Removing Water Droplets In Medical Images Using Textured Spectral Analysis

Abstract

Some disease in human body is caused by accumulation of water molecules at particular place in an organ or entering of unwanted toxic foreign bodies. This may be in large or in negligible amounts. Many methods like CT, MRI can detect these abnormalities if present in appreciable amount. But if present in small amount, human vision on these scanned images cannot detect them. The visual appearance of moving water droplet is very complex. Each water droplet refracts and reflects both scene radiance and atmospheric illumination toward an observer. Water droplets are randomly distributed in space and move at high velocities. Thus, water droplets produce spatial and temporal intensity fluctuations in videos. Modelling, analysing and detecting these unwanted water molecules may benefit avoiding the negligence of presence of disease. In this paper, the proposed system that detect water molecules in the images of affected organ like lungs automatically .The crux idea is to exploit textural properties of droplets or fluid. To perpetrate this idea, we are aiming to model these droplets by laws of physical science and reveal this through block processing of image pixels. For partially occluded image portions, information of the image may be used to be applied in transform like DCT, blending functions and retrieve it. For fully occluded image, image completion techniques can be used. By using this we can detect the droplets even if they are in micron size.