Edge Detection and Extraction for SAR Images

Edge information is useful for various remote sensing applications : classification, relief reconstruction, image analysis. A lot of methods have been proposed to process edge extraction, with respect to the hypothesis of additive noise. All these methods are usually based on differential operators followed by a non local-maxima suppression or a zero-crossing search. In SAR imagery, presence of speckle noise makes derivative methods not applicable because of non constant false alarm results. Usually, ratio operators are used for SAR image processing. In this paper, we propose a new edge detector based on a ratio. This operator has the particularity to deliver signed results. Then we apply two edge extraction methods : the first one based on a non local-maxima suppression, and the second on a non zero-crossing suppression. INTRODUCTION Structural information is necessary for various applications (classification, image analysis, relief reconstruction), and edges are the main part of this information. Edges are usually modeled as an intensity step. In optical imagery, derivative methods have been used for a long time to extract edges. A classical method is : Computation of the gradient of the image Extraction of local maxima along the gradient direction. Hysteresis thresholding. Second order derivatives can also be used. In this case, the second step would be the extraction of zero-crossing along the gradient direction. In SAR imagery, this kind of techniques can not be applied because speckle noise is multiplicative. Ratio operators are mainly used [1]. In the first part of this paper, we propose a new ratio operator, which is signed. In the second part, we use the sign of this operator to compute a non local-maximum, and a zero-crossing extraction along the direction of the ratio (perpendicularly to the edge). We finally show first results of detection and extraction methods. I. EDGE DETECTION In this section, we first remind that differential operators provide a non constant false alarm ratio when used on SAR images, and that ratio operators permit to solve this drawback. Then we propose a new ratio-based operator. A. Difference Of Averages Let us consider the difference of average (d) , which is the most commonly used detector in optical imagery :