Modelling of X-Ray Image Retrieval System using Zone Based Transform Techniques
P. Nalini1, B. L. Malleswari2
1P. Nalini, Department of Electronics and Communication Engineering, Mahatma Gandhi Institute of Technology, Hyderabad, India.
2Dr. B. L. Malleswari, Department of Electronics and Communication Engineering, Sridevi Woman’s Engineering College, Hyderabad, India.
Manuscript received on October 04, 2017. | Revised Version Manuscript Received on October 16, 2017. | Manuscript published on October 20, 2017. | PP: 13-17 | Volume-4 Issue-6, June 2017. | Retrieval Number: F0764054617/2017©BEIESP
Open Access | Ethics and Policies | Cite
© The Authors. Published By: Blue Eyes Intelligence Engineering & 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: Medical image retrieval based on its visual attributes is one of the prime aspects for the clinical decision making process. It can be beneficial and important to find other images of the same modality, the same anatomic region in disease identification. This results in improved healthcare system with increased efficiency. In this paper we proposed a model of medical image retrieval system that works on the features extracted from transformed coefficients by applying four different transformations viz. Discrete Fourier Transform (DFT), Discrete Sine Transform (DST), Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT). Visual content of X-Ray images better expressed with textural features. The texture attributes computed from transformed domain shown superior performance over spatial domain computed GLCM matrix based features. So in this paper we presented the comparison analysis of the four transformation techniques, DFT, DST, DCT and DWT in analyzing the performance of medical image retrieval system. IRMA 2009 X-Ray image database was used for experimentation. X-Ray image retrieval worked well with regionally computed attributes over global features, transform coefficients computed by partitioning the images into 64 regular regions of 16 x 16 in size. Images retrieved by finding the similarity between the region wise feature vectors. Among the four transformations, DWT and DCT result with more than 75% Mean Average Precision (MAP) at 100% recall rate.
Keywords: CBMIR, RBIR, Transform techniques, Zoning.