Color image compression based on vector quantization using PCA and LEBLD

This paper presents a new algorithm for the compression of color images, which uses PCA (principal component analysis) and LEBLD (least error boundary line detection). PCA is a statistical technique for linearly reducing the dimensionality of input vector while retaining as much of the information present in the input vector as possible. LEBLD is a new technique that detects optimal boundary line to divide a data set into two groups. By applying PCA to VQ (vector quantization) codebook design, the superior codebook is generated fast. We design the three separate codebooks for three color components Y, Cb and Cr to generate effective codebooks by adjusting the number of nodes for each tristimulus value Y, Cb and Cr according to the correlation among components