In the image retrieval, user must search a database of many thousands, or millions, of images. User goals vary to find a particular image. The optimal interface would provide a very flexible query mechanism, perhaps through a natural language interface. In fact, many “search engines” currently provide such an interface to their collections. Advertisers and publishers present a description of their requirements, for example “an image of the beach with athletic people playing volleyball”. Human clerks then scan many images by hand using keywords. In some of these systems a user interact with the retrieval engine through example images, in others the user is also asked to weight a set of “intuitive” features, such as color, texture and shape. This paper provides a content-based image searching engine based on (FMD) Feature Matching Descriptors. Feature Matching Descriptors, are invariant to image scaling and transformation and rotation, and partially invariant to illumination changes and affine, present the local features of an image. Therefore, feature keypoints are extracted more accurately by using Feature matching descriptors than color, texture, shape and spatial relations feature. To decrease unavailable features matching, a dynamic probability function replaces the original fixed value to determine the similarity distance and database from training images. Then, by using pretreatment of the source images, the keypoints will be stored to the XML format, which can improve the searching performance. Finally, the results displayed to the user through the HTML. The demand and need of the hour is to formulate the strategies to cope up with the challenge posed by the semantic gap of image retrieval. The semantic-based image retrieval is a better way to solve the “semantic gap” problem, so the semantic-based image retrieval method is stressed in this paper.
