基于共焦图像序列的自适应聚焦度量场景深度估计方法

Adaptive focus measure method to scene depth estimation based on confocal image sequence

为了估计场景深度信息,提出了基于共焦图像序列的自适应聚焦度量方法。首先,利用相机阵列获取场景多视角图像,用相机阵列模拟一个虚拟孔径相机并合成共焦图像序列。接着,采用自适应聚焦度量方法计算共焦图像序列中每个像素点的聚焦度量值并找出最大度量值作为清晰点。最后,根据相机阵列与共焦图像序列的几何配置估计出场景的深度。此方法提高了场景中深度不连续处的准确度,能较准确地恢复出场景的深度信息。

In order to estimate the depth map of scene, an adaptive focus measurement method based on confocal image is proposed. First, we acquire the multivew images of the scene using camera array. And we use the camera array to imitate a virtual aperture camera and synthesize confocal image sequence. Then, we calculate the focus measurement of each pixel in each confocal image by using adaptive focus measurement method and get the best focused points. At last, using the best focused points we get dense depth map of the scene. As the results show, the method improve the estimation accuracy at the depth discontinuity and give a well depth estimation of the scene.