Adaptive Minimum-Maximum Exclusive Mean Filter for Impulse Noise Removal

TL;DR

This paper introduces the AMMEM filter, a fast and adaptive method for impulse noise removal that improves detection accuracy, image restoration quality, and computational efficiency, even at very high noise levels.

Contribution

The paper proposes a novel adaptive filter that overcomes limitations of existing methods by using variable window sizes and flexible parameters for efficient impulse noise removal.

Findings

Significant improvement in noise detection accuracy.

Effective image restoration at noise levels up to 95%.

Enhanced computational efficiency compared to traditional filters.

Abstract

Many filters are proposed for impulse noise removal. However, they are hard to keep excellent denoising performance with high computational efficiency. In response to this difficulty, this paper presents a novel fast filter, adaptive minimum-maximum exclusive mean (AMMEM) filter to remove impulse noise. Although the AMMEM filter is a variety of the maximum-minimum exclusive mean (MMEM) filter, however, the AMMEM filter inherits the advantages, and overcomes the drawbacks, compared with the MMEM filter. To increase the various performances of noise removal, the AMMEM filter uses an adaptive size window, introduces two flexible factors, projection factor P and detection factor T, and limits the calculation scope of the AVG. The experimental results show the AMMEM filter makes a significant improvement in terms of noise detection, image restoration, and computational efficiency. Even at noise level as high as 95%, the AMMEM filter still can restore the images with good visual effect.