A novel approach of adaptive thresholding for image segmentation on GPU

摘要

In the recent decades, graphical processing unit (GPU) improvises the programmability in the area of image processing for reducing the computational time as compare to CPU. GPUs can solve large data parallel problems and tends to give most robust and energy efficient segmentation method. It also gives the concept of memory pre-allocation and vectorization and handled the large image data set very profoundly for accelerated image segmentation method. In this paper, investigation results are obtained on 100 gold standard dermoscopy images which reveal that the proposed novel method of image segmentation is benefited from GPU processing in terms of speed and accuracy for lesion detection.