Nose bug egg geometric structure segmentation with anisotropic filtering and edge enhancement

This paper presents a method to segment geometric structures from the Meccus Phyllosomus bug eggs. These structures are of interest because they may help on the taxonomy of the bug families and their possible hybrids. The segmentation method is based on the Perona - Malik anisotropic filtering to enhance the images. A tune in parameter process is proposed for the filter. After this filtering, proposed techniques to improve edge definition, removal of false positives and edge linking are applied to the bug images. Four segmentation performance metrics are used to evaluate the proposed method. Over 90% of the processed images present metrics greater than 0.7 indicating a high level of segmentation.