List of selected sample programs

• visualize the structure tensor of an image (compile and run app/structureTensorComputation.C)
  This small program will take an image as input, compute its structure tensor, and display the result on screen.

• visualize application of the various filters used to discern image textures in segmentation (compile and run app/filter2DTests.C)
  This small program will take an image as input, convert it to greyscale, and compute and visualize some filtered versions of it that are useful for texture discrimination when segmenting images. These are: structure tensor, anisotropically smoothed image (total variation variant) and scale measure.

• generate a stack of 2D images that the library can read as a 3D volume (compile and run app/generate_stack.C)
  This small utility program takes a list of image filenames, read the corresponding image files and generates a "stack" of images that it writes to file in its own format. A stack can be read into an Image as a 3D image, where each of the images in the stack constitutes one "z-slice" of the 3D volume.

• see the development of an image as TV-flow (total-variation-diminishing flow) is applied to it (compile and run app/tvfilter.C)
  This program will demonstrate the effect of total-variation-diminishing flow on an image. Note the segmentation-like effect obtained.

• see the effect of line integral convolution on a 2D image
  (compile and run app/LIC_test.C)
  This small program demonstrates the effect of applying line integral convolution of a vector field and a Gaussian kernel function on an image consisting of random noise. Note how this leads to an intuitive visualization of the vector field.

• see the effect of line integral convolution on a 3D image
  (compile and run app/field_smooth.C)
  This small program demonstrates the effect of applying line integral convolution of a Gaussian kernel function and a 3D vector field generated by a dipole on a 3D volume consisting of random noise. Note how this leads to an intuitive visualization of the vector field.

• random partitioning of a rectangular domain into a given number of fragmented regions (compile and run app/randomVoronoiTest.C)
  Small program demonstrating how the function random_scattered_voronoi() can be used to generate a random partitioning of a region into a given number of fragmented subregions. This is potentially useful for initializing an automatic segmentation process with multiple regions.

• 2D segmentation with 2 or more regions regions (compile and run app/segmentation2D.C) This program carries out segmentation on a user-specified image according to various options that the user can define on the command line. Run the program without arguments to get a list of arguments to use. The result from a segmentation can be saved afterwards. Masks can also be used, defining which part(s) of the image that should participate in the segmentation process.

• anisotropic, curvature-preserving smoothing of images (compile and run app/smoothcurvpres.C)
  This is a 2D/3D implementation of the greycstoration algorithm described by David Tschumperle in his 2006 paper ''Fast Anisotropic Smoothing of Multi-Valued Images using Curvature-Preserving PDE's''. In essence, it is a curvature-preserving anisotropic smoothing algorithm using line integral convolutions to smooth an image along natural flow lines (edges, etc.) The results can be quite remarkable. Try to play around with the parameters and see what it does. An explanation of the parameters is found in the paper linked to above. The algorithm takes as input a picture in any common image format, or a stack of images, which can be generated by the utility program generate_stack.

---