Real time ultrasound image denoising using NVIDIA CUDA

摘要

Image filtering is a process of reducing noise which degrades the performance of image processing. In some applications such as segmentation or classification, denoising has been designed to smooth the homogeneous areas while keeping and enhancing the edges. In several applications such as video analysis, image-guided surgical interventions or visual servoing, real-time denoising is needed. The devoted filter was suggested to adjust the filter of anisotropic diffusion to the features of the multiplicative noise existing within the ultrasound video and also to make easier the filtering process. The purpose of this study is to decrease the processing time implementation of the diffusion function using parallel processors through the optimization of the Graphics processor unit (GPU). The results show that the suggested method is very effictive in terms of real-time compared to a standard central processing unit (CPU) implementation. Our proposed model magnifies the acceleration of the image filtering to 7X compared to the sequential calculation.