Image edge-preserving filtering method based on adaptive neighborhood shape

Abstract

The invention discloses an image edge-preserving filtering method based on an adaptive neighborhood shape. The method is characterized by, based on line shift, drawing lines, the length of which is k,in different directions on each pixel point, wherein the lines are in radial movement in their respective directions, and it must be ensured that each of such shift lines comprises a current pixel point; finding the cleanest line in shift lines in each direction; then, combining the cleanest lines in different directions to form a clean neighborhood of the current pixel point; and according to cleanliness of the lines, calculating weighting coefficient of the lines in forming filtering guide value of the current pixel point, wherein line length directly controls the scale of filtering, that is, texture and noise under such scale are filtered out. The method has a function of filtering texture and large-scale noise interference in the image, can keep significant meaningful edges of the image clear, and can be widely applied to image detail enhancement, image cartoonization, and image target recognition and the like.

Description

technical field [0001] The invention relates to an image edge-preserving filtering method, in particular to an image edge-preserving filtering method based on an adaptive neighborhood shape. Background technique [0002] Image edge-preserving filtering is a very critical preprocessing step in many computer vision application systems. It can filter out the interference of texture and noise in the image, while keeping the significant and meaningful edges in the image clear, for subsequent visual object segmentation, semantic Provide necessary technical support for analysis and three-dimensional space reconstruction. [0003] When the change scale of texture or noise in the image is relatively small, the bilateral filtering method (C.Tomasi, R.Manduchi, "Bilateral filtering for gray and color images," in Proc.of ICCV, 1998) or isotropic based on partial differential equations The diffusion filter method (L.Rudin, S.Osher and E.Fatemi, "Nonlinear total variation based noise rem...

AI Technical Summary

Problems solved by technology

But this method will have problems when the pixels are located in a narrow strip or inside a sharp corner, because it is difficult to find a "clean" image block around these pixels, which will blur the narrow strip and sharp corner structure in the image

Images