Spatial filtering for speckle reduction, contrast enhancement, and texture analysis of GLORIA images

This paper reports a comparative study of digital enhancement techniques using spatial filtering to improve the geologic interpretation of side-scan sonar GLORIA images. Seven algorithms for speckle reduction with window sizes of 3/spl times/3-7/spl times/7 pixel and various numbers of iterations were tested for cosmetic purposes, and also to improve subsequent image processing. The filtered images were evaluated using both quantitative and qualitative techniques. It was determined that a normalized inverse gradient weighted smoothing scheme, with a 3/spl times/3 pixel filter and five iterations, allows a significant speckle reduction without blurring the edges in the GLORIA image which correspond to geological structures. Three local contrast enhancement techniques were also tested and evaluated to increase the perception of these geologic structures. Subtracting the gradient magnitude twice, calculated with spatial filters of a 5/spl times/5 pixel on smoothed images, was found to enhance most GLORIA images. Texture analysis methods developed for GLORIA images of mid-oceanic ridges and based on edge detection and orientation determination by spatial filtering are also presented. It enables the GLORIA mosaic of the Rodriguez triple junction (Indian Ocean) to be partitioned into regions of preferred orientation corresponding to the different seafloor fabrics generated at each arm of the triple junction.<
>