Performance analysis of image smoothing methods for low level of distortion

Abstract

Image enhancement is an initial and basic preprocessing step in image processing. The success of this primary module determines the accuracy rate of higher level of image processing task. This module is application specific in nature. So determining a suitable image enhancement for real time images is a challenging one. Common distortions that are found in the real time images are contrast variation, blur, salt and pepper noise and so on. It is a challenging task to enhance the low level of blur distortion in an image. Hence in this paper the existing image smoothing methods (mean filter, Gaussian filter, anisotropic diffusion, median filter, adaptive median filter, conservative smoothing, and alpha trim mean filter) suitable to enhance low level distortion in images are evaluated in detail. Images of low distortions along with its ground truth image were taken from the CSIQ dataset. Performance of the existing image smoothing methods was measured using Peak Signal to Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM). Results showed that PSNR and SSIM were higher for conservative smoothing method followed by the adaptive median filtering method. Hence, this study concludes that conservative smoothing technique and adaptive median smoothing among neighborhood processing techniques could be deployed as a suitable image enhancement algorithm for blur effect removal.