Department of Electrical and Computer Engineering, and the Cooperative Research Centre for Sensor, Signal, and Information Processing, University of Queensland, Brisbane, QLD 4072, Australia.
IEEE Trans Image Process. 1998;7(6):925-31. doi: 10.1109/83.679446.
Our noncausal, nonparametric, multiscale, Markov random field (MRF) model is capable of synthesizing and capturing the characteristics of a wide variety of textures, from the highly structured to the stochastic. We use a multiscale synthesis algorithm incorporating local annealing to obtain larger realizations of texture visually indistinguishable from the training texture.
我们的无因果关系、非参数、多尺度、马尔可夫随机场 (MRF) 模型能够综合和捕捉各种纹理的特征，从高度结构化到随机化。我们使用一种多尺度综合算法，结合局部退火，以获得与训练纹理在视觉上无法区分的更大纹理实现。
