Structural similarity analysis for brain MR image quality assessment

Abstract

Abstract. Brain MR images are affected and distorted by various artifacts as noise, blur, blotching, down sampling or compression and as well by inhomogeneity. Usually, the performance of pre-processing operation is quantified by using the quality metrics as mean squared error and its related metrics such as peak signal to noise ratio, root mean squared error and signal to noise ratio. The main drawback of these metrics is that they fail to take the structural fidelity of the image into account. For this reason, we addressed to investigate the structural changes related to the luminance and contrast variation (as non-structural distortions) and to denoising process (as structural distortion)through an alternative metric based on structural changes in order to obtain the best image quality.

Algorithm for Analyzing the Microenvironment Surrounding Melanoma

“…2. The Structural Similarity Index (SSIM) measures the perceived changes in structural information, and it is computed using the luminance (l), contrast (c), and structural texture (s) [13].…”

“…, 𝜇 𝑥 , 𝜇 𝑦 , 𝜎 𝑥 , 𝜎 𝑦 , 𝜎 𝑥𝑦 , are the local means, standard deviation and cross-covariance for images x, y. 𝐶 1 , 𝐶 2 and 𝐶 3 are constants, and 𝛼, 𝛽, 𝛾 are default exponents [12], [13].…”