Abstract

When an image is very noisy and/or has very low contrast, it is difficult to detect the features of interest. The adaptive wedge filtering algorithm is capable of extracting linear features (lines and edges) from a noisy image when no other common edge detector will work. First, the unprocessed image is divided into several subimages, where every subimage is overlapped with its neighbors. Then, several wedge filters are used to detect and modify the Fourier spectra of linear features in each subimage. Artifacts may be produced by wedge filters, but these artifacts can be sup pressed by introducing a threshold for the detection of significant spectral energies in each wedge filter. The final enhanced image is reconstructed from each enhanced subimage. Several examples are presented to show the results of enhancing different kinds of image.

© 1986 Optical Society of America