Multi-dimensional signal-processing research program

In the area of image segmentation and classification, a hierarchical segmentation scheme was developed for processing images with several region classes. Adaptive contrast enhancement was used as a preprocessor for image segmentation based on texture rather than gray level. These techniques were also applied to aerial images degraded by light cloud cover and haze. The primary effects of this type of degradation are a reduction in contrast and an increase in the local average intensity of the image. Two aspects of the iterative implementation of multi-dimensional digital filters are also discussed. This first concerns spatial truncation effects which occur during the iteration because the image frame buffer has a finite storage capability. The errors caused by this truncation are closely related to the solution of a boundary value problem with boundary conditions specified on the frame edges. These errors can be eliminated by including the boundary conditions in the spatially truncated iteration. Potential multiprocessor architectures are discussed.