Comparison of different coding schemes for SAR image compression

• Published in: ISIE '99. Proceedings of the IEEE International Symposium on Industrial Electronics (Cat. No.99TH8465) Vol. 1; pp. 114 - 117 vol.1
• Main Authors: Gleich, D., Planinsic, P., Gergic, B., Banjanin, B., Cucej, Z.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 1999
• Subjects: Adaptive optics; Image coding; Optical distortion; Optical noise; Optical sensors; Quantization; Statistical distributions; Synthetic aperture radar; Testing; Wavelet domain
• ISBN: 0780356624; 9780780356627
• DOI: 10.1109/ISIE.1999.801768

Recently, a new class of image coding algorithms (EWZ SPIHT) based on embedded zerotree method (EWZ) has appeared. Methods use zerotree-structured quantization of subbands. These algorithms were developed and tested on optical images. The authors investigated their behaviour on amplitudinal synthetic aperture radar (SAR) data images. SAR data images have different statistical distribution than optical images. They contain a lot of typical noise called speckles, so energy in the wavelet domain gets differently disposed than in optical images, i.e., details contain more energy. The authors compared performances of these methods and compared them with the classical method that employs almost optimal bit allocation using scalar quantization of subbands followed by Recency Rank coding. EWZ based methods exploit dependency parent-children relationship between multiresolution subimages. SPIHT uses a sorting algorithm in hierarchical trees. The proposed methods, based on zero-tree quantization are numerically simple and give better results in the rate distortion sense than the classical method with Recency Rank coding.