Wavelets work on volumetric data
Wavelets are finding increasing use in analyzing, encoding, and modeling two-dimensional images and signals (see Vision Systems Design, Sept. 1996, p. 52). Now, Kara Kytle, a graduate student at Harvard University (Cambridge, MA), has developed software that allows wavelet transforms to be performed on 3-D data.
As an extension to the Vox-L visualizer from Vox-L (Woburn, MA), the WaveApp package uses the Vox-L Application Programmers Interface (API). WaveApp allows wavelet transformations to be performed in two or three dimensions on any dataset.
At present, the package is optimized for image edge detection. This is achieved in two steps. First, a forward wavelet transform breaks down original image data into three high-frequency components representing regions of high horizontal, vertical, and diagonal content. These three components contain information about the edges of the image. Thresholding is then applied to this high-frequency data to convert the information to binary form. In the second step, processed data are subjected to an inverse wavelet transform and re mapped over the original image, allowing visual evaluation of the edges.
To help developers achieve the best edge detection possible, Wave App provides an experimental mode that allows the decomposition and reconstruction steps to be performed separately. By doing so, the results of the forward transform and the effectiveness of different filters and the degree of enhancement that produces the most appropriate visual results can be evaluated.
After these parameters have been established, WaveApp can be run in an automatic mode. This is useful for volume transformations where entire volumes can be processed by iterative transforms of two-dimensional slices of data. For more information contact fringe_child@ dataspace.com.
