Low level feature selection for a content based digital mammography image retrieval system

Content Based Image Retrieval (CBIR) systems enables to retrieve images from large image archieves based on its contents as well as external attributes associated to each image. This study aims at extracting low level attributes to be used in a CBIR model that enables the utilization of low level image based attributes together with high level concepts. The contribution of this study is to develop an infrastructure for the selection of best low level attribute set to be used in the CBIR method by considering model performance. Within the scope of this study: segmentation of mammogram images, development of a mammogram database, low level attribute extraction from the segmented images and breast type estimation by means of machine learning algorithms are realized.