The retrieval of medical images plays a significant role specifically in medical diagnosis. The need for an effective and fast retiieval system is essential to retrieve most appropriate images from a template image to a query image for a particular case. This paper proposes a new effective image retrieval algorithm depending on local directional extrema based pattern for retrieval of CT images. The principle local binary pattern approach defines the relationship between the centre pixel and its corresponding neighbours by comparing the values of the grey level. In this article, a new feature descriptor is defined which extracts local directional extrema edge information’s along $0^{0},45^{0},90^{0}$ and 1350 directions. In the proposed method, first the directions of the existing extremas are investigated in a local neighborhood within a fixed radius around the center pixel. The feature descriptor is formed taking the local directional extremas into consideration, the directions in which extrema pixels are present and where they aren’t and encoded into a four bit binary pattern. The pattern value is computed by applying a uniform weighting function on the four-bit binary pattern which significantly decreases the dimension of the feature descriptor and solves the large size constraint problems. The proposed methodology is compared with local binary pattern and local diagonal extrema pattern for CT class of image database and it is observed that even with smaller feature dimension, the proposed feature descriptor gives better retrieval results outperforming both LDEP and LBP.
Medical Image Retrieval Using Modified Directional Extrema Based Pattern
2020-11-05
363581 byte
Conference paper
Electronic Resource
English
Valuation of image extrema using alternating filters by reconstruction [2568-10]
British Library Conference Proceedings | 1995
|British Library Online Contents | 2009
|Extrema Principles Of Dissipation In Fluids
NTRS | 1991
|Medical image retrieval based on complexity analysis
British Library Online Contents | 2010
|Content-Based Medical Image Retrieval using Dynamically Optimized Regional Features
British Library Conference Proceedings | 2005
|