Improvement of Object Detection from SAR Image Using Speckle Filter

In this paper, a convolutional neural network (CNN) based SAR image classifier is proposed. Synthetic aperture radar (SAR) images are often contaminated by speckle noise. Therefore, in the proposed method, a speckle filter is placed as a pre-process of the classifier. The network was trained after the speckle noise removal from SAR images. This approach helps the classifier to improve the performance. From an experimental evaluation, the significance of the proposed method is verified.