PROJECT TITLE :

Weighted Guided Image Filtering With Steering Kernel

ABSTRACT:

The guided image filter (GIF) is prone to halo artefacts at the margins because of its local characteristic. As a workaround, a weighted guided image filter (WGIF) with edge-aware weighting has recently been offered as an alternative. Edge-preserving performance is improved by combining the advantages of local and global operations. These guided filters do not, however, take into account the edge direction, a critical aspect of the guidance image. To get around this problem, we've developed a new version of GIF that makes better use of the edge direction. Our filter's behaviour can be improved by including adaptive learning results obtained from the steering kernel into the filtering process. With the proposed strategy, performance can be improved while edges are preserved and halo artefacts are reduced. Studies that included edge-aware smoothing, detail improvement, denoising and dehazing came to similar outcomes.