Invariant low level feature with emantics based image mining using Multi level object relational Semantic similarity measures

Abstract

The representation of information is always an important issue in modern society. Earlier days, the information was represented in textual manner. The growth of information technology has allowed the information in the form of images. For example, the ancient stories were drawn in the form of images which can be explained better than human. The entry of image based representation is applied in several problems from generic to medical solutions. Any information written on a document can be converted into images which cannot be erased easily. Similarly, the organizations maintain most of their information in the form of images or scanned copies in huge database. Such images would be recovered or retrieved from the huge data base whenever required. As the size of database increases, the retrieval of images relevant to the query is a quiet challenging one. The relevancy of result for a submitted query is most important. The image mining is a hallenging issue, when applied in several areas. The image mining is the process of retrieving images relevant to the query being submitted. The problem of image mining has been performed in several ways. The methods can differ on the feature being used for the measurement of similarity and for classification. The color features used for imilarity measurement between different images produces higher false ratio and irrelevancy. It is necessary to consider the low and high level features in measuring the similarity between the images. There are a number of approaches identified for the problem of image mining, by considering different features like color, shape and texture. However, the methods have deficiency in producing relevant images for the query being submitted. newline