Combined Spatial Filtering and Boolean Operators Applied to the Processing of Real Images.

摘要

Several new and seemingly successful scene analysis techniques for application to 'real' image processing are presented and discussed. These techniques consist of particular combinations of spatial low pass filtering, global thresholding and Boolean operators, specifically the 'AND', 'OR' and 'NOT' operators. These combinatorial operators, hereafter referred to as the 'Boolpass' operators, perform the task of picture energy/information reduction, while retaining the fundamental picture primitives such as edges which characterize the images. Over 150 figures are included which illustrate the results obtained from application of the Boolpass techniques to eight different natural scenes. These results indicate that the Boolpass operators do display great potential as important components of a larger more comprehensive pattern recognition machine. Such a machine would encompass further processing (for target classification/recognition) of the resulting Boolpass operator information. (Author)