SAR Image Despeckling Based on Convolutional Denoising Autoencoder

TL;DR

This paper introduces a convolutional denoising autoencoder approach for SAR image despeckling, which learns directly from corrupted images and outperforms some existing methods in quality metrics.

Contribution

The paper proposes a novel SAR despeckling method using C-DAE with batch normalization, effective on limited datasets, improving over traditional techniques.

Findings

Outperforms some existing despeckling methods in PSNR and SSIM.

Efficient training achieved with batch normalization.

Effective despeckling demonstrated on SAR images.

Abstract

In Synthetic Aperture Radar (SAR) imaging, despeckling is very important for image analysis,whereas speckle is known as a kind of multiplicative noise caused by the coherent imaging system. During the past three decades, various algorithms have been proposed to denoise the SAR image. Generally, the BM3D is considered as the state of art technique to despeckle the speckle noise with excellent performance. More recently, deep learning make a success in image denoising and achieved a improvement over conventional method where large train dataset is required. Unlike most of the images SAR image despeckling approach, the proposed approach learns the speckle from corrupted images directly. In this paper, the limited scale of dataset make a efficient exploration by using convolutioal denoising autoencoder (C-DAE) to reconstruct the speckle-free SAR images. Batch normalization strategy is integrated with C- DAE to speed up the train time. Moreover, we compute image quality in standard metrics, PSNR and SSIM. It is revealed that our approach perform well than some others.