One-iteration dejittering of digital video images

Abstract

We propose several very fast algorithms to dejitter digital video images in one iteration. They are based on an essential disproportion of the magnitude of the second-order differences along the columns of a real-world image and all its jittered versions. The optimal row positions are found using non-smooth and possibly non-convex local criteria, applied on the second-order differences between consecutive rows. The dejittering iteration involves a number of steps equal to the number of the rows of the image. These algorithms are designed for gray-value and color natural images, as well as to noisy images. A reasonable version of these algorithms can be considered as parameter-free. We propose specific error measures to assess the success of dejittering. We provide experiments with random and structured jitter. The obtained results outperform by far the existing methods both in quality and in speed (the ours need around 1 second for a 512 × 512 image on Matlab). Our algorithms are a crucial step towards real-time dejittering of digital video sequences.