Image vector quantization using Hadamard transform subspace
Chi-Kwong Chan, Lai-Man Po · 2002
A new technique for reduction of the dimensionality of an image vector quantizer is proposed. In this technique, the convention space domain distortion measure is replaced by a transform domain distortion measure. Due to the energy compaction properties of transforms such as cosine or Hadamard transforms, the dimensionality of the transform domain distortion measure can be reduced drastically without affecting the performance of the quantizer. A modified LBG algorithm incorporating the new distortion measure is proposed. The performance and design considerations of a real-time image encoder using the new technique are investigated. A four times speed up in both codebook design time and search time are obtained along with a four times reduction in the size of the fast RAM.>