Pixel enlargement in high-speed camera image acquisition based on 3D sparse representations

We propose an algorithm that enhances the pixel number in high-speed camera image acquisition. In high-speed cameras, there is a principle problem that the number of pixels reduces when the number of frames per second (FPS) increases. To suppress this problem, we first propose an optical setup that randomly selects some percent of pixels in an image. Then, the proposed algorithm reconstructs the entire image from the selected partial pixels. In this algorithm, we exploit not only sparsity within each frame but also sparsity induced from the similarity between adjacent frames. Based on the two types of sparsity, we define a cost function for image reconstruction. Since this function is convex, we can find the optimal solution by using a convex optimization technique, in particular the Douglas-Rachford Splitting method, with small computational cost. Simulation results show that the proposed method outperforms a conventional method for sequential image reconstruction with sparsity prior.