中文摘要：

三边滤波器是在双边滤波器的基础上引入了梯度信息,能够很好地保留图像细节,但其时间开销较大.为此,提出了一种基于亮度分层的快速三边滤波算法.通过对三边滤波器中的梯度滤波阶段进行加速,把图像按照像素亮度值分成多层,并对每一层计算2幅中间图像,可在中间图像上进行高斯滤波得到梯度滤波的结果.实验结果表明,使用该算法进行高动态范围图像的色调映射,速度较原三边滤波器提高约10~15倍,而获得的低动态范围图像的视觉质量接近于原三边滤波器.

英文摘要：

By introducing a gradient field into the bilateral filter,the trilateral filter favorably preserves image details at the cost of high computational consumption.In this paper,we present a fast trilateral filtering algorithm based on intensity-layer decomposition.The proposed approach accelerates the gradient filtering in three stages： decomposing the image intensity into several layers,computing two intermediate images for each layer,and performing Gaussian filtering on these intermediate images.Experimental results show that our approach is about 10 times faster than the standard solution in the context of tone mapping,and achieves similar image quality.