Numerous remote sensing applications require the detection of elements structuring the image. Among these, shiny targets or points, contours and lines are preeminent. In fact, strong reflectors corresponding to objects are often manmade; lines can be described in numerous themes, whether they are the road and hydrographic networks or geological structures. As for contours, they make dividing the image into areas of interest possible, such as parcel plans, or they characterize regions in relation to their density, such as in urban areas. This chapter describes the techniques making the detection of these three types of primitive elements on radar images possible.

Despite its strong image readability degradation, speckle is interesting because it is statistically well modeled, as we have seen in Chapter 5. This aspect is important and the majority of methods presented in this chapter are based on statistical analysis for the development of detection criteria. Thus, considering a measure m, which indicates the presence, or not, of a target, a contour or a line, we will try to establish the conditional distribution of m according to magnitudes implied: contrast between target and background, or contrast on one side and the other of the contour, or size of the neighborhood used. Then, we will be able to predict, according to the threshold values on m, detector performances in terms of detection probability P _d and false alarm P _fa. When it is not possible to obtain an analytical expression of the distribution, simulations may...