Method of estimation and prediction in time series of images movement.

摘要

Abstract invention patent: "method of estimation and prediction of the movement in temporal series of images.The present invention is a method of forecasting a time series of images, where, given a time series of images acquired at equally spaced times, obtains a picture of future time, can the method be applied to time series of images of solid objects and / or FL Uidos, providing a modeling of variations of gray levels, displacements,Growth and decrescimentos regions in the images, even if all these events occur simultaneously in the same time series. The processing is made through the following steps. The first (1), the images are processed by wavelet transformed input 2D (2). Then, are processed by the transformed 2D DCT (3).Next, the neural network NARX (4) processes each pixel of the images processed as a sampling of a function, by extrapolating the future value corresponding to the pixel. These pixels generate an image of the future time in the frequency domain, where applying the transformed 2D inverse DCT (5) and (6), 2D inverse wavelet to obtain the image of the future time (7).