Improvement of Spatial Resolution Using Block- Matching Based Motion Estimation and Frame Integration

High frame-rate images typically possess a high temporal correlation between their motion vectors. By use of this characteristic, high accuracy motion estimation techniques for high frame-rate images were proposed in our previous work. Compared with conventional methods, our proposed methods are relatively resistant to spatial aliasing errors. However these methods are still inadequate for resistant to spatial aliasing errors. This paper proposes a new matching based motion estimation method that further improves spatial resolution. The proposed motion estimation method is more resistant to spatial aliasing errors and is suitable for reconstructing a high spatial resolution image using multiple high frame-rate images. The method uses multiple previous frames for motion estimation. First, motion vectors are independently estimated between the current frame and multiple previous frames using the basic block matching method and parabola fitting. Next, a timeaxially-averaged motion vector is approximated from multiple motion vectors using the least-squares method. Finally, the motion vector in the current frame is calculated by the approximation straight line. The results of motion estimation experiments conducted using images with spatial aliasing errors show that the proposed method improves accuracy. Further, the results of super-resolution experiments conducted using high frame-rate images show that the proposed method also improves spatial resolution.