Abstract

An adaptive algorithm for the enhancement of patterns consisting of families of curved lines was implemented. The algorithm was designed to simulate the human visual system’s ability to detect and follow such patterns even in adverse conditions. Sections of a striated pattern that are not perceivable alone may be detected if viewed contiguous to suprathreshold parts of the same pattern. The algorithm implements filters that are well localized in both the space and spatial frequency domains. These bandpass, orientationsensitive fan filters are similar in point spread functions to responses recorded from visual cortical cells. The filters are used both to detect the orientation of the striation in a small window and to enhance the image in that orientation. To speed up calculations and reduce error rates, the search for orientation is limited to a small angular range based on orientations previously found in neighboring, overlapping windows. The results of application of the algorithm to low-contrast striated patterns and to noisy images are presented. Difficulties and solutions to the problems encountered in windows with multiple orientations are discussed.

© 1986 Optical Society of America