Texture Study of Synthetic Aperture Radar (SAR) Images of Ocean Surfaces

摘要

This report describes the first phase of a study whose objective is to establish the relation between the texture in SAR images of the ocean surface and the relevant physical conditions giving rise to this texture. The initial concern is the selection of a technique which leads to a reliable separation of images into regions of different texture. Section II of the report discusses the role of texture in image analysis. Section III introduces the concept of a feature and establishes rationale for the selection of gray level co-occurrence matrix approach to feature extraction. Section IV describes the proposed classification algorithm and Section V gives the results for several test images. It is found that this approach works will in classification of land scenes. For ocean images which exhibit only subtle differences between regions the approach's discriminating power is less satisfactory than for land scenes. In Section VI alternate approaches are suggested for ocean images.(Keywords include: Texture, Co-occurrence, Probability matrix, Synthetic aperture radar, and Ocean surface.