Analysis of Colour Constancy Algorithms for Improving Segmentation of Malaria Images

Abstract

Malaria is a very serious disease that caused by the transmitted of parasites through the bites of infected Anopheles mosquito. Malaria death cases can be reduced and prevented through early diagnosis and prompt treatment. Currently, microscopy-based diagnosis remains the most widely used approach for malaria diagnosis. The appearance of the infected red blood cells (RBCs) and their morphological features are very important for recognising the presence of malaria parasites. However, it is difficult to identify the presence of malaria parasites as well as observing its morphological characteristics due to the non-standard preparation of the blood slides; producing colour varieties in different slides. Thus, this study aims to apply colour constancy algorithms for standardisation of blood images in order to enhance segmentation of malaria parasites. In this paper, four different colour constancy algorithms namely Gray-World, white patch, modified white patch and progressive algorithms have been analysed to identify colour constancy algorithm that can give the significant segmentation performance. The experimental results show that segmentation on Gray-World images has successfully segmented 100 malaria images with average segmentation accuracy, sensitivity and specificity of 99.60%, 91.26% and 99.85%, respectively.