An Evaluation Method Based on Image Texture and KL Divergence for SAR Image Quantization

Raw images obtained from Synthetic Aperture Radar (SAR) imaging processing typically exhibit a broad dynamic range (14-16 bits). Subsequent processing requires these images to be compressed and quantized into 8-bit grayscale images, which can result in the loss of detailed information. Consequently, when selecting quantization algorithms, it is essential to employ quantitative evaluation criteria to assess their performance. However, traditional evaluation methods designed for optical images are not suitable for SAR images due to their inherent characteristics, such as poor texture information and a concentrated grayscale distribution.In this paper, we propose a method that leverages the local texture information of an "ideal image" to construct the "distance" between the quantized image and the "ideal image," based on the Kullback-Leibler (KL) divergence. We also present experimental results demonstrating the numerical values obtained through this quantitative measurement algorithm.