Adaptive bilateral filter for sharpness enhancement and noise removal

Area of Science:

• Image Processing
• Computer Vision

Background:

• Traditional sharpness enhancement methods like unsharp mask (USM) can introduce artifacts such as overshoot and undershoot.
• Existing filters often struggle to simultaneously enhance edges and remove noise effectively.

Purpose of the Study:

• To introduce the adaptive bilateral filter (ABF) for superior sharpness enhancement and noise reduction.
• To present a novel approach to edge slope restoration without edge detection.

Main Methods:

• The adaptive bilateral filter (ABF) enhances edge slopes by transforming the histogram using a range filter with adaptive parameters.
• ABF parameters are optimized through a training procedure.
• The method avoids explicit edge detection, orientation analysis, or profile extraction.

Main Results:

• ABF significantly sharpens images and removes noise, outperforming the standard bilateral filter.
• Compared to the optimal unsharp mask (OUM), ABF achieves comparable edge sharpness without halo artifacts.
• ABF demonstrates superior noise removal capabilities compared to both the bilateral filter and OUM.

Conclusions:

• The adaptive bilateral filter (ABF) offers an effective solution for simultaneous sharpness enhancement and noise removal.
• ABF provides artifact-free sharpening and robust noise reduction for diverse image types, including natural and text images.