Image Super-Resolution

In image super-resolution (SR), one tries to obtain a high resolution image from several low-resolution (LR) images with overlapping fields of view.  SR algorithms first estimate the relative disparity among the different  LR images to achieve a precise registration and then try to obtain a high-resolution image by properly combining the registered images. In this project we are investigating the use of new sampling schemes to achieve a very accurate registration of very low-resolution multi-view images and then we achieve super-resolution by exploiting some a-priori knowledge of the properties of the original image (See the figures below for an example).

To probe further, check-out  our video showing the performance of our super-resolution algorithm in Loic Baboulaz homepage.

(a) Original Image (2014x3039)	(b) ROI of (a)  (128x128)	(c) Super-res. image (1024x1024)

• L. Baboulaz and P.L. Dragotti,  Exact Feature Extraction using Finite Rate of Innovation Principles with an Application to Image Super-resolution, IEEE Transactions on Image Processing, to appear.
• L.Baboulaz and P.L. Dragotti, Local Feature Extraction for Image Super-Resolution, in Proc. of IEEE International Conference on Image Processing (ICIP), San Antonio, Texas, September 2007.
• L. Baboulaz and P.L. Dragotti, Distributed Acquisition and Image Super-Resolution Based on Continuous Moments from Samples, in Proc. of IEEE International Conference on Image Processing (ICIP), Atlanta, USA, October 2006 (to appear).