Skip to Main Content
Moving object is too fast or a long exposure of a moving object will cause a "streak" of the moving object in the recorded video sequence. In this case, it's hard to recognize the detail of objects, and not necessary to encode the detail of motion which usually consumes lots of bits, especially under the situation of high resolution sequence. In this paper, we propose an upsampling based encoding method for the high resolution sequence with rapid motion. When rapid motion is detected in the current sequence, only a down-sampled version is encoded and sent to the decoder to save bit-rate. The decoder recovery the high resolution sequence through up-sampling the received low resolution bit stream. To avoid the visible artifacts and merge the synthesis into nowadays' I, P, B frame based codec, a synthetic region selection strategy is employed, which can help achieving a smooth transition between the synthetic frames and the adjacent non-synthetic frames. This scheme has been integrated into H.264, and experimental results show large amount of bit-rate is saved at similar visual quality levels compared with H.264.