PROJECT TITLE :

Anisotropic Guided Filtering

ABSTRACT:

They've been widely used in numerous Image Processing and computer vision applications because of their low complexity as well as good edge-preservation capabilities. "Detail halos" appear when the different forms of the guided filter are unable to deal with increasingly aggressive filtering strengths, despite this success. These current filters, on the other hand, function badly when the input and guide images are structurally inconsistent. For example, the guided filter operates as a variable-strength locally-isotropic filter that, in fact, serves as a weak anisotropic filter on the picture, as demonstrated in this paper. When guided filter versions such as the adaptive guided filter (AGF), weighted guided image filter (WGIF), and gradient-domain-guided-image filter (GD-GIF) are used, they tend to exhibit this behaviour (GGIF). AnisGF (Anisotropic Guided Filter) is a new filter that uses weighted averaging to maximise diffusion while maintaining sharp image edges. In order to provide strong anisotropic filtering while keeping low computational costs, the proposed weights have been optimised based on local neighbourhood variances. The results of synthetic experiments reveal that the suggested method tackles the issue of detail haloes and the management of inconsistent structures reported in prior iterations of the guided filter.. Scale-aware filtering, texture removal and upsampling of chroma all highlight the benefits of the technique in practise.