Comparative study of automatic seed selection methods for medical image segmentation by region growing technique

摘要

Seeded Region Growing technique is very attractive for medical image segmentation by involving the high-level knowledge of image components in the seed selection procedure. However, the Seeded Region Growing technique suffers from the problems of automatic seed generation. A seed point is the starting point for region growing and it's choose is very crucial since the overall success of the segmentation is dependent on the seed input. In this work three automatic seed placement methodologies are reviewed, evaluated and compared on three distinctive medical image databases. The first method is based on region extraction approach, the second one is based on features extraction approach and the last method is based on edge extraction approach. Our results showed that the region extraction approach performs well on the three tested databases. The features extraction approach gives good results with only two databases. Edge extraction approach gives correct results just on one database.