A fast route for application of rate-distortion optimal quantization in an MPEG video encoder

The freedom which is offered by the adaptable MPEG video encoders is meaningfully exploited by means of a “rate-distortion optimal” framework using a Lagrange formalism, as is well known from the literature. One such feature of interest is “optimal quantization”. We propose some practical short-cut solutions, which reduce the computational complexity typically associated with “rate-distortion optimal” coding. The Lagrange multiplier, which is the most important parameter controlling the encoding of a single picture, is successfully predicted on the basis of a set of histograms for the amplitude distribution of each DCT-coefficient. In combination with a block-based bit-rate control the selection of the optimal quantizer step-size is further simplified