No abstract

We appreciate very much the constructive comments from the reviewers, which have had helped us to revise and improve the paper greatly. One-to-one responses and corresponding revisions are given below.

This paper proposes a generalized modeling and 1 simulation approach for correlated SAR texture based on the 2 Gaussian coherent scatterer model. It is rooted in the physics-3 based coherent scatterer assumption where each observation 4 in a SAR image is a coherent sum of multiple underlying 5 Gaussian scatterers. The proposal generalizes existing single-6 point statistical models by allowing the number of scatterers 7 to be a correlated random field. It can also generate the desired 8 spatial correlation texture by stipulating the structure in both 9 the Gaussian scattered field and the number of scatterers. This 10 generalized model is derived theoretically and then validated by 11 both simulations and experiments with SAR data from actual 12 sensors. 13 Index Terms-SAR statistical modeling, Coherent scatterer 14 model, Correlated SAR texture. 15 I. INTRODUCTION 16 S TATISTICAL modeling of SAR images is the theoretical 17 basis of SAR image processing and interpretation: image 18 filtering, image classification, object detection, and other oper-19 ations rely on tractable and expressive models. A good statis-20 tical model should be an effective and efficient representation 21 of the often complicated SAR data. It mainly involves two 22 aspects: the single-point distribution of image pixel intensities, 23 and two-point statistical relations, i.e. the correlation function, 24 which characterizes image texture information. 25 The mainstream models of SAR image statistical modeling 26 are reviewed in [1]. It summarizes more than twenty single-27 point distributions [2]-[13] for different physical scenes, as 28 well as some correlation function models [6], [14]-[18]. It is 29 found that the existing statistical models have three drawbacks, 30 namely, 31 1) many of them are phenomenological methods which do 32 not take into account the actual physical process of 33 scattering; 34 2) each statistical model is usually suitable for only one type 35 of specific scenario of terrain surface e.g. the Rayleigh 36 distribution for homogeneous areas [14], the K distri-37 bution for intermediate inhomogeneous area [4], the G 0 38 Manuscript received XXX; revised XXX.

No abstract

No abstract