Improved Bilateral Texture Filtering with Edge-aware Measurement.

Area of Science:

• Computer Vision
• Image Processing
• Computational Imaging

Background:

• Current texture filtering methods rely on axis-aligned windows for texture measurement, often leading to reduced quality due to misalignment with structure edges.
• Existing techniques analyze pixel-level intensity contrast, neglecting the linear characteristics of structure edges, which limits measurement effectiveness.
• This results in a trade-off where small windows fail to smooth large textures, and large windows remove small structures.

Purpose of the Study:

• To develop novel edge-aware measures for enhancing texture measurement quality in image filtering.
• To address the limitations of existing texture filtering methods that struggle to differentiate between large textures and small structures.
• To improve the accuracy and effectiveness of texture filtering algorithms.

Main Methods:

• Constructing edge-aware windows that are optimally contained within texture regions.
• Incorporating the linear characteristics of structure edges into the texture measurement process.
• Employing large box windows for overall texture filtering and narrow edge-aware windows for precise texture measurement.

Main Results:

• The proposed edge-aware measures significantly improve texture measurement quality.
• The method successfully filters large-scale textures while preserving fine structures, resolving the existing dilemma.
• Experimental results demonstrate superior performance compared to conventional texture filtering techniques.

Conclusions:

• Edge-aware measures offer a more robust approach to texture measurement for image filtering.
• The developed technique effectively balances the smoothing of large textures with the preservation of small structures.
• This advancement provides a more effective solution for texture filtering applications in image processing.