【摘要】 在马尔可夫随机场纹理模型的基础上,提出一个基于金字塔结构的动态多尺度标记随机场模型,应用这个模型实现纹理图像分割.算法应用由粗至细的处理策略,实现了图像的区域紧凑性,克服了一层GRF标记场结构的运算复杂性,减少了运算时间,并且能够获得较好的分割结果.更多还原

【Abstract】 In the framework of Markov random field texture models. The paper proposes an unsupervised texture segmentation approach based on a variable pyramid label structure modeling. This approach applied a processing strategy that from coarse to fine realized region compactibility and overcome the computational complexity of single Gibbs Random Field label structure, decrease computational time, and acquired a better segmentation result.更多还原