基于格子波尔兹曼模型的内镜图像高光修复

在内窥镜检查中,由于器官内腔粘膜表面黏液对光源的反射,高清图像中常常存在高光。高光的修复对内窥镜检查和诊断具有重要的临床意义。本文提出一种基于四元数和格子波尔兹曼模型的高光修复方法,首先通过四元数表达高清图像的彩色信息,然后利用四元数旋转操作表达两个像素的色差,并将色差引入到格子波尔兹曼模型的松弛因子设计中,最后利用格子波尔兹曼模型的扩散实现高光的修复。实验结果表明,该方法修复高光效果更好,局部区域细节保持效果提升了50%,颜色保持效果提升了5倍。

Specular image highlight inpainting based on lattice Boltzmann model

During endoscopic examination procedure, due to the specular reflection of the mucus on the surface of organ internal chamber, highlights often appear in high definition images. Specular highlight inpainting has crucial clinical significance in endoscopic examination and diagnosis. A specular highlight inpainting method based on quaternion and lattice Boltzmann model is proposed. In the first step, the quaternion is used to express the color information of the high definition images. Then, the quaternion rotating operation is adopted to express the chromatism of two pixels. Furthermore, the chromatism is introduced into the relaxation factor design of lattice Boltzmann model. Finally, the diffusion of lattice Boltzmann model is used to repair the highlights. Experiment results show that the proposed method has good image highlight inpainting effect, the detail preservation ratio of local region is increased by 50%and the color-preservation effect is enhanced by 5 times.