Call of Papers for Current Volume **************** OnLine Submission of Paper

Volume & Issue no: Volume 3, Issue 7, July 2014


Column Transform based Feature Generation for Classification of Image Database
Author Name:
H.B.Kekre, Tanuja K Sarode and Jagruti K Save
ABSTRACT Designing computer programs to automatically classify images using low level or high level features is a challenging task in image processing. This paper proposes an efficient classification technique which is based on image transforms and nearest neighbor classification. The database of images which has been used for experimentation is large containing 2000 images (20 classes) with wide variety in them. Since the performance of classifier is largely depends on the feature vector, a lot of research is going on the feature generation methods. This paper analyses the different transforms in this application domain. Initially Transforms like Discrete Fourier Transform(DFT), Discrete Cosine Transform(DCT), Discrete Sine Transform(DST), Hartley Transform , Walsh Transform and Kekre Transform applied to the columns of three planes of color image. Then using fusion technique, feature vector is generated. For more dimension reduction, the size of vector is further reduced. Nearest neighbor classification with Euclidean distance as similarity measure is used for classification task. The performance of other similarity measures like Manhattan distance, Cosine correlation measure and Bray-Curtis distance are also tested. The paper also discusses the accuracy obtained for variation in the size of feature vector, the size of training set and its impact on the result. Keywords: Image classification, Image Transform, Nearest neighbor Classifier, Similarity Measure, Feature vector generation, Row mean vector
Cite this article:
H.B.Kekre, Tanuja K Sarode and Jagruti K Save , " Column Transform based Feature Generation for Classification of Image Database" , International Journal of Application or Innovation in Engineering & Management (IJAIEM) , Volume 3, Issue 7, July 2014 , pp. 172-181 , ISSN 2319 - 4847.
Full Text [PDF]                           Back to Current Issue