Content Based Image Retrieval System

Image Retrieval System at the beginning of its development which is around the late 1970s, still use text to mark or give a description (annotation) to the image. First image captioned text form and then to do the retrieval process used DBMS (Database Management System)-based text. Provision of such information has the disadvantage that is: if a collection of images has a very large number, it becomes inefficient because the process is done manually and the information provided in the image is subjective, highly dependent on the perceptions of informants. To overcome these problems, in the early 1990s began to be developed CBIR (Content-Based Image Retrieval) that perform the retrieval based on visual content (visual content). The charge distribution can be either visual color, texture or form (shape) owned the image. In general CBIR system can be expressed in the diagram in figure below

visual content of images in the database is extracted, then described as a feature vector (feature vector) and stored in the database feature. To get back an image, users give input to the system in the form of an example image to be searched, the process is called the QBE (Query By Example). The system then change the sample image into the form of characteristic vectors and compare the level of similarity (similarity comparison) with feature vectors in the feature database. In the process of comparing the similarity indices used in the feature vectors in order to access the database more efficient. Process is then performed based image retrieval and sorting the resulting value to the process of benchmarking the level of similarity. Retrieval systems today also have involved the feedback from the user whether an image retrieval results are relevant or not (relevance feedback) is used as reference to modify the retrieval process in order to obtain more accurate results.