To communicate each other and store the information, we frequently use the multimedia data as much as the text data in the past. As the efficiency of image retrieval gets much more weight, the key features for representing the information are needed. ...
To communicate each other and store the information, we frequently use the multimedia data as much as the text data in the past. As the efficiency of image retrieval gets much more weight, the key features for representing the information are needed. Therefore, it is essential to classify the feature of shape which makes the image retrieval possible and to know how to measure the shape of image. In this dissertation, we consider shape descriptors proposed in MPEG-7 and simulate the one among these shape descriptors.
The purpose of MPEG-7 is to standardize the expression of multimedia data and to support the wide applications. The distinctive feature of shape descriptor is to reflect the feature of object as it is although the object transfers or scales or rotates. There are roughly two shape descriptors. One is based on the outline and The other is based on the region. In this paper, we use the chain code string based on the outline among these features of shape descriptor, the peak points and filtering.
So, at first, we divide features of shape by the feature based on an outline and the feature based on a region.
Secondly, we also consider how each of six shape descriptors which are proposed in MPEG-7 and estimate the similarity between a query image and images in database.
Thirdly, we simulate the standard of proposed shape descriptors called "Curvature Scale Space(CSS)". In this case, the shape descriptor recognizes the transformational images of a query image and the image before transforming very similarly although we apply rotating, scaling, flipping vertically and filtering of twisting, zigzaging to a query image.