picture Coding Experts Group (MPEG) Reduction Technology, making it possible to shrink an image including multiple video, text and picture elements to a specified size directly at the MPEG data stream level.
The new technology has been applied to an Internet-based teleconferencing system linking up to nine sites. Instead of decoding the data to original image, shrinking it and then encoding it again, the new technique applies shrinking process to the decoding stage, that is, to discrete cosine transform (DCT) data. For example, to halve an image in both horizontal and vertical directions, one 4-pixel x 4-pixel low-frequency component block is extracted from each of four 8-pixel x 8-pixel DCT blocks, and the four 4-pixel x 4-pixel blocks are combined to form a single 8-pixel x 8-pixel DCT block which is again quantized and coded to the variable length code.
As a result, the processing volume is cut to 1/20th that of the conventional scheme, so that it is possible to reduce the size of any of the nine images being displayed simultaneously. The new approach offers much better image quality than shrinked imagery generated by simple downsampling.
The firm also developed I/P Frame Conversion Technology to maintain smooth frame rates for key images when Internet throughput (effective transfer rate) is low, and Media-Specific Optimized MPEG Generation Technology to enhance decoded image clarity by allocating more encoding bits to character and line drawing regions than to video area. |
