Indian Journal of Science and Technology

Abstract

Objectives: To investigate on choice of suitable image quality assessment techniques and effect of parameters such as gradient threshold, time step, number of iterations etc. on anisotropic diffusion filtering. Methods/Statistical Analysis: Various diffusivity functions and the individual contribution of associated parameters are compiled and evaluated. Edge preservation in smoothing is an important issue while filtering digital images. A recently proposed image quality metrics is used for evaluating the edge preserving ability of filters. A comparative study is carried out on the basis of experiments and their performance is tested on standard test images using 16 different image quality metrics. Findings: The experimental findings show that most of the true edges got lost for higher number of iterations along with more smoothening of images in techniques namely, Perona and Malik diffusion, Speckle Reducing Anisotropic Diffusion (SRAD) and Weickert anisotropic diffusion. Further, in robust anisotropic diffusion it is found that conduction coefficient plays significant role in filtering process. Higher value of conduction coefficient results in more smoothening with blurring of sharp details and edges. It is concluded that image quality measures such as universal Quality index (Q), Edge Retrieval Index (ERI) and Structural Content (SC) can give significant information about image quality even in cases where conventional image quality metrics such as Picture Signal To Noise Ratio (PSNR), Mean Square Error (MSE), Root Mean Square Error (RMSE), geometric average error (GAE), normalized absolute error (NAE) etc. fail to do so. Proposed approach can give significant information when conventional metrics fail to assess the filtered image quality. Application/Improvements: The behavioural characteristics of techniques studied will be suggestive when applied on real time noisy images such as ultrasound images which inherently contain speckle. 

Keywords: Diffusion, Diffusivity Function, Edge Preservation, Edge Retrieval Index, Noise Filtering, Speckle