Abstract

Visual information plays an important role in modern applications as the usage of multimedia becomes more widespread from day to day. However, impulse noise is ubiquitous in real life and often distorts an image. This paper presents a new noise-filtering algorithm called efficient nonparametric switching median (ENPSM) filter, which is capable of reducing the effect of low level random-valued impulse noise up to 30% of the corruption rate in digital images. The proposed filter is composed of a nonparametric impulse noise detector and a recursive pixel restoration technique. Initiated with the impulse noise detector to determine any potential noise pixels, the filter will then replace the detected noise pixels with the median value of the surrounding pixels. Yet, only the noise-free pixels are considered, in order to determine the median value in a way that the filter is more efficient in the pixel restoration process. Based on the simulation results, it has been shown that the ENPSM method performs better than some of the existing state-of-the-art methods by giving better filtering performance, both in terms of qualitative and quantitative evaluations. Hence, this ENPSM filtering algorithm could possibly be used as a preprocessing module in the electronic imaging products.
 
Key words: Image processing, impulse noise, digital image, noise filtering, nonparametric switching median filter.

Abbreviation

ENPSM, Efficient nonparametric switching median filter; SM, standard median; AM, adaptive median; CWM, centre weighted median; SWM, switching median; CWSWM, centre weighted switching median; LSM, laplacian switching median filter; ERID, enhanced rank impulse detector; MSM, multi-state median; TSM, tri-state median; MSWM, modified switching median; PSNR, peak signal-to-noise ratio; MSE, mean-squared error; MAE, mean of absolute error.