DefinePK

Abstract

Image segmentation plays a crucial role in the medical imaging analysis to diagnosis diseases. The FCM algorithm is a widely used technique in this field and still being under improvement by researchers either at accuracy or at time execution. BCFCM (bias field correction FCM) is a robust variant of FCM that segment and corrects the intensity in homogeneity artifact on medical images. However the algorithm is always a time consuming problem because of the powerful treatment requirement. GPU-based parallelism is one of the used solutions to enhance his efficiency in terms of execution time. In this paper we have studied and modeled the behavior of CPU and GPU hardware against the sequential BCFCM and the parallel PBCFCM implementations. The modeled results for i7 3.5 Ghz CPU and GTX 760 GPU considering the clusters number variation show interesting behaviors and are in good accordance to experimental ones.