The Research Of Annotation Model For Web Images Based On Semantic

Faced with an enormous amount of web multimedia resources, especially the widely used and rapidly increased web image resources, it is necessary for people to quickly and accurately find out the interested ones, and then rapidly and completely understand their semantic content. Therefore, how to effectively represent and organize the semantic content of these large-scale web image resources, and then quickly discover the useful knowledge has became a hot research field.Generally, people always obtain knowledge based on domain/event, but the image resources of single domain/event are on a vast scale and still in rapid growth. Therefore, this article proposes a Three-Level Association Link Network Annotation Model for the high-level semantic content of web image, and then put forward a method to calculate the association weight among web images based on complementary semantic. Our purpose is to build the association network among web images and help web users obtain the needed knowledge quickly. The main research contents in this paper are as follows:1. The initial expression model for web imageHow to exactly represent a web image is the important step, it is also the basis for image clustering. Firstly, we discussed four kinds of text representation models: Vector Space Model, Non-negative Matrix Factorization Model, Association Link Network Model, and Power Series Model; Then, we discussed and put forward four initial representation model of web images; Lastly, the advantages and disadvantages of the four models were discussed.2. Three-Level Semantic Annotation Model for web imageCurrently, the main way of web users obtaining image resources is the Search Engine, and the image resources are around by the short describing information. It may be several discrete keywords or a paragraph of text, but the high-level semantic content of image is highly rich, this kind of image annotation method is unable to complete and detailed expression the semantic content of image. This paper presents Three-level Semantic Annotation Network to express the content of web images.3. The calculation method of web association weightIn this paper, we put forward a method to calculate the association weight among web images based on complementary semantic keywords. Our aim is to construct an association network among images.