Probabilistic Decision Based Improved Trimmed Median Filter to Remove High-Density Salt and Pepper Noise
- PDF / 6,903,805 Bytes
- 15 Pages / 612 x 792 pts (letter) Page_size
- 83 Downloads / 182 Views
Probabilistic Decision Based Improved Trimmed Median Filter to Remove High-Density Salt and Pepper Noise A. P. Sena,* and N. K. Routa,** a
School of Electronics Engineering, KIIT, Deemed to be University, Bhubaneswar, India * e-mail: [email protected] ** e-mail: [email protected]
Abstract—This paper focuses on the removal of salt and pepper noise from a contaminated image. A Probabilistic Decision Based Improved Trimmed Median Filter (PDITMF) is proposed here. The proposed PDITMF algorithm resolves the conflict regarding an even number of noise free pixel of Trimmed Median Filter. The proposed algorithm makes use of two estimation techniques for de-noising, namely, Improved Trimmed Median Filter (ITMF), and Patch Else Improved Trimmed Median Filter (PEITMF) depending upon noise density. The algorithm experiments with many standard sample images. Simulation results show the proposed algorithm is capable of de-noising the image very efficiently. The algorithm has a better visual representation and it outperforms the existing well-known algorithms in context to peak signal-to-noise ratio (PSNR) as well as image enhancement factor (IEF) with lower execution time (ET) at all noise densities. Keywords: improved trimmed median, patch else improved trimmed median, probabilistic approach, salt and pepper noise DOI: 10.1134/S1054661820030244
1. INTRODUCTION In the present era of expanding digitalization, it is an important area of research to renovate an image from a contaminated image. This factor of preserving images through image processing techniques attracts most of the researchers to de-noise for better visual perception. There are many pieces of literature reported in this area. The quality of image pre-processing depends on the removal of the noise from the corrupted image without destroying the edges [1]. Image corrupted through noise can occur due to many reasons such as during acquisition, transmission, and storage. Different types of noises that may corrupt an image are salt and pepper noise, additive Gaussian noise, and Poisson noise. Among them, salt and pepper noise affect the most where the corrupted pixel may be replaced by the maximum or minimum value with a certain probability [2]. A number of linear and non-linear filters have been developed for de-noising a corrupted image. Linear model [3] works well when noises are additive in nature. Its benefit is its speed while the linear model is unable to retain edges of the image in an efficient manner. These are recognized as discontinuities and blur on the image. Therefore, the research got directed towards nonlinear filters [4]. The Standard Median Filter (SMF) [3, 4] is the best example of it. The SMF works well in lowdensity noise but ceases to perform as the noise density (ND) increases. This algorithm is unable to distinguish
Received August 11, 2019; revised January 28, 2020; accepted January 31, 2020
between thin lines and impulses and as a result; thin lines are interpreted as impulses and removed. Then, the filter with an adaptive wi
Data Loading...
