Performance Analysis of Morphological Operation in CPU and GPU for Medical Images

As far as the problem of processing real time images studies showed that there can be two ways of analyze. On one hand is about the central processing unit (CPU) and the other is about the graphics processing unit (GPU). To obtain the highest possible performance they have to be used at the same time. Comparing the two processes, the result is that in the CPU, assignments as sequential operations and input-output are involved, in contrast to the GPU which is used for processes that can be made at the same time. Transposing these concepts of the real-time processing in medical digital images can be proved that the processing of an image is a response aproved within a certain time. The time we are speaking about lasts a few seconds or more less and images can be evaluated in a real-time medical side. Real-time image processing has become a very important objective and studied on the strength of its use in innovative media production, video calls, mobile phone cameras and video communication. The main purpose of this work is to improve the researching of a real-time digital image processing system applied on medical images. In this paper we try to present a method which is going to expose healthy and malignant tissues. Even if the physicians decisions remain the common method of detecting nonpalpable cancer, the method proposed in this paper plan to generate the fastest, easiest, and more accurate route, than in the case with classical techniques, by means of medical procedures. Our method of digital image processing by real-time processing is performed in order to allow medical specialists in radiology to detect abnormal tissues or cells, and then determine if a tumor is a malign or benign. The research illustrated in this paper involves an application toolbox with different filters, including segmentation, as well as various new tools for morphological filtering. More specifically, this paper focuses on CUDA as its parallel programming platform.