Learning a Dilated Residual Network for SAR Image Despeckling

Qiang Zhang,Jie Li,Zhen Yang,Qiangqiang Yuan,Xiaoshuang Ma

doi:10.3390/rs10020196

Qiang Zhang, Jie Li + Show 3 more

Open Access

https://doi.org/10.3390/rs10020196

Copy DOI

Journal: Remote sensing	Publication Date: Jan 29, 2018
Citations: 153	License type: CC BY 4.0

Affiliation: Wuhan University, Anhui University

Abstract

In this paper, to break the limit of the traditional linear models for synthetic aperture radar (SAR) image despeckling, we propose a novel deep learning approach by learning a non-linear end-to-end mapping between the noisy and clean SAR images with a dilated residual network (SAR-DRN). SAR-DRN is based on dilated convolutions, which can both enlarge the receptive field and maintain the filter size and layer depth with a lightweight structure. In addition, skip connections and a residual learning strategy are added to the despeckling model to maintain the image details and reduce the vanishing gradient problem. Compared with the traditional despeckling methods, the proposed method shows a superior performance over the state-of-the-art methods in both quantitative and visual assessments, especially for strong speckle noise.

Highlights

A synthetic aperture radar (SAR) is a coherent imaging sensor, which can access a wide range of high-quality massive surface data
Skip connections are added to the despeckling model to maintain the image details and avoid the vanishing gradient problem
Rather than using log-transform [28] or modifying training loss function like [29], we propose a novel network for SAR image despeckling with a dilated residual network (SAR-DRN), which is trained in an end-to-end fashion using a combination of dilated convolutions and skip connections with a residual learning structure

Summary

Introduction

A synthetic aperture radar (SAR) is a coherent imaging sensor, which can access a wide range of high-quality massive surface data. For the purpose of removing the speckle noise of SAR images, scholars firstly proposed spatial linear filters such as the Lee filter [5], Kuan filter [6], and Frost filter [7] These methods usually assume that the image filtering result values have a linear relationship with the original image, through searching for a relevant combination of the central pixel intensity in a moving window with a mean intensity of the filter window. Due to the nature of local processing, the spatial linear filter methods often fail to integrally preserve edges and details, which exhibit the following deficiencies: (1) unable to preserve the average value, especially when the equivalent number of look (ENL) of the original SAR image is small; (2) the powerfully reflective specific targets like points and small surficial features are blurred or erased; and (3) speckle noise in dark scenes is not removed [8]

Methods

Results

Conclusion