As the next generation of broadcast system, there are many technical difficulties need to be resolved in 3DTV, in which depth estimation is one of the key techniques. To improve the depth estimation precision, a novel depth estimation method based on graph cut is proposed in this paper. A local window with adaptive support weights is used to compute the average value difference of the pixels' luminance, which is a part of data item of the cost function. To resist mismatch caused by brightness difference and protect the edge information, gradients between current pixel and its eight neighboring pixels for each pixel in the reference view and current view are computed also. Then, the pixel-wise depth map is classified into two categories by cross detection, one is reliable and the other is unreliable, which guides the post processing iteratively. Rendering experiments show that the depth maps estimated by the proposed method are more precise than DERS 5.1.%3DTV作为下一代视频广播系统,还有许多技术难点有待解决,其中深度估计是3DTV的关键技术之一.为了获取高质量的深度图,提出基于图割(graph cut)的深度估计方法.该算法在构建能量函数的数据项时,通过对窗口内各个像素赋予自适应权重,引入梯度信息以抑制因亮度差异导致的误匹配问题并保护边缘信息.然后,经过交叉检测将深度图像素分为可靠点与不可靠点两类.对检测后的深度图进行后处理迭代优化,从而提高所获取深度值的可靠性.实验表明此算法估计出的深度图用VSRS绘制虚拟合成视时比标准的深度估计软件DERS5.1可有效提高虚拟视质量.
展开▼