Optimized algorithms for real time medical image registration

Many applications in the medical image processing field use huge volumes of data. Usually this kind of huge data suffers from problems such as linear and nonlinear distortion, Partial data losses during acquisition. Registration is a basic step for successful processing of this data. Usually it consumes few minutes to align two images together therefore and consequently it takes few hours to register huge data. In this work, Real MRI images with linear distortion are used and then registration problems are fixed with rigid image registration algorithms based on cross correlation algorithm. TMS320C6000 processor is used for hardware implementation and DSP profile to estimate the execution time of the different parts of the c source code. Therefore the part of the code that consumes the main part of execution time is written with assembly code to reduce the total execution time with parallel processing methods. The results show that reduction in time is more than 30 % between optimized and non-optimized algorithms on embedded DSP platform that reduce the computation time of image registration to a fraction of a second.