New Clustering Algorithm for Vector Quantization Using Hybrid Haar Slant Error Vector

Abstract

Vector Quantization is an effective and simple method for lossy image compression. Codebook generation acts as important phase in Vector Quantization (VQ). The codebook is used to encode the image blocks for image compression. The codebook generation algorithm is generally preferred to have minimum distortion between the original image and obtained the reconstructed image. The paper presents an effective clustering algorithm to generate codebook for vector quantization. In Kekre's Error Vector Rotation (KEVR) while splitting the cluster every time new orientation is introduced using error vector sequence. As this error vector sequence is binary representation of numbers, cluster orientation change slowly in every iteration. The Kekre's Error Vector Rotation using Walsh (KEVRW) uses Walsh sequence to rotate the error vector. Because of this cluster orientation change rapidly in every iteration. The proposed VQ codebook generation technique Thepade's Hybrid Haar Slant Error Vector Rotation (THHSlEVR) is based on KEVR algorithm. Here the error vector used for splitting the clusters in Vector Quantization is proposed to be prepared using discrete Slant transform matrix and Haar matrix. The proposed VQ codebook generation methodology is tested on different images for various codebook sizes. The obtained results show that proposed VQ codebook generation algorithm gives less MSE and less distortion as compared to KEVR, KEVRW indicating better image compression.