IMcProfileEdgesEdgeFindingMethod Property

Property Value

Type: mcProfileEdgeFindingMethod

In its weighting phase, FindEdges uses one of two general methods for analyzing a profile to generate a weighted profile (see FindEdges for an overview discussion). The two methods are a weighted average method and a correlation method. The four flavors of the weighted average approach will find luminance peaks, valleys, rising edges or falling edges (the corresponding EdgeFindingMethod values are mcpefmPeaks, mcpefmValleys, mcpefmRisingEdges and mcpefmFallingEdges). The other method, specified by the mcpefmMatchPattern EdgeFindingMethod, will detect good matches between the luminance profile and a supplied luminance pattern. The weighted average methods are simpler and generally work well for profiles with sharp luminance transitions (for mcpefmRisingEdges or mcpefmFallingEdges) or with narrow luminance peaks or valleys (for mcpefmPeaks or mcpefmValleys). Two properties affect these weighted average methods: Weights and MatchLength. The defaults for these usually work adequately: the MatchLength is 5 pixels while the Weights are 0.6, 0.8, 1.0, 0.8, 0.6 (that is, the center point gets full weight and the profile samples on either side are given slightly less weight). The Weight property array is automatically mapped onto the MatchLength, so MatchLength could be changed to 7 for example without needing to change the Weights property (this would favor more gradual transitions). The mcpefmMatchPattern EdgeFindingMethod, performs a correlation between the luminance profile and a supplied luminance pattern to come up with the weighted profile. The Pattern property specifies the shape of the luminance pattern being sought, while the MatchLength property specifies its length (the Pattern property array is stretched or shrunk as necessary to map onto the MatchLength). The SizeWeighting property is a value between 0 and 100 that controls the degree to which profile amplitude changes are considered in the weighting (0 is a pure correlation, profil amplitude does not matter, only profile shape; 100 is a pure convolution, larger profile amplitude yields larger weighted result). SizeWeighting is 50 by default. In addition, two of the WeightingFlags can be turned on to force an even closer match to the supplied Pattern: mcpewfPointsForSameSize will cause profile patterns that are the same amplitude as the Pattern property to be favored over either smaller or larger amplitudes, and mcpewfPointsForSameOffset will cause profile patterns that are the same mean luminanance as the Pattern property's mean luminance to be favored over either dimmer or brighter sections of the pattern. Both flags are off by default; you may set one or both of them. When EdgeFindingMethod is assigned mcpefmMatchPattern, then the MatchLength property is forced to be at least 2 (MatchLength is allowed to be 1 for the weighted averaging methods, but it must be 2 or more for the mcpefmMatchPattern method). When EdgeFindingMethod is assigned either mcpefmPeaks or mcpefmValleys, then the OptionFlags mcpeofAbsoluteThresholding bit is cleared. This is done because thresholding against the AbsoluteMaxWeightValue is rarely appropriate for these weighted averaging methods.