Synthetic aperture radar (SAR) image compression using the wavelet transform

Li, Ying (1997) Synthetic aperture radar (SAR) image compression using the wavelet transform. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (4MB)

Abstract

Along with the success of Synthetic Aperture Radar (SAR) imaging systems and the large amount of data which they routinely produce, SAR image compression has begun to attract the attention of researchers seeking solutions for efficient image transmission and storage. In this thesis, two key characteristics of SAR images, namely their speckle noise and varied image contents are addressed. We develop a selective soft-thresholding method and a multi-rate compression scheme to compress SA.R images more efficiently, using the wavelet transform to accomplish both the noise smoothing and image compression tasks. -- We initially approach the SAR image compression problem by studying the effects of SAR image characteristics using standard compression techniques, such as Joint Photographic Expert Group (JPEG) method, the Efficient Pyramidal Image Coder (EPIC) and the Embedded Zerotree Wavelet (EZW) coder. We find that speckle noise tends to break the inter-pixel correlation in SAR images and thus has negative effects on compression. In order to compress SAR images more efficiently, we need to smooth speckle noise and to enhance inter-pixel correlation prior to image compression. To this end, we develop a selective soft-thresholding method which refines Donoho’s overall soft-thresholding method. The proposed method makes use of the correlation structure of wavelet coefficients to select edge coefficients in the wavelet domain and then protects them from soft-thresholding. Test results show that this method can smooth speckle noise and preserve edges and hence enable more efficient compression. -- Another issue in SAR image compression is concerned with the varied scene contents within large-size SAR images. We propose a multi-rate compression scheme on top of the EZW algorithm. This scheme partitions an image into several regions in the wavelet domain. Each region is assigned a different bit budget according to the relative importance of the information each region contains. A highlighted region can be encoded with higher accuracy and only the major structures in the background regions transmitted. This scheme, when combined with the selective soft-thresholding method. Can provide better visual quality for the highlighted regions; major structures in the background can be easily picked up while finer details and speckle noise corrupted coefficients are eliminated at low hit-per-pixel rates. -- This work demonstrates applications of the wavelet transform in SA.R image processing and compression. Many issues for future work are also recommended.

Actions (login required)

View Item