The Research Of Image Retrieval Method Based On Semantic Network

With the rapid development of multimedia technology and network and a large number of images increase, the effective management and retrieval of images is becoming increasingly important. For users, how to retrieve the image quickly and accurately from the large image database, is an important topic in current research.From the early 20th century,90s, content-based image retrieval of multimedia technology on becoming one of the hot. Content-based image Retrieval technology gradually become one of the hotspots of multimedia technology.CBIR techniques mainly use visual features (texture, shape, color and spatial relations between objects) to retrieval images.But the existence of a semantic gap is the major obstacle of image retrieval technology.Thus semantic-based image retrieval technology came into being.Image retrieval systems, as derived from its larger set, the multimedia information systems, inherit the same trait as a multidisciplinary field from computer science.The ever-growing knowledge comprises elements from database management, signal processing, computer vision, natural language processing, networking and human computer interaction. Many systems, such as QBIC, Photobook, VisualSEEK and Blobworld attempt to adapt human perceptual capabilities. Early approaches in image retrieval systems also adopt many techniques from the pattern recognition field. In an image retrieval system, user information need is expressed using multiple types of query.Unfortunately, due to user subjectivity perception to visual features and semantic depths of images, the conventional query submitted to the system encounter difficulties to identify user information need. For example, query by sketch and query by image example do not represent semantic content of the targeted image.Query and matching are done using images visual features only. Meanwhile, query by text annotation faces the ambiguity of user description that poses difficulty for Natural Language Processing techniques. Different mapping and lack of uniform correlation between system's annotated images and user's own constructed annotation in the query pose problem for image retrieval systems. The blooming of interest in semantic image retrieval requires current research direction in image retrieval system to be more concerned into semantics.Currently, vast querying techniques are used to accommodate user query to different retrieval systems.For querying to an image retrieval system based on text, the query item consists of metadata, keywords and textual annotation, which describe the image. For image retrieval system based on visual features, queries fall under the query by sketch and query by example types.Category browsing is used for both systems depending to the database size and the respective system classification abilities.Content-based image retrieval and text-based image retrieval are two Fundamental approaches in the field of image retrieval.Nowadays, Image retrieval systems, emphasize the combining of high level semantics and low level visual features in the image indexing process. Recently, the researchers use the combining approaches and semiautomatic image retrieval, using the user interaction in the retrieval cycle. The proposed approach can reply different requests in the image retrieval domain based on a hierarchical semantic network and doing a new kind of learning process by the feedbacks given by user.