Cost Parameter Analysis and Comparison of Linear Kernel and Hellinger Kernel Mapping of SVM on Image Retrieval and Effects of Addition of Positive Images

Abstract

Listen

Translate article

paper we have brought out the analysis and comparison of cost parameter validation in Support vector machine using two different kernel mappings i.e. the linear and the Hellinger kernel. This paper also shows and discusses the results of the addition of positive images to the respective class of images with different cost parameters. The analysis is carried out using Matlab R2009a and C environment. The results obtained show that the increase in cost parameter for linear kernel gives much better results whereas for Hellinger kernel the performance decreases as cost parameter is increased. In the other hand, two classes of images are taken and they are tested by increasing the number of positive images gradually and the results show that the addition of positive class of images to a database can increase the performance of the system employed. 1 INTRODUCTION he growth of the World Wide Web have led to the huge online digital images and videos, so there is a strong demand for developing an efficient technique for image retrieval to exploit maximum benefit from this huge amount of digital information. With the advancement in technology, a large amount of information in the form of images is being generated daily in various fields like architecture engineering designs biometrics satellite imagery. Also with the increasing capacity of the storage devices, the database of the image information is expanding allowing a huge amount of database images to be stored quite easily. Even though storing these images is now easier, accessing or retrieving these images as per the requirement is a tedious job. Image retrieval has been an active topic for research for the past three decades. The goal of an image retrieval system is to retrieve a set of images from a collection of images such that this set meets the users requirements. The users requirements can be specified in terms of similarity to some other image or a sketch, or in terms of keywords. An image retrieval system provides the user with a way to access, browse and retrieve efficiently and possibly in real time, from these databases. Well-developed and popular international standards, on image coding have also long been available and widely used in many applications. The challenge to image indexing is studied in the context of image database, which has also been researched by researchers from a wide range of disciplines including those from computer vision, image processing, and traditional database areas for over a decade. This paper analyses the most important parameter i.e. the cost parameter of the support vector machine for various classes of images. The cost parameter is significant in a SVM classifier and has to be studied very carefully to train and test the dataset and obtain satisfactory results. The other sections of the paper can be put up as follows: SECTION 2 describes the Support Vector Machine and the Significance of the cost parameter for various kernels, SECTION 3 Provides information about the Database, SECTION 4 explains Feature Extraction, SECTION 5 gives the results and discussion and finally SECTION 6 gives the conclusions.