Image Sharpness Candidate Points Candidate Points are the points with the highest local greyscale variation in the region of interest. These points are candidates at which a sharpness measurement will be made if the local variation is sufficient. The number of candidate points is, by default, automatically set by the tool based on size of the region of interest. When the tool is executed, it first scans the region of interest and identifies many candidate points where the local standard deviation is the highest. It then evaluates the sharpness at each of the candidate location that has a local standard deviation above the Standard Devi- ation Threshold.
The locations where the sharpness is actually measured become Meas- urement Points. l Candidate points are, by default, set automatically by the tool. When the default, and recommended, Automatic setting is enabled, the tool uses 500 Candidate Points for a region of interest over 320x240 pixels in size. If the area is smaller than 320x240, the number of Candidate Points is equal to: width*height (500 / 320x240) l The number of candidate points can be set manually by entering a value for the Can- didate Point Count parameter.
Standard Deviation Threshold Standard Deviation Threshold sets the minimum standard deviation required for a Candidate Point to be used as a Measurement Point for calculating the average image sharpness. When the tool is executed, it scans the region of interest and identifies candidate locations where the local standard deviation is the highest. The number of candidate locations is set by Custom Candidate Points. Points having a standard deviation equal to or above the threshold are used by the tool as the measurement points for calculating the average image sharpness. Kernel Size Sets the size of the kernel of the operator for the sharpness process.
The default kernel size of 5 is usually appropriate in typical applications. However, the Kernel should be larger than the number of pixels over which a typical contrast is spread. A larger kernel may be used for blurrier images, for example, in the case where the blurriness is larger than the default kernel value. This is illustrated in the following figure, where the con- trast is about 6-8 pixels wide. Note the difference in values obtained with different kernel sizes. With a 7x7 kernel, all the candidate points are used as measurement points. Larger ker- nels subsequently have almost no impact on the Sharpness result. ACEUser's Guide, 3.6.x, 18316-000 Rev A Page 532