A Modified Shape Feature Extraction Technique for Image Retrieval
A. Srinagesh1, K. Aravinda2, G. P .Saradhi Varma3, A. Govardhan4, M. Sree Latha5

1A.Srinagesh, CSE Department, RVR & JC College of Engineering, Guntur, India.
2K.Aravinda, CSE Department, RVR & JC College of Engineering, Guntur, India. 522019, India.
3G. P. Saradhi Varma, IT Department, SRKR College of Engineering,Bhimavaram,54 204 India.
4A.Govardhan,CSE Department, JNTUH,Hyderabad, India.
5M.Sree Latha, CSE Department, RVR & JC College of Engineering,Guntur, India.

Manuscript received on June 11, 2013. | Revised Manuscript received on June 15, 2013. | Manuscript published on June 25, 2013. | PP: 9-13 | Volume-1 Issue-8, June 2013. | Retrieval Number: H0338061813/2013©BEIESP

Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Published By: Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: Semantic based Image retrieval is an emerging research area and is currently the mainstay in variety of applications or domains. In recent times, there exists a lot of gap between the high level semantics and low level features. The process of Features extraction is Application-specific or options are limited. In this paper, we propose a new Modified Shape Descriptor (MSD) feature extraction technique which is used as descriptive feature to discriminate Objects in an image database. In Object recognition after initial Pre-processing, feature extraction is the next crucial step which determines the efficiency of the technique or method. In our approach, a test image is taken from the database, which is then divided into 8×8 Blocks each; shape structure is detected using edge detection technique with Threshold method to generate the shape feature vector. Then, texton-based texture, color features are extracted using the existing Multi-texton Histogram (MTH) method. To form the final discriminating feature vector for that image in total, three features are extracted namely shape, texture and color for that particular image to form a discriminating feature vector which this then stored in a feature library. When a query image is given Euclidean distance between the query image and the test images feature values available in the feature library are computed. Based on the similarity characteristics top-k images are retrieved. Our proposed method gives better results when compared with other existing techniques.
Keywords: Content-based image retrieval, Pattern Recognition, Image Retrieval, Multi-texton Histogram, Shape Descriptor.