一种彩色图像的主基元图模型和提取算法

A model and extraction algorithm for primal sketch of color images

文章针对一种灰度图像的主基元图模型和提取算法的不足,提出了一种彩色图像的主基元模型和提取算法,该算法是建立结构和纹理的统一表达模型,在遵循最小描述长度的基本原则下,利用匹配追踪(matching pursuit)算法,从给定一幅图像中推理主基元图。实验结果表明,彩色图像主基元图模型比灰度图像主基元在表示上更加完备,能表达在灰度图像主基元图中被"漏检"的主基元。

By analyzing the limitations of the model and extraction algorithm for primal sketch of gray images ,a modified model and extraction algorithm for primal sketch of color images is introduced . The presented primal sketch unifies the representation of both image structure and texture .Under the minimum description length（MDL） principle ,the primal sketch is inferred from a given input image by using the matching pursuit algorithm . T he experimental results show that the proposed primal sketches of color images are more compact and meaningful than the counterpart of gray images ,and can represent the missing primal sketches in gray images .