Video encoding processor LSI for multimedia

为了实现快速的MPEG-2编码，1998年开发了一种快速运动估计（ME）算法，称为“中断-中断搜索（BOS）”算法。通过在传统的乐趣搜索（FS）算法中引入计算约束，大大减少了计算量。在改进了BOS算法的基础上，1999年又提出了改进的中断-中断搜索（MBOS）算法。采用自适应分配阈值的MBOS算法可将处理速度提高3.4倍。它不仅可以保持与FS方法相同的视觉质量，而且有时可以获得比FS更高的PSNR。MBOS算法可以消除由于场景变化而导致的图像质量和速度的严重下降。此外，速度的大幅提高大大降低了ME阵列的功耗。采用MBOS方法，采用0.6 μm CMOS工艺制备了40.5 mhz、2.5 v、ME测试芯片。测量到的功耗降低到410 mW，约为使用FS的me阵列的1/14。

For fast MPEG-2 encoding, a fast motion-estimation (ME) algorithm, called the "breaking-off-search (BOS)" algorithm, was developed in 1998. The number of computations is drastically reduced by introducing computational constraints in a conventional fun-search (FS) algorithm. By improving the BOS algorithm, "modified breaking-off-search (MBOS)" was also developed in 1999. The MBOS algorithm with threshold values, which are adaptively assigned, can increase processing speed by a factor of 3.4. It can not only keep the same visual quality as that of the FS method, but can sometimes achieve a higher PSNR than FS. The MBOS algorithm can eliminate serious degradations of picture quality and speed associated with BOS due to the presence of scene changes. Furthermore, a drastic speed improvement substantially reduces the power dissipation of the ME array. A 40.5-MHz, 2.5-V, ME test chip using the MBOS method was fabricated by using the 0.6-μm CMOS process. The measured power dissipation was reduced to 410 mW, which is about 1/14 that of the ME-array using FS.