High compression of chrominance exploiting its correlation with luminance

The paper reports research on original techniques for compression of chrominance data in video sequences transmitted at very low bit rates. Two chrominance components are converted into one scalar chrominance signal using vector quantization in chrominance plane. Three methods for scalar chrominance compression based on transform and predictive coding are described. The important issue is that presented algorithms exploit mutual correlation between vector quantized chrominance and the luminance component of video data. In transform coding techniques, the activity of reconstructed luminance estimated for its transform coefficients controls quantization steps of respective transform coefficients of scalar chrominance. In the approach based on predictive coding, detection of active areas in the luminance component allows for efficient coding of chrominance prediction error, where unnecessary transmission of zero-valued errors is avoided. The paper presents experimental results obtained for the proposed methods.