NN-Based Prediction of Sentinel-1 SAR Image Filtering Efficiency

Abstract

Images acquired by synthetic aperture radars are degraded by speckle that prevents efficient extraction of useful information from radar remote sensing data. Filtering or despeckling is a tool often used to improve image quality. However, depending upon image and noise properties, the quality of improvement can vary. Besides, a quality can be characterized by different criteria or metrics, where visual quality metrics can be of value. For the case study of discrete cosine transform (DCT)based filtering, we show that improvement of radar image quality due to denoising can be predicted in a simple and fast way, especially if one deals with particular type of radar data such as images acquired by Sentinel-1. Our approach is based on application of a trained neural network that, in general, might have a different number of inputs (features). We propose a set of features describing image and noise statistics from different viewpoints. From this set, that contains 28 features, we analyze different subsets and show that a subset of the 13 most important and informative features leads to a very accurate prediction. Test image generation and network training peculiarities are discussed. The trained neural network is then tested using different verification strategies. The results of the network application to test and real-life radar images are presented, demonstrating good performance for a wide set of quality metrics.

Reduction of Spatially Correlated Speckle in Textured SAR Images

“…Obviously, this should be metrics applicable for grayscale images and possessing good adequateness with respect to peculiarities of human vision system (HVS). In this sense, we can further employ earlier experience [22,26] concerning visual quality metrics [27][28][29][30] and their use in SAR image processing [26].…”

“…Then, to get quantitative evaluation of speckle suppression efficiency, we need means for modelling speckle with the determined properties. Methods to do this have been earlier described in our papers [23,26]. In Fig.…”

“…4, we are more interested in special kinds of textural images. This is because it is known that the task of improving image quality due to denoising is the hardest just for highly textured images [18,26]. One way to simulate them is to apply fractals.…”

“…that characterizes dependence of the local variance on local mean; then, for pure multiplicative noise one has For the modified DCT-based filter[13,26] adapted to speckle spatial correlation, one has frequency…”

Optimal parameters selection of the Frost filter based on despeckling efficiency prediction for Sentinel SAR images

“…In this paper we consider the Frost filter [7] since it is still widely used in various toolboxes like SNAP [11]. Our previous studies have shown that it is possible to predict despeckling efficiency beforehand with high accuracy for several filters [12]. We have already demonstrated how to select the window size for the well-known local statistics Lee filter based on despeckling efficiency prediction [10].…”