Prism Utilities These are a few examples of utility files provided with Prism. Three of the various types of files are shown here. Prism can also load palettes or "color look up tables" such as those that can be used by NIH Image. ----------------- The kernel files included here are also compatible with NIH Image. They provide Gaussian smoothing with various standard deviations (given in pixels in the file names, and ranging from 0.4 pixels to 10.5 pixels). In addition, the "DOG" file directly calculates the difference of Gaussians, used in edge finding operations. ----------------- The PCMD files are separately compiled image processing drop-ins that can perform a variety of advanced processing and measurement operations, either to an entire image or to a region of interest. The ones provided here include: 3D_Plot: Draws an isometric plot of the pixel values displayed as elevation. Holding down the option key suppresses the autoscaling, and pressing the shift key inverts the values. This PCMD requires two images (not necessarily of the same size). The second image, with the 'checked' name, is not affected. The frontmost window is erased and has the plot drawn in it. ASTM_GS: A very fast intercept-length measurement procedure for determining the ASTM Grain size from an image with incomplete grain boundaries, and without the need for skeletonization to thin the lines. The image is expected to be thresholded to binary, showing dark boundaries and white grains. Blend: Adds together two images in any proportion. The second (checked) image is not affected. Contours: Draws multiple contour lines at equi-spaced brightness levels, each in their appropriate grey scale or color value. Frei&Chen: An advanced edge-finding algorithm that responds to the "Edgeness" of pixels irrespective of the absolute value of the grey scale. Grid: Draws a grid of vertical and horizontal lines on the image (user specifies the number of lines or their spacing). Haralick: An advanced texture-extraction operator, that responds to local pixel diferences to produce a grey scale image that can be thresholded. HitOrMiss: A flexible binary image erosion/dilation routine that allows specifying any particular pattern of nearest neighbors as a target. Hough: An implementation of the linear Hough transform for finding linear alignments of features. The forward Hough creates the transform in the frontmost image from the binary features in the checked image. The reverse Hough (depress the option or shift keys) draws the lines corresponding to black points in the checked image onto the frontmost image. The normal use sequence would be to threshold the original image, check it, create a new window, call the forward Hough, examine it and mark the maximum points using a black pencil, check the image, select the original image and call the reverse Hough to see the lines superimposed on the features. Hurst: An advanced texture-extraction operator that computes the local fractal dimension of each pixel in the image and creates a new grey scale image that can be thresholded. HybridMedian: An advanced median filter that preserves edges and corners. LocalEqualize: Expands local contrast by performing a histogram equalization in an 11 pixel wide circle around each pixel in the image, applying the result to just the central pixel. Mask: Copies pixels from the checked image into only those pixels in the original image that are set to black (or, with the shift key depressed, white). Ramp: Creates a brightness grey scale across the top of the image, useful for photographing or printing. RandPoints: Marks random points on the image (number specified by the user). RandValues: Fills the image with random grey values. If the Option key is depressed, the values are Gaussian random instead of uniform random. Rank: Applies a ranking operation to a 5 pixel wide circle centered on each pixel in the image. It can create a new image with the brightest or darkest pixel, the median, the range, the variance or standard deviation, or maximize contrast using the mode. This is useful in various situations including background levelling by removing small features, responding to edges or texture, etc. ReduceResn: Averages pixels and/or divides down the brightness values to show the effect of reduced pixel or grey scale resolution. TopHat: Locates features smaller than 3 pixels and locally brighter or darker than the surroundings, either to mark (keep) the features, or to eliminate them by replacement with the local median, in which case this is more commonly called a rolling ball filter. ----------------- The Transform files allow the direct measurement of size, shape, brightness and position values not provided in the already extensive built-in list of measurement parameters. A utility for creating these transforms from measured values is provided with the program, along with dozens of examples. These may be loaded and used for ranking, etc., just as any other parameter is selected. This capability is also used with the regression/principal components routines in the data analysis program provided as part of the package to create feature recognition scripts that perform identification and classification. The examples provided here illustrate a few of the possibilities: Indentation Depth (the mean depth of bays around the feature periphery, measured from the bounding polygon or "taut string" perimeter). Mean Intercept Length (a stereological parameter for features). Modification Ratio (the ratio of the inscribed to the circumscribed circle fitted to the feature outline, used primarily in textile quality control). Brightness Range (the difference between the darkest and brightest pixel within a feature).