A novel hybrid decision-based filter and universal edge-based logical smoothing add-on to remove impulsive noise

Abstract

This paper presents a novel hybrid filter along with a universal extension to remove salt and pepper noise even at very high noise density. The proposed filter initially specifies a threshold and then de-noises the image using combination of linear, non-linear and probabilistic techniques. Further, to improve quality, a universal add-on is presented which uses edge detection and smoothening techniques to brush out fine details from the restored image. To evaluate the efficacy, the proposed and existing filtering techniques are implemented in MATLAB and simulated with benchmark images. The simulation results show that proposed filter is able to restore image details even at extremely high noise density of 99%. Further, the proposed filter provides admirable results on natural as well as medical images from very low to very high noise density. Finally, it is observed that the proposed filter on an average improves the PSNR by 11% over the state-of-the-art technique.

Autoencoder-based Image Denoiser Suitable for Image of Numbers with High Potential for Hardware Implementation

“…However, using unnecessary intensive computational techniques for highly needed and widespread applications, such as those dealing with numbers and simple images, may be discourageable for hardware implementation. In this case, the only required features beyond the high accuracy are the fast processing time and the computational simplicity that guarantees the possibility to be hardware implemented [7], [8]. In light of the discussion mentioned above, this research aims to propose a simple autoencoder-based image denoiser with three and five layers and also discuss and investigate the conventional image filtering techniques to be a reference for evaluating the efficiency of the proposed technique.…”