International Journal of Computer Theory and Engineering

Abstract

Determining the number of cell in blood sample, especially with high degree of overlapping, are remaining an obstacle need to be overcome, however. This study presents a new method for splitting (extracting) clumped cells into single (individual) cells supporting useful information for classification and detection infected cells. The proposed method is mainly focused on rapidly detecting central point using the distance transform value. Additionally, a boundarycovering degree of each center point was applied to select the best potential center points. Single cell extraction is employed in order to estimate the average size of cell. Finally, an effective algorithm was designed split correctly and speedily. The robustness and effectiveness of our method have been assessed through the comparison with more than 400 images labeled manually by experts and exhibiting various clumped cell. As the result, the F-measure generally reaches 93.5% and more than 82% clumped cells can be tolerated in the condition of non-distorted shape and well-focused images.