Detecting Edges

Detecting edges and region boundaries is a fundamental requirement for automated image interpretation. A segmentation algorithm divides up an image into regions. An edge detector finds lines which separate areas with different characteristics: these edges do not necessarily join up to form boundaries of closed regions.

An edge-map is a binary image in which each pixel indicates whether or not an edge is present. It can be produced directly from the raw data by an edge detection algorithm, or (usually much more reliably) by identifying the region boundaries in a segmentation.

The scanedge routine

The scanedge function detects edges in a raw SAR image via a combination of a scanning window and a scanning edge. The algorithm used has been shown theoretically [2] to give optimal detection of edges in a raw SAR image, for the window size specified. The sensitivity of the detector can be altered by setting the probability of false alarm (pfa) -- the probability of finding edges where none exist. Smaller pfas lead to fewer edges being found.

The scanedge algorithm can be performed upon images with more than one image layer. scanedge is accessed via the filter function: use method=''scan''. Full Python script and function syntaxes for cfar are given in section A.4.10.

Figure 4.12 shows a raw DLR ESAR amplitude image, and edge-maps produced from it for several combinations of filter size and pfa. It is clear that the dominating parameter is the filter size: we recommend using a size of 7 by 7.

The edge routine

Alternatively, by using the edge utility, you can produce an edge-map of the region boundaries from a segmentation. An edge-map overlaid on its corresponding cartoon may show up region boundaries you would otherwise overlook: two adjacent regions may have similar intensities.

Do not use scanedge to convert a cartoon image, the output of a segmentation function, into an edge-map. Use edge to perform this task.

An example of the use of edge is shown in Figure 4.12(f). The edge algorithm can be performed upon images with more than one image layer. Full Python script and function syntaxes for edge are given in section A.4.5.

Figure 4.12: (a) Raw ESAR Image (b) scanedge output with filter size 3 by 3 and pfa $10^{-5}$ (c) 7 by 7 (d) 10 by 10 (e) 7 by 7 with pfa $10^{-7}$ (f) Edges found from the segmented image using scan

Figure 4.12(a) Figure 4.12(b)

Figure 4.12(c) Figure 4.12(d)

Figure 4.12(e) Figure 4.12(f)