Semi-supervised Super-pixels classification for White Blood Cells segmentation

In this paper, we focused on annotating images at the super-pixel level using a semi-automatic segmentation approach. This process requires more important expert interaction. White Blood Cells segmentation is one of the most challenging topics in cytological image processing. In this work, we have proposed a semi-automatic process for White Blood Cells segmentation. The segmentation process is based on super-pixel classification using the semi-supervised co-Forest classifier. The first step of our process is about super-pixels generation by SLIC algorithm. In the second step, we apply the features extraction for each super-pixels, the features are extracted from color information (RGB) and Color Mean (CM). In the last step, we realize a semi-supervised classification to classify each super-pixel into nucleus and cytoplasm region. The super-pixel classification results gives a good segmentation of nucleus and cytoplasm of different nature and structure.