Content Based Image Retrieval Using Geometrical-Shape of Objects in Image

Although some major advances have been made in text searching; only preliminary work has been done in image search. The field of Image search is rooted in Artificial intelligence, digital signal processing, statistics, natural language understanding, databases, psychology, computer vision, and pattern recognition. However none of these fields can solve the problem of image search alone but the solution lies at the crossroads of these parent fields. In our paper we are presenting a method of Contents Based Image Search where geometrical shapes of the objects in the image are considered as contents of image. Each object is separated from the image by segmentation. Then the geometrical shape of the object is estimated and compared with a predefine set of shapes of different categories. Number of objects in an image and geometrical shape of the objects are used as contents of the image which is used for retrieval and searching. Number on objects in the image is used for first level of indexing in search process. Currently we have restricted our objects to a fix number of basic geometrical shapes for simplicity but in futures these shapes can be extended and linked to the real world objects by using more complex equations and other features like color, texture and concept of correlation. Most of the existing image retrieval systems are based on text search using keywords that are annotated manually which involve the intellectual and emotional sides of the human. But in our proposed system this process is somewhat automatic.