Regressive Damped Linear Filterative Spatial Convoluted Edge Smoothing for Image Quality Enhancement

Abstract

Image quality enhancement aims to improve the rich details from degraded images, which is applied in many fields, such as medical imaging, video surveillance, criminal investigations, remote sensing, etc. Natural images captured under varying light conditions have poor contrast, low brightness, hidden colors, and high noise.  Therefore, image processing methods are developed for image enhancement. Image processing is the method of performing the analysis and manipulation of digitized images for increasing the image quality by minimizing noise and contrast enhancement. Numerous techniques have been developed for image enhancement. However, these techniques are only suitable for enhancing the images but it fails to remove the artifact-free quality improved results for various other types of images. Therefore, to meet this aim, in this paper, an automatic image enhancement technique called Piecewise Regressive Damped Linear Filterative Spatial Convoluted Edge Smoothing (RDL-SCES) is introduced for image preprocessing to enhance the image quality with the higher peak signal-to-noise ratio and minimum error. The proposed RDL-SCES technique performs image preprocessing that includes two processes namely filtering and edge smoothing.  In the RDL-SCES technique, number of natural images are collected from the dataset and considered as input. Then, every natural image gets pre-processed by using Piecewise regressive damped Bryson–Frazier Fixed Interval Filter. The designed filter employs the series of measurements observed over the different states including the statistical noise.  The proposed filtering technique performs image pixels analysis at every observation state to determine the smoothed image and covariance with help of piecewise regression and the Damped Least-Squares method. After the noise removal, the RDL-SCES technique performs the edge smoothing by using spatial convolutive Marr–Hildreth edge smoothing. This in turn helps to enhance the image quality. Experimental evaluation is carried out using natural images with different factors such as mean square error, peak signal-to-noise ratio, pre-processing time, and memory consumption with respect to a number of natural images and sizes.