Distance Functions Association for Content-Based Image Retrieval using Multiple Comparison Criteria

The comparison operators available in traditional database management systems (DBMS) are not adequate to handle complex data such as images, rather comparing them using similarity operators is the option of choice. Similarity operators need a way to measure the similarity between pairs of objects. Although there are many interesting works dealing with similarity queries and functions to measure similarity, they all rely on a single similarity function that must be applicable over the whole dataset. However, images from medical exams often require several ways to measure similarity, depending on many factors, such as the particular pathological condition being searched, or the existence of specific clinical condition revealed in the images compared. Therefore, the ability to handle several ways to compare images by similarity is important in medical software handling images. This work develop a technique to allow several similarity functions to be combined when indexing a large set of images, allowing queries to probe the dataset regarding distinct comparison criteria. This technique also allows a flexible way to pose queries supporting fast retrieval of the answers