基于多尺度纹理能量测度的单幅图像深度估计

Depth estimation of single image based on multi-scale texture energy measure

为了更好地描述纹理渐变线索与场景深度间的关系,提出了一种基于多尺度纹理能量测度的深度估计算法。采用具有旋转不变性的纹理能量测度方法捕捉多个尺度下场景图像的纹理能量,再分析各尺度下纹理能量的变化规律,得到用于估计纹理区域相对深度的算法。实验结果表明,该算法对近似同分布的非结构化纹理区域的深度估计具有较好效果。

To adequately describe the relationship between the texture gradient cues and the scene depth,a new approach for depth estimation from a single image based on multi-scale texture energy measure is proposed.First,a rotationally invariant texture energy measure is adopted to capture the texture energy fromscene images of different scales.By analyzing the multi-scaled texture energy variation, the algorithm for estimating the relative depth of two textured region is brought forward.Experimental results demonstrate the effectiveness of relative depth estimation for regions with uniformly distributed and unstructured texture.