基于分层级各向异性滤波的图像景深渲染算法

A layered image depth of field rendering algorithm using anisotropic filtering

提出了一种基于分层级各向异性滤波的图像景深渲染算法。基于光学薄透镜的景深成像模型,分析得到了更为准确的景深弥散圈光强分布模型。对于输入的深度图,构建了深度图金字塔,修复了其中的不连续区域,结合弥散斑分布模型,确定了场景中各点的模糊半径参数。根据深度信息对深度图进行分层渲染,每层的滤波核半径与弥散圈半径一致,最终采用分离式的各向异性高斯滤波快速得到渲染结果。与典型的景深渲染算法进行了对比,实验结果表明,所提出的算法具有较高的运行效率,图像渲染结果更接近真实景深效果。算法图像景深渲染视觉质量表现优异,具有较好的适用范围和算法稳定性。

A layered depth of rendering algorithm using anisotropic filtering is proposed.Based on the analysis of thinlens depth of field model,a more accurate light intensity distribution function is obtained.For the input depth map,image pyramid is constructed to fix discontinuous regions and combine it with CoC（Circle of Confusion）model to determine the blur parameters.The image is rendered hierarchically with depth information using separable anisotropic Gaussian filter.The result shows that the proposed method could perform fast with satisfying rendering quality that is similar to real depth of field effect,and has wide application range and good algorithm stability.