Texture synthesis based on Direction Empirical Mode Decomposition

摘要

Texture synthesis is an essential aspect in virtual reality, making it an important research area in recent years. A new texture synthesis algorithm is proposed in this paper. Using the Direction Empirical Mode Decomposition (DEMD), the inherent direction of the sample textures is calculated and the samples are decomposed into a series of images. Based on the feature value in the direction, every level of the decomposed images is then compounded according to the pattern correlation searching. Composing all levels of decomposed images, a synthesized image corresponding to the direction can finally be generated. The experiments proved that this algorithm is simple, fast and high quality.