Abstract: The realization and application of a windowless nonlinear image filter are studied. In the windowless nonlinear image filter, an input gray-scale image is transformed into a group of binary parts by threshold decomposition, and then nonlinear windowless filter operation is performed on these binary images. As it is usually very easy to segment image structures in a binary image, so we can handle the image based on meaningful image structures instead of distinct image pixels. But the original windowless filter algorithm needs very heavy computation, so it is unpractical. To overcome this drawback, two solutions are given: 1)by reducing the amount of the binary images of threshold decomposition, one can cut down a lot of computation of threshold decomposition and reconstruction; 2)in image structure analysis, instead of the binary images, the contour lines are used, so the complex threshold decomposition and reconstruction process are bypassed. As a real application, the results of car license plate location by the new filter are reported.

Key words: image filtering, nonlinear filtering, image window, license plate location