Reduced-complexity multiview prediction scheme with content-adaptive disparity vector estimation

摘要: Disparity estimation is a highly complex and time consuming process in multiview video encoders. Since multiple views taken from two-dimensional camera array need to be coded at every instance, the complexity of encoder plays an important role besides its rate-distortion performance. In previous papers we have introduced new frame type called D (derived) that exploits strong geometrical correspondence between views, thereby reducing encoder. By employing frames instead some P prediction structure, significant gain can achieved if threshold value, which keystone element adjust cost quality and/or bit-rate, selected wisely. A adaptive method calculate value automatically existing information during encoding presented. this method, values are generated for each block increase accuracy. The algorithm applied several image sets 20.6% using without compromising or bit-rate.