Cơ sở dữ liệu hình ảnh - Chương 4

4. SEGMENTATION AND EDGE DETECTION4.1 Region OperationsDiscovering regions can be a very simple exercise, as illustrated in 4.1.1. However, more often than not, regions are required that cover a substantial area of the scene rather than a small
Nội dung trích xuất từ tài liệu:
Cơ sở dữ liệu hình ảnh - Chương 44. SEGMENTATION AND EDGE DETECTION4.1 Region OperationsDiscovering regions can be a very simple exercise, as illustrated in 4.1.1. However, moreoften than not, regions are required that cover a substantial area of the scene rather than asmall group of pixels.4.1.1 Crude edge detectionUSE. To reconsider an image as a set of regions.OPERATION. There is no operation involved here. The regions are simply identified ascontaining pixels of the same gray level, the boundaries of the regions (contours) are at thecracks between the pixels rather than at pixel positions.Such as a region detection may give far for many regions to be useful (unless the numberof gray levels is relatively small). So a simple approach is to group pixels into ranges ofnear values (quantizing or bunching). The ranges can be considering the image histogramin order to identify good bunching for region purposes results in a merging of regionsbased overall gray-level statistics rather than on gray levels of pixels that aregeographically near one another.4.1.2 Region mergingIt is often useful to do the rough gray-level split and then to perform some techniques onthe cracks between the regions – not to enhance edges but to identify when whole regionsare worth combining – thus reducing the number of regions from the crude regiondetection above.USE. Reduce number of regions, combining fragmented regions, determining whichregions are really part of the same area.OPERATION. Let s be crack difference, i.e. the absolute difference in gray levels betweentwo adjacent (above, below, left, right) pixels. Then give the threshold value T, we canidentify, for each crack 1, if s < T w= 0, otherwisei.e. w is 1 if the crack is below the threshold (suggesting that the regions are likely to bethe same), or 0 if it is above the threshold.Now measure the full length of the boundary of each of the region that meet at the crack.These will be b1 and b2 respectively. Sum the w values that are along the length of thecrack between the regions and calculate: ∑w min ( b1 ,b2 )If this is greater than a further threshold, deduce that the two regions should be joined.Effectively this is taking the number of cracks that suggest that the regions should bemerged and dividing by the smallest region boundary. Of course a particularly irregularshape may have a very long region boundary with a small area. In that case it may bepreferable to measure areas (count how many pixels there are in them).Measuring both boundaries is better than dividing by the boundary length between tworegions as it takes into account the size of the regions involved. If one region is very small,then it will be added to a larger region, whereas if both regions are large, then the evidencefor combining them has to be much stronger.4.1.3 Region splittingJust as it is possible to start from many regions and merge them into fewer, large regions.It is also possible to consider the image as one region and split it into more and moreregions. One way of doing this is to examine the gray level histograms. If the image is incolor, better results can be obtained by the examination of the three color valuehistograms.USE. Subdivide sensibly an image or part of an image into regions of similar type.OPERATION. Identify significant peaks in the gray-level histogram and look in thevalleys between the peaks for possible threshold values. Some peaks will be moresubstantial than others: find splits between the best peaks first.Regions are identified as containing gray-levels between the thresholds. With colorimages, there are three histograms to choose from. The algorithm halts when no peak issignificant.LIMITATION. This technique relies on the overall histogram giving good guidance as tosensible regions. If the image is a chessboard, then the region splitting works nicely. If theimage is of 16 chessboard well spaced apart on a white background sheet, then instead ofidentifying 17 regions, one for each chessboard and one for the background, it identifies16 x 32 black squares, which is probably not what we wanted.4.2 Basic Edge DetectionThe edges of an image hold much information in that image. The edges tell where objectsare, their shape and size, and something about their texture. An edge is where the intensityof an image moves from a low value to a high value or vice versa.There are numerous applications for edge detection, which is often used for variousspecial effects. Digital artists use it to create dazzling image outlines. The output of anedge detector can be added back to an original image to enhance the edges.Edge detection is often the first step in image segmentation. Image segmentation, a field ofimage analysis, is used to group pixels into regions to determine an images composition.A common example of image segmentation is the magic wand tool in photo editingsoftware. This tool allows the user to select a pixel in an image. The software then draws aborder around the pixels of similar value. The user may select a pixel in a sky region andthe magic wand would draw a border around the complete sky region in the image. Theuser may then edit the color of the sky without worrying about altering the color of themountains or whatever else may be in the image.Edge detection is also used in image registration. Image registration aligns two images thatmay have been acquired at separate times or from different sensors. roof edge line edge step edge ramp edge Figure 4.1 Different edge profiles.There is an infinite number of edge orientations, widths and shapes (Figure 4.1). Someedges are straight while others are curved with varying radii. There are many edgedetection techniques to go with all these edg ...