Nowadays the main hurdles facing Content-based Image Retrieval Systems are: 1) the semantic gap between the low-level visual features and the high-level semantic concepts and 2) human subjectivity in regard to visual content. This work is inspired on the WLSP-C±1 image similarity model, proposed by Stejic (2003). The main characteristics are: the region-based image comparison, the use of feature combinations, and the image similarity measure’s adaptation to the user’s criteria using weights that reflect their relevance and irrelevance concepts and, also the undesirability of similarities. The main objectives of this work are: to improve the retrieval precision and to increase the methods speed. Experiments have shown that the proposed method provides a better performance when compared to Stejic’s model which was affirmed to as having a greater efficiency than many of the existing methods, as observed by the authors. Tests in a database of 4200 images have shown that the system is very efficient, indicating the possibility of World Wide Web application, using the force of parallel computation to minimize search time.
|