Performance prediction in radar image filtering and lossy compression. Part I: Filtering
Владимир Васильевич Лукин, Oleksii S. Rubel, Sergey Abramov, Alexander Zemliachenko · 2017 IEEE First Ukraine Conference on Electrical and Computer Engineering (UKRCON) · 2017
This paper analyzes possibilities and recently proposed methodology of performance prediction for denoising and lossy compression of radar images under assumption that they are corrupted by pure multiplicative noise. Methods based on discrete cosine transform in blocks are considered. Characteristics of the noise (speckle) are assumed a priori known or pre-estimated with high accuracy. They are taken into consideration in setting parameters of denoising techniques (thresholds, parameters of homomorphic transforms) and compression methods such as quantization step. It is shown that simple statistics of DCT coefficients in 8×8 pixel blocks can be used for prediction of improvement (reduction) of peak signal-to-noise ratio. Curves for prediction are obtained by regression into scatter-plots in off-line mode. Applicability of the proposed prediction approaches is confirmed by real-life examples.