Image Metadata Feature Extraction And Its Application In Retrieval

As network resource is expanding today, the requirement of resource application and management is imminent. To deal with different structure system and various type of resource, the construction of metadata standard is emerging as the times require.In the field of education, with regard to construction of educational resource, several types of metadata standard have been set up too. The image is crucial part of construction of educational resource, but up to now the construction of it's metadata standard is still based on text retrieval at it's exterior.There is no intersection between the image application of content based and the metadata standard construction, which currently represent two different way of development. The paper herein is trying to find the connection point between the aforesaid independent methods, we define it as IMAGE METADATA.The definition of metadata is referred from the concept of metadata standard construction. The concept of image is differentiated from meaning of the traditional image, we specially call it the image of content-based. In the aspect of image feature extracted, the technique of content- based visual information retrieval is usually implemented, in terms of image defined, it is referred from description specification, language and method of construction of metadata standard.The main points of this paper manage to solve are summarized as follows:1. For metadata feature extracted of image color, RGB color histogram, HSV color histogram and HSI color histogram are extracted respectively;2. For metadata feature extracted of image texture, we set up four-direction gray scale co-occurrence matrix and then extract five feature vectors;3. For metadata feature extracted of image shape, we deal the image with wavelets for triple times. And then extract seventh order matrix of image;4. For image feature description; firstly we set up a model of image features descriptors and image description, using XML & RDF languages to describe the image, and then we set up a trial modeling system and finally to retrieval image based on color, texture and shape.5. In the end, we give image retrieval result and analyze it.The conclusions to content-based image metadata retrieval of the paper are derived as...