부산대학교 도서관

Identification of objects of interest is most sought problem in computer vision related applications. This is in particular needed, when large volumes of data are available and a decision is to be made regarding relevance of an object to a specific region. In medical related applications, analysis of structural variations is much required for disease identification and progression. Manually delineating the affected portions is time consuming and prone to error. In the current paper, a novel algorithm is proposed to extract most significant tissue of human brain, Hippocampus. The algorithm uses labeling algorithm which is simple of its kind and does not need any prior knowledge. The segmented results are further compared with ground truth image using most prominent similarity indices, Dice Similarity Coefficient (DSC) and Jaccard coefficient.