Worm egg segmentation based centroid detection in low contrast image

In this paper, we propose silkworm eggs image segmentation technique based on centroid detection in low contrast image. The objective of this work is to identify individual objects from clustered objects for quantity evaluation. The technique can be divided into three parts, The first part is to separate foreground edge of worm egg image from its background. The second part is to convert the foreground binary edge image to reconstruct gray scale image. In this process, a circular weight mask which has the same size of a worm egg object is created and applied to the binary edge image. Shade color that obtained from the gray image result will describe different region between egg's object centroid and its boundary. In the last part, an iterative thresholding operation together with several criteria such as object size and distance is applied to the gray shade image to determine centroid positions in clustered worm egg objects. The last two part is the main idea of this paper which demonstrates a simply way to solve a problem of how to detect touching objects that has slightly color different between foreground and background. In experimental result, the accuracy rate is 87.9% obtained from 8558 object number.