Adaptive Sequentially Weighted Median Filter for Image Highly Corrupted by Impulse Noise

Jiayi Chen,Yinwei Zhan,Huiying Cao

doi:10.1109/access.2019.2950348

Abstract

To tackle the difficulties in the detection and removal of impulse noise faced by the existing filters, and to further improve the denoising performance, we propose an adaptive sequentially weighted median filter for image corrupted by impulse noise. In the proposed method, a noise detector employing the 3σ principle of normal distribution and the local intensity statistics, is proposed; and a sequentially weighted median filter with a neighborhood of adaptive size, is proposed for noise removal, in which the weighted operator is derived in reference to the spatial distances from central noisy pixel, i.e., the weighting coefficients are sequentially inversely proportional to the spatial distances. The experimental results confirm that the proposed method outperforms the existing filters, excelling in the capability of noise removal, structure and edge information preservation.

Highlights

An image is often corrupted by impulse noise in the process of acquisition and transmission; and there are two types of impulse noise: fixed-valued impulse noise and randomvalued impulse noise [1]
In a black image region, we employ local intensity statistics for noise detection based on these two observations, that is, if a pixel takes minimum intensity, and the minimum intensity accounts for the majority in its neighborhood, this pixel is labeled as noise free, otherwise as noisy
The empirical validation for the proposed adaptive sequentially weighted median filter (ASWMF) is conducted by performing thorough comparative analyses with the stateof-the-art filters proposed recently in literatures, which are different applied median filter (DAMF) [11], neighborhood decision based impulse noise filter (NDBINF) [13], pixelvariation gain factors (PVGF) [18], adaptive dynamically weighted median filter (ADWMF) [22], radial basis functions interpolation (RBFI) [24], SVM classification based fuzzy filter (SVMFF) [25], NAISM [27], and iterative nonlocal means filter (INLM) [30], in terms of noise detection accuracy, peak signal to noise ratio (PSNR), structural similarity index (SSIM) [33], edge preservation index (EPI) [34], image entropy H [35], visual perception, and computational time

Summary

INTRODUCTION

An image is often corrupted by impulse noise in the process of acquisition and transmission; and there are two types of impulse noise: fixed-valued impulse noise and randomvalued impulse noise [1]. The traditional median filter was found unable to obtain a thorough noise removal and structure information preservation simultaneously, especially for high density noise, because it processes all pixels regardless of whether they are noisy or not, destroying the noise free pixels To address this problem, some researchers initially proposed. The existing filters inevitably have inherent shortcomings, and are not necessarily effective, especially for high density noise: they either overly smooth the image, or are unable to restore effectively the structure and edge information, so that they still could not satisfy the high requirements of image analysis and application To tackle this problem and provide high quality image for analysis and application, we proposed an adaptive sequentially weighted median filter (ASWMF) for image highly corrupted by impulse noise; the contributions of the proposed ASWMF are briefly described as follows.

RELATED WORKS

NOISE DETECTION BY 3σ PRINCIPLE AND LOCAL STATISTICS

EXPERIMENTS

NOISE DETECTION PERFORMANCE OF FILTERS

CONCLUSION