Abstract
In this paper, we address the problem of mixed Gaussian and impulsive noise reduction in color images. A robust filtering technique is proposed, which is utilizing a novel concept of pixels dissimilarity based on the reachability distance. The structure of the denoising method requires the estimation of the impulsiveness of each pixel in the processing block using the introduced local reachability concept. Furthermore, we determine the similarity of each pixel in the block to the central patch consisting of the processed pixel and its neighbors. Both measures are calculated as an average of modified reachability distances to the most similar pixels of the central patch and the final filtering output is a weighted average of all pixels belonging to the processing block. The proposed technique was compared with widely used filtering methods and the performed experiments proved its satisfying denoising properties. The introduced filtering design is insensitive to outliers and their clusters introduced by the impulsive noise process, preserves details and is able to efficiently suppress the Gaussian noise while enhancing the image edges. Additionally, we proposed a method which estimates the noise contamination intensity, so that the proposed filter is able to adaptively tune its parameters.
Highlights
In the recent years the topic of image denoising has been extensively studied in computer vision and digital image processing fields
In a pixel affected by impulsive noise, each RGB channel was randomly replaced by a value drawn from uniform distribution in the range [0, 255]
The images were first distorted by Gaussian noise with standard deviation in the range 10–50, and 10–50 % of the pixels was replaced by random-valued impulsive noise, so that every RGB channel of a corrupted pixel was assigned a value drawn from uniform distribution in the range [0, 255]
Summary
In the recent years the topic of image denoising has been extensively studied in computer vision and digital image processing fields. The enhancement of image quality is a crucial. As the denoising is the first step in the image processing pipeline, the effective restoration allows to successfully accomplish its further stages. The noise filtering methods used for color image enhancement can be divided into component-wise and vector-based techniques. The component-wise filters process the color image channels independently, neglecting the usually strong inter-channel correlation. The advantage of this approach is that many methods used for the greyscale image denoising can be directly applied to the color image channels and the processing results are merged to obtain the final restored output.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
