The rate control strategy in hybrid DPCM/DCT video codec like MPEG, has a great effects on the quality of the reconstructed image sequence. Especially, in a poorly predicted picture sequence on account of scene changes, zooming, and fast motion, image quality degradation can occur due to the little correlation to the previous picture.
This paper presents a simple approach to improve the performance of the MPEG codec in these cases. When these cases occur, picture target bit is adjusted in the direction to minimize the distortion in a GOP on the Rate-Distortion relation obtained from the spatial domain variances for each P-picture or macroblock coding mode. When the bit shortage could be occur, the proposed method extends the current GOP to the next. The proposed algorithm can prevent the buffer from the abnormal operations such as over/under-flow by the target bit adjustment using the current and the next predicted buffer fullness. This paper also proposes a hardware implementation method of the proposed bitrate control algorithm. The algorithm can be applied to the existing MPEG codecs and real-time applications easily. From the experimental results, the proposed algorithm gives 0.5~3 dB improvement in PSNR and shows a small fluctuation in quality after scene change compared with the MPEG-2 TM5 rate control algorithm. This also guarantees a stable operation of the encoding buffer in the poorly predicted picture sequence.
