Lossless Index Coding for Image Vector Quantization Using Huffman Codes

Image vector quantization (VQ) has many current and future envisioned applications, such as digital image and signal compression, watermarking, data hiding, speaker identification. The idea of lossless index coding is to enhance the performance of image VQ by further encoding the index map of the VQ without introducing extra distortion. A novel lossless index coding scheme, called Index Associated List Coding (IALC), is proposed in this paper. When we encode the given current index, the first part of our scheme searches for an equivalent index among the neighboring indices. If such an index is found, the current index can be encoded with bits that represent the position of the equivalent index. Furthermore, we propose a coding structure, called Index Associated List (IAL), to encode the current index efficiently. The performance of IALC is evaluated by comparing it with three famous lossless index coding schemes: SOC, LCIC and LVIC. IALC outperforms the other three schemes in terms of average bit rate. Compared with conventional image VQ, IALC reduces the bit rate from 0.50 b/pixel to 0.26 b/pixel with a codebook size of 256.