A New Method for Multi-frame Super-Resolution Reconstruction

In this paper, a new method for multi-frame super-resolution reconstruction is proposed. It builds up with three steps: motion estimation, frame interpolation and fusion, deblurring. Based on the pure translations motion model, a simple motion estimation method is firstly proposed. All measured frames are subsequently located to a fixed grid and interpolated using an effective technique called steering kernel regression. Then, the reconstruction frame is obtained by fusion high-resolution pixels selected from all of the interpolated frames. Finally, the reconstructed frame is deblurred at the last step. Experimental results on simulate and real data confirm the effectiveness of the proposed method and demonstrate the Steering Kernel Regression´s superiority to other interpolate methods such as cubic interpolation and the method of.