Accelerate Image Processing Algorithms using parallelized processing

Abstract

One of the most embarrassing feature of the lung X-ray images is the shadows of the bones. It could be disturbing at the medical examination if it is the lung which is being examinated, not the bones. These shadows can hide abnormalities and can decrease the chance of detecting them. During the evolution of the technology, it is available nowadays to take digital, high-resolution pictures of the lung, which can be processed by computers in order to make Computer Aided Diagnose (CAD) systems which help the radiologists with new features like detecting abnormalities or eliminate bone shadows. The disadvantages of these algorithms are that they are too slow to be used in a realtime application. A bone shadow eliminating function can be also useful for a abnormality detector program in order to increase its efficiency but it is unuseless for the doctor if he has to wait for minutes to get the result. The application of the graphics cards with tremendously high computing capabilities is an up-to-date, persistently developing technology. In my presentation, I introduce the NVIDIA® CUDA™ system, and how to make an algorithm faster with using parallel computations.