与纹理无关的散焦测距方法的研究

计算机视觉中，景物三维重建的关键之一是从景物图像中获得景物目标到摄像机的距离．在此介绍一种基于散焦图像来计算景物距离的新算法．该方法通过改变镜头的光圈系数得到同一景物的2幅散焦程度不同的图像，用图像信号时域能量值表示其功率谱密度；应用一种动态参照技术，得到模糊参数值C，从而得到景物距离的估计值．这种方法具有标定简单（只有一个值需要标定）、计算量小、精度较高并且得到的估计结果与景物的纹理无关的特点．实验结果表明了该方法的有效性.

Depth from defocus independent of object texture

In computer Vision, the key of 3-D reconstructing is to figure out the depth information of a scene from the images. A new method for depth from defocus images is proposed. Two images of the same scene with different defocus are taken by changing the aperture of the lens. The power spectrum density is expressed by time domain energy. The fuzzy parameter C is calculated using the dynamic referencing technique, so that the depth of Scenery is estimated. This method has high accuracy, and low computational work and only one unknown constant involved. Furthermore ,the resulting distance to object is independent of the object texture. Experimental result proves the effectiveness of the method.