Discriminative Switching Filter

S.S. Agaian, R.R. Sifuentes, and L. Lamoureux (USA)

Keywords

Nonlinear filter, noise detection, andswitching filter

Abstract

We propose a new switching method of non-linear filtering to suppress various types of noise while focusing on the preservation of image detail and features. The new approach uses new local noise identification measures in combination with adaptive algorithms which are able to switch filters depending on the type of noise that is found to be present. The introduced identification measures and adaptive algorithms are derived from a combination of median-based local statistics. Testing and analysis was done using a small database of 100 color and 100 grayscale images varying in size, color, format with a mix of different types and levels of noise such as impulsive and Gaussian. Simulation results show that our method performs superior in comparison to other existing algorithms like the traditional median [1], center weighted median [1], switching ROM filters [2] [3], and entropy based switching filters [4] [5]. Other applications include audio filtration, filtering video transmission, and noise classification.

Important Links:



Go Back