Title: Medical image up-sampling using a correlation-based sparse representation model

Authors: Katragadda Vamsi Krishna; Barjinder Singh Saini; Barjinder Singh Saini; Barjinder Singh Saini

Addresses: Electronics and Communication Engineering, Dr. B.R. Ambedkar National Institute of Technology Jalandhar, Punjab (144011), India ' Electronics and Communication Engineering, Dr. B.R. Ambedkar National Institute of Technology Jalandhar, Punjab (144011), India ' Dr. B. R. Ambedkar National Institute of Technology Jalandhar, GT Road, Amritsar Bypass Road, Jalandhar, Punjab 144011, India ' Dr. B. R. Ambedkar National Institute of Technology Jalandhar, GT Road, Amritsar Bypass Road, Jalandhar, Punjab 144011, India

Abstract: In this paper, a correlation guided sparse representation model is proposed for medical images that can be used for up sampling the images. For estimating some of the parameters of the sparse model, the repetitive patterns in the image are analysed. The relation between sparse model and content estimation of the image is explored and this approach is used to adapt the parameters of the existing sparse model. The sparse model works on the basis of dividing the image into several numbers of patches and sparse dictionary learning and then repeating this for a specific number of iterations. The patch size to be taken is passed into the model based on repetitive content in the image thus making the model to rely on the image characteristics. The numerical results obtained by applying this method on brain magnetic resonance images confirm the proposed approach as a better method compared to existing methods.

Keywords: correlation; sparse dictionary; medical images; image up-sampling; sparse representation model; brain MRI scans; magnetic resonance imaging.

DOI: 10.1504/IJMEI.2017.080926

International Journal of Medical Engineering and Informatics, 2017 Vol.9 No.1, pp.73 - 86

Available online: 12 Dec 2016 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article