PDF《Now and Beyond》

55 (Mar. 2006)

13 Creating HD Pictures with H.264 Coding Method The current video coding technology, i.e. H.264, is two to three times higher in terms of compression performance than MPEG-2.Accordingly, it is gaining a considerable degree of attention regarding its adoption as a video coding technology for the Blu-ray Disc and HD DVD formats as well as digital broadcasting. This technology, however, requires ten times the number of encoding calculations than MPEG-2;

therefore, it is very important to work out how to do these calculations efficiently and how to raise the compression ratio while maintaining image quality. Focusing our attention on applications such as hard-disk recorders and DVD cameras, encoding LSIs, image-archiving systems, and video-surveillance systems, Hitachi is pushing forward the development of high-speed and high-quality H.264 coding methods. As regards natural images, an image texture comprises complicated parts and even parts within a picture. And as regards moving images, the subject comprises points of sudden movement and other points of relatively slow movement. On coding such images by H.264, though noise in the images is generated by compression of the data volume, about regions with large movements where the image pattern is complicated, noise in the eyes of people is hard to spot, and about regions with small movements where the image pattern is even, noise in the eyes of people is easy to spot. As shown in Fig. 4, with the developed coding method (i.e. H.264/AVC), the complexity of the image input at the time of coding is analyzed, and in the case that the regions where noise in the eyes of people is easy to spot, the coding amount is multiplied and allocated, and the image quality is improved. And in the case that the noise is easy to spot, the coding amount is cut back, and the compression efficiency is improved. Moreover, by feeding back the relationship between the generated code quantity, the remaining hypothetical buffer of the decoder, and the target code quantity, this control method does not fail even if the picture moves rapidly(2). In addition to the above method, we have developed a method that obtains the optimum quantization parameter for the coding amount and picture quality by iterative calculation, thereby improving compression efficiency(3). And we are concentrating our efforts on developing technologies―such as realizing long-term recording that maintains image quality―aimed at product applications. FUTURE TECHNOLOGIES FOR VIDEO CODING AND THEIR PROSPECTS Future Video Coding Technologies At the current MPEG/JVT meeting, on SVC (scalable video coding)―which is a coding technology for multi-resolution images―is being actively discussed as an extension to the H.264 standard. This transmission method multiplexes a number of image Fig. 3―Effect of Error-control Technique for Motion Compensation. By introducing control variables, the reddening problem (due to accumulation of rounding errors by motion compensation) is solved. Fig. 4―Overview of High-quality Coding Method H.264/AVC. As regards natural images, regions of easy-to-spot noise and not-so-easy-to-spot noise exist. By distinguishing these areas and giving them appropriate coding amount, coding with high- picture quality is possible. (a) Without error control Reddening (b) With error control Example of coded image Parts where movement is small and image (pattern) is even Noise is easy to spot. Coding amount is allocated and image quality is greatly improved. Parts where movement is large and image (pattern) is complex. Noise is difficult to spot. Less coding amount is used and compression efficiency is increased. Video ........