Real-Time Super-resolution with an application to Video enhancement
The high density pixel sensors of the latest imaging systems provide images with high resolution, but require long exposure times, which limit their applicability due to the motion blur effect. Recent technological advances have lead to image sensors that can combine in real-time several pixels together to form a larger pixel. Larger pixels require shorter exposure times and produce high-frame-rate samples with reduced motion blur. In this work, we propose ways of configuring such a sensor to maximize the raw information collected from the environment in real-time, and methods to process that information and enhance the final output. In particular, a super-resolution based approach, for motion deblurring on an adaptive image sensor, is proposed and investigated that can operate under real-time constraints. An overview of the proposed system is shown below:
The figure below shows an example of a reconstruction of the car sequence (synthetic). The image on the left is the motion-blurred output of a traditional image sensor, the image on the middle is the result of applying bicubic interpolation to increase the resolution of an image, where the image on the right is the result of the proposed hardware system (super-resolution based) under real-time constraints.
Real-time Ego-motion estimation for an unmanned plane
This project focuses on the development of a system that estimates the ego-motion of an unmanned plane using a single camera. Using the video feed captured by the camera, information regarding the translation and rotation of the plane is extracted. Issues the need addressing are the complexity of such module, the imposed real-time constraints, and robustness to noisy video feed.
Key to the succesful estimation of the ego-motion of the place is the selection of a set of feature points, from which the ego-motion parameters can be extraceted. In our recent work  we demonstrate that by imposing appropriate distance constraints in the feature selection process leads to a significant increase on the precision of the ego-motion parameters. In , a real-time FPGA-based ego-motion system is presneted using the above distance constraints leading to a system that achieves high precision in the ego-motion parameter estimation, meeting at the same time the hard real-time constraints.
A photo of the UAV that was used for the collection of the data is shown below, along with a sequence of video frames captured by the on-board camera. The data used for the evaluation of the proposed system can be found here .
 M. Angelopoulou and C.-S. Bouganis, “Feature Selection with Geometric Constraints for Vision-Based Unmanned Aerial Vehicle Navigation,” in IEEE International Conference on Image Processing (ICIP), Sep. 2011, pp. 2357 –2360.
 M. Angelopoulou and C.-S. Bouganis, “Vision-Based Egomotion Estimation on FPGA for Unmanned Aerial Vehicle Navigation,” in IEEE Transactions on Circuits and Systems for Video Technology (to be published).
- Engineering and Phyisical Sciences Reasearch Council (EPSRC)
- Technology Strategy Board
Dr. Theocharis Theocharides, University of Cyprus, Department of Electrical and Computer Engineering, Machine learning acceleration using reconfigurable hardware, 2012
Keynote in the "Workshop on Modern Circuits and Systems Technologies"Title: "Computing with FPGAs", Thessaloniki, Greece, 2012