Extraction Technique of Cell Targets from Marine Coscinodiscus Microscopic Images

Abstract

Cell extraction is the most important link to construct microscopic image identification system of phytoplankton, and the extraction quality directly determines the accuracy of recognition. In this paper, a new technique that combines high-precision threshold segmentation with proper morphological operation for automatically extracting Coscinodiscus cells from microscopic images is put forward. In particular, firstly, the input microscopic image to grayscale image is converted and then the gray value of the grayscale image linearly to enhance its contrast is mapped. Next, threshold segmentation using the between-class maximum discrete measure matrix trace method is performed to get the binary image, whose noise is removed by successive morphological close operation and hole-filling operation. Owing to the possibility of multiple cells in the image, those larger connected components are further extracted. The obtained image and the original grayscale image are merged into a single image by using logical AND operation. The final extraction result is an image with black background and grayscale cell foreground. The extensive qualitative and quantitative experimental results demonstrate that the proposed strategy can effectively remove various types of noise and accurately extract Coscinodiscus cells with clear edge and complete shape.