Proposal of Local Automatic Weighing Attribute in CBIR

Abstract

Lung cancer is the most common malignant lesion and the principal cause of cancer-related death worldwide. This problem encourages researchers to build computer-aided solutions to help diagnose lung cancer. Content-based image retrieval (CBIR) systems are very promising in this context due to a large number of image generated everyday. However, semantic gaps have limited CBIR applicability. This work proposes a new approach to automatically adjust CBIR attribute weights to reflect users' semantic interpretation on retrieval process, minimizing the semantic gap problem and improving retrieval accuracy.