文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Recovering from Bit Errors in Scalar-Quantized Discrete Wavelet Transformed Images
侵权投诉

Recovering from Bit Errors in Scalar-Quantized Discrete Wavelet Transformed Images

In this paper we study the effects of transmission noise on fixed-length coded wavelet coefficients. We extend our previous work on maximum a posteriori detectors [1] to include interbitplane information and determine which transmitted codeword was most li

Recovering from Bit Errors in Scalar-Quantized Discrete Wavelet Transformed Images Alexander E. Mohr Eve A. Riskin Richard E. Ladner Electrical Engineering Electrical Engineering Computer Science University of Washington University of Washington University of Washington Seattle, WA 98195{2500 Seattle, WA 98195{2500 Seattle, WA 98195{2350 amohr@isdl.ee.washington.edu riskin@isdl.ee.washington.edu ladner@cs.washington.edu

Abstract In this paper we study the e ects of transmission noise on xed-length coded wavelet coe cients. We extend our previous work on maximum a posteriori detectors 1] to include interbitplane information and determine which transmitted codeword was most likely corrupted into a received erroneous codeword. We present a simple method of recovering from these errors once detected and demonstrate our restoration methodology on scalar-quantized wavelet coe cients that have been transmitted across a binary symmetric channel. The wavelet transform 2] is a tool for spacefrequency analysis that uses wavelet basis functions to decompose signals in L2 (R) into distinct frequency subbands, each of which corresponds to di erent scale information. The 2-dimensional discrete wavelet transform (DWT) decomposes an image into high and low frequency components along each image dimension. Figure 1 shows a 3-level DWT. Wavelets used for lossy transform compression have provided some of the best compression performance to date 3, 4]. The question of robustness to transmission noise and error is beginning to be explored; many of these This work was supported by U. S. Army Research Ofce Grant DAAH004-96-1-0255. Published in the 1998 International Conference on Image Processing (ICIP'98), October 4-7, 1998, in Chicago, Il. For more publications, visit doc.guandang.net

1 Introduction

encoded bit streams are not robust when transmitted without explicit error control codes 5]. We use a discrete wavelet image that has been quantized by an organized quantizer to present a simple method of recovering from bit errors introduced by a binary symmetric channel. We use xed-length scalar quantization codebooks that have been organized progressively. As a result, errors in the most signi cant bits of a codeword index result in large changes in the coe cient value, which are easily detected, while errors in the least signi cant bits of a codeword index result in only minor changes in the coe cient value, rendering them less important to correct. Our motivations for using scalar quantization are that (1) it has been widely studied and (2) it provides a clear framework within which to study the e ects of channel noise on codeword indexes. Our algorithm exploits the redundancy, or correlation, that remains in the low-frequency subbands. Because of that redundancy, a coe cient's neighbors can give some insight into the expected range of values for the central coe cient. If a coe cient has a value that is outside of the expected range, i

t is possible that the codeword index has been corrupted by bit errors. For low bit error rates (less than 5%), it is also probable that an erroneous codeword index di ers in only one bit from the correct one. Our correction algorithm has two components. The rst component corrects bit errors in codeword indexes, while the second modi es coe 1

2 Approach

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email