A New Low Complexity Distance and Its use For Vector Quantization of Images

A new distance norm [5] is used with the LBG algorithm and its eectiveness in achieving minimum distortion is studied with respect to Euclidean and L 1 norms. Partial Sum and Nearest Neighbouring Distance (NND) approaches are used with this new norm to reduce vector quantization (VQ) codebook design time. The mean square error (MSE) achieved with this new norm is similar to that of the conventional LBG algorithm which uses the Euclidean norm. A CLBG algorithm is proposed which uses both new and LBG algorithms sequentially which gives lower MSE values than all other methods. This norm can also be used with a binary splitting operation in the initial codebook generation process. Experiments are carried out on dierent real images.