Evaluation of compressed synthetic aperture radar imagery

Three widely employed image compression techniques (Discrete Cosine Transform, DCT; Block Truncation Code, BTC; and Vector Quantization, VQ) were applied to 23 high resolution synthetic aperture radar images (SAR). Image quality metrics (Impulse Response, IPR, and Contrast Ratio, CR) were applied to the original (uncompressed) and the processed (compressed/reconstructed) images. Image fidelity measurements (Signal-to-Noise Ratio, SNR, and Mean Square Error, MSE) were also made to compare the two image sets. A constant false alarm rate (CFAR) automatic target cuer (ATC) algorithm, employing the Target-to-Background Interference Ratio, TBIR, statistic, was used to quantify scene clutter. Twelve trained and experienced imagery analysts served as subject matter experts in assessing the image interpretability of the two image sets. The study demonstrated the sensitivity of the IPR measurement to compression technique. Neither of the image fidelity measures were found to be useful predictors of either image quality or image interpretability. Significant changes in the expected FAR and image interpretability ratings were, however, found to be caused by the compression process.<
>