Low-Complexity Image Coder/Decoder With An Approaching-Entropy Quad-Tree Search Code For Embedded Computing Platforms

Authors

Tao Ma, University of Nebraska LincolnFollow
Pradhumna Shrestha, University of Nebraska Lincoln
Michael Hempel, University of Nebraska-LincolnFollow
Dongming Peng, University of Nebraska LincolnFollow
Hamid Sharif, University of Nebraska-LincolnFollow

Date of this Version

2011

Citation

2011 18th IEEE International Conference on Image Processing; Digital Object Identifier: 10.1109/ICIP.2011.6116287

Comments

Copyright 2011 IEEE

Abstract

In this paper, we propose a fast, simple and efficient image codec applicable for embedded processing systems. Among the existing image coding methods, wavelet quad-tree is a foundation leading to an efficient structure to encode images. By searching significant coefficients along quad-trees, an embedded efficient code can be obtained. In this work, we exploit hierarchical relations of the quad-tree structure in terms of searching entropy and present a quad-tree searching model that is very close to the searching entropy. By applying this model, our codec surpasses SPIHT [1] by 0.2-0.4 db over wide code rates, and its performance is comparable to SPIHT with arithmetic coding and JPEG2000 [2]. With no additional overhead of arithmetic coding, our code is much faster and simpler than SPIHT with adaptive arithmetic coding and the more complicated JPEG2000 algorithms. This is a critical factor sought in embedded processing in communication systems where energy consumption and speed are priority concerns. Our simulation results demonstrate that the proposed codec is about twice as fast with very low computational overheads and comparable coding performances than existing algorithms.