Institute of Signal Processing
e-mail: Iivari.Kunttu@tut.fi |
|
Keywords: Histogram, Indexing, Content-based image retrieval, Paper defect images
Abstract
Image
retrieval can be considered as a classification problem. Classification is
usually based on some image features. In the feature extraction image
segmentation is commonly used. In this paper we introduce a new feature for image
classification for retrieval purposes. This feature is based on the gray level
histogram of the image. The feature is called binary histogram and it can be
used for image classification without segmentation. Binary histogram can be
used for image retrieval as such by using similarity calculation. Another
approach is to extract some features from it. In both cases indexing and
retrieval do not require much computational time. We test the similarity
measurement and the feature-based retrieval by making classification
experiments. The proposed features are tested using a set of paper defect
images, which are acquired from an industrial imaging application.