Segmentation of Malaria Parasite Candidate from Thickblood Smear Microscopic Images using Watershed and Adaptive Thresholding

Abstract

Segmentation of malaria parasite on thick blood smear is a critical intermediate step in automation process of malaria detection. Most of the thick blood smear have low quality that characterized by high noise, the low-intensity difference between background and foreground, and the presence of artifacts. This situation makes the segmentation process becomes difficult. In this paper we proposed a new segmentation strategy for microscopic images of malaria parasite obtained from thick blood smear using watershed and adaptive thresholding. The proposed method consists of two main stages: image enhancement and segmentation. Enhancement process used Low-pass filtering and contrast stretching. Meanwhile, the segmentation used combination watershed segmentation and adaptive thresholding. The performance was evaluated on 253 parasite candidates, cropped from 22 thick blood smear microphotographs. The experimental results showed that the average segmentation accuracy of the proposed algorithm was 95.2%. Further analysis showed that the nucleus and cytoplasm of the malaria parasite were successfully extracted, thus the method is suitable for being used on detection of malaria parasites.