Improving the performance of MPEG compatible encoding at low bit rates using adaptive neural networks

A new approach is presented in this paper for improving the performance of MPEG encoders, especially in videophone or videoconferencing applications, through allocation of a greater number of bits in objects that belong to th e foreground of image frames, than in objects that belong to the background . A human face and body detector followed by a neural network classifier ar e used for foreground/background object extraction. The derived image segme ntation is used to modify the rate control of MPEG schemes so as to allocat e more bits to foreground objects than to background, while retaining compa tibility with MPEG encoders. Experimental results are presented, including image sequences with complex backgrounds, which illustrate the performance of the proposed scheme. Both a subjective image quality improvement and a P SNR increase of about 1.35 db on average have been obtained. (C) 2000 Acade mic Press.