Interpolation is one of the fundamental tasks in image processing. While many interpolation techniques have been proposed for that task, most of them comprise an averaging effect and therefore not optimized for texture image
Interpolation is one of the fundamental tasks in image processing. While many interpolation techniques have been proposed for that task, most of them comprise an averaging effect and therefore not optimized for texture images. In this work, we address the interpolation of texture images using an autoregressive second order wide sense Markov model, utilizing its ability to characterize the correlation among neighboring pixels in an image and thus to take an advantage of the statistically repetitive nature of texture images. Specifically, we explore these characteristics within color images and the extent of their applicability in different color spaces. Finally, we propose several robust methods for color texture interpolation. These methods jointly present an extensive generalization of a single-channel previously presented interpolation model, and outperform currently available methods both visually and according to a suitable fidelity criterion. Furtheremore, we show how obtained Markov models emphasize the correlation between intensity channels in color images.
Block diagram
Operations research
Black and white images interpolation results
Colored images interpolation results