1Synthetic Aperture Radar ImageCompressionByMagesh ValliappanGuner Arslan2Synthetic Aperture Radar (SAR)✔ SAR ?– Active imaging system– Working in the frequency range 1-10 GHz– All-weather system– High resolution compared to real aperture radar✔ Applications– Agriculture, ecology, geology, oceanography,hydrology, military...✔ Nature of SAR images– High volume of data– Speckle noise– More information in high frequencies than opticalimages3Lossy Image Compression Techniques✔ Joint Photographic ExpertsGroup (JPEG)– Discrete Cosine Transform– Fast implementation– Blocking artifacts✔ Set Partitioning InHierarchical Trees (SPIHT)– Discrete Wavelet Transform– Good visual quality– Ringing effect for high compressionratios4Quality Metrics for SAR Images✔ Standard Metrics– Mean Squared Error (MSE)– Signal to Noise Ratio (SNR)– Peak Signal to Noise Ratio (PSNR)✔ Other Metrics for SAR Images– Weighted Signal to Noise Ratio (WSNR)– Linear Distortion Quality Measure– Correlation of Edge Information5Simulations✔ Space borne Imaging Radar-C and X-BandSynthetic Aperture Radar✔ 512 x 512 Sub-Images✔ 8 bit grayscale✔ Pre-filtered by a modified σ-filter– adapted to handle spot noise6Estimation of a Linear Model✔ Linear Least Square Estimate✔ Linear Model is needed to– compute the Noise Image– estimate the Distortion Transfer Function (DTF)✔ Drawbacks– Model assumes uncorrelated additive noise– Variance of the estimateHNoiseImageSARImageCompression De-compression7Linear ModelsSPIHTJPEGCSF8Correlation of Edge InformationJPEGSPIHTOriginal9Results - WSNR and PSNR(dB)10Results - Linear Distortion Measure11Results - Correlation12Conclusions✔Standard metrics does not give resultsconsistent with visual quality✔A framework for evaluation of SAR Images– Weighted Signal to Noise Ratio– Linear Distortion Measure– Distortion of edge information✔SPIHT outperforms
View Full Document