期刊名称:Journal of Theoretical and Applied Information Technology
印刷版ISSN:1992-8645
电子版ISSN:1817-3195
出版年度:2014
卷号:66
期号:1
出版社:Journal of Theoretical and Applied
摘要:In recent days the content-based image retrieval plays an important role in image retrieval and has achieved great development. This paper proposes a novel diagonal direction feature descriptor for content based image retrieval (CBIR). In Existing Local Tetra pattern, the relationship between the referenced pixel and its neighbors are encoded using first order derivatives only in vertical and horizontal direction. The image retrieval results are further improved by considering diagonal pixels for derivative computation in addition to vertical and horizontal direction. The proposed system includes the following phases 1) Pre-processing of an image using resize method and calculation of the direction of a pixel by computing the first order derivatives along with , and 2) Derivation of Local Octa Pattern from the direction 3) Construction of Magnitude Pattern using magnitude of first order derivatives 4) Calculation of feature vector on combining LOP and magnitude pattern and then similarity is measured. Experimental results show the proposed method is capable of effectively retrieving relevant images thus providing superior performance than several existing approaches.
关键词:Feature Vector; Local Tetra Pattern (LTrP); Local Octa Pattern (LOP); Local Binary Pattern.