| Περίληψη: | The aim of this study is the development of a statistical denoising method, to reduce noise in scintigraphic images, preserving image quality characteristics such as contrast, and resolution. The method is based on principal component analysis (PCA) reduces the volume of image data, preserving a large amount of useful information, by considering that a small number of independent image components contain useful information (signal), whereas a large number of independent components contain statistical noise. Therefore, applying PCA and discarding the image components, which correspond to noise, noise reduction can be achieved.
PCA is a multivariate correlation analysis technique which explains algebraically a variance-covariance structure of observed data sets with a few linear combinations of original variables [28-30]. The motivation behind PCA is to find a direction, or a few directions, that explain as much of the variability as possible. This is achieved because each direction is associated with a linear sum of the variables, which are linear sums of the initial variables. Thus, the first principal component is the linear sum corresponding to the direction of greatest variability. The search for the second principal component is restricted to variables that are uncorrelated with the first principal component.
To assess the performance of the proposed denoising method was compared to four conventional noise reduction methods, employing quantitative image quality characteristics (noise and spatial resolution characteristics). Specifically, the linear filter (smooth 3x3 and smooth 5x5), and the non-linear filter (median 3x3 and median 5x5) were used. Additionally to demonstrate the applicability of the proposed method, it was applied to clinical planar scintigraphic images.
|
|---|
