Image compression based on periodic principal components

In the present paper, the almost periodicity of the first principal components is studied, with the aim of being able to use less information in order to obtain acceptable reconstructions of compressed images. The results of this study show that by working with the periodic principal components of images under analysis, it is possible to obtain an additional reduction to that obtained by using the original principal components. Specifically, it is shown that if the principal components that are considered periodic are replaced by their period plus a trend, it can be said that the reconstruction achieved using these periodic principal components is very close to the reconstruction achieved using the original principal components.