Abstract:Vector Quantization (VQ) is one of effective image compression techniques. High compression quality depends on designing a good VQ codebook. Based on two optimality conditions, the conventional K-means algorithm is a widely used scheme to generate a codebook. However, the initial codebook greatly affects both the convergence speed and the quality of the generated codebook. Therefore, many researchers have presented all kinds of methods to obtain a good initial codebook, including splitting-based, pruning-based, pairwise nearest neighbor-based, random initialization-based and maximum distance initialization-based schemes. However, all these techniques do not take into account the features of the training vectors and some of them need high extra computational load. This paper presents a simple classification based technique to obtain a better initial codebook. During the initial codebook generation process, we classify the training vectors according to their edge orientations (eight templates) and their contrast information (binary decision), and thus we can obtain sixteen subsets. Then we just randomly select several codewords from each subset proportionally to the number of training vectors in the subset. Experimental results demonstrate that, compared with the conventional and modified K-means algorithms under the random initialization strategy, our scheme converges to a better codebook with fewer iterations.

Multi-Stage Codebook Optimization for LSF Quantization with Inter-Stage Prediction

Subframe Interpolation Optimized Coding Of Lsf Parameters

Codebook Sharing in Multi-Stage Vector Quantization

Joint Optimization Algorithm of Codebook Size in Multi-Parameter, Multi-Mode Quantization

Joint optimization algorithm of multi-parameter codebook size based on superframe mode

Mode Based Multi-Stage Vector Quantization with Shared Codebooks

High Efficiency MSVQ for Prediction Linear Spectrum Frequency Parameters with Inter-Frame and Inter-Stage Prediction

Research on Low Delay Low Bit Rate Speech Coding Algorithm

Improvement of Delayed Decision Coding for LSF Difference Quantization

A novel adaptive mutation PSO vector quantization

Efficient LSP Quantization Algorithm for Very Low Bit Rate Speech Coding

Single-Codec: Single-Codebook Speech Codec towards High-Performance Speech Generation

VQ Codebook Design Using Modified K-means Algorithm with Feature Classification and Grouping Based Initialization

A Low-Complexity 2.4 kb/s Speech Coding Algorithm Based on Sinusoidal Excitation Linear Prediction

Fast Codebook Design Method for Image Vector Quantisation

Predicting Multi-Codebook Vector Quantization Indexes for Knowledge Distillation

Research of the Line Spectrum Frequency Quantization of the G.729 Speech Coding Algorithm

8kbit/S Low Delay Speech Coding Algorithm With Adaptive Codebook

A Method Of Optimizing Codebook Based On Codeword Use Frequency

High Efficient Quantization of the Energy Parameter in 0.6 Kb/s Vocoders

Edge and Contrast Classified K-means Algorithm for Image Vector Quantizer Design.