Region-of-interest based H.264 encoding parameter allocation for low power video communication

H.264 is the state-of-the-art in modern video compression standards. Its extremely high compression ratio meets the requirements the video communication between portable devices. Since the power is limited in portable devices, the huge computation of H.264 is a critical problem for hardware design. According to the human visual system (HVS) research, human vision is only able to focus on one area in a frame, which is defined as region-of-interest (ROI). In most cases, human face attracts the most attention of the device user. This phenomenon gives a chance to code all macroblocks unequally. In this work, the ROI is detected with an encoder-oriented fast algorithm, using chrominance and texture contrast features. After the ROI is detected, the encoder will allocated the coding parameters respectively in ROI and non-ROI. As a result, it keeps fine quality in ROI, saves much throughput in non-ROI, and greatly reduces the computation. The ROI detector and the encoder are also designed to be decoding-friendly and hardware-friendly.