Research On The Algorithms For Depth Form Defocusing

Depth from defocus (DFD) is one of the research focuses in the field of computer vision. It was first introduced by Pentland in 1987. The DFD requires only two images acquired at two different lens settings. Because this technique belongs to monocular vision, it avoids the matching of the feature points existed in stereoscopic vision. Thereby in relation to stereoscopic vision, this method is more practicable in many fields.Our depth estimation method is based on the DFD algorithm proposed by Subbarao. After introducing the basic principle briefly, we discuss two kinds of algorithms. One algorithm applies S transform to image processing. The image function is approximated as a third order polynomial function in small regions. The parameter of points spread function is defined to be the moments of the unitary points spread function. So the parameter is simple to get. This article points out that the spatial-domain operation of the S transformation is independent of different object texture.The result of experiments confirm the same. Another algorithm uses dynamic referencing technique. This method has high accuracy at low computational costs and involving only one unknown constant 'K' which makes the calibration easier and the result distance is regardless of object texture.Employ the Virtual Instrument to carry on experiments of the two algorithms. The experimental result and analysis chart have shown the validity of the algorithm and certain limitation. The images used in the experiment are obtained by aperture of the lens. It can avoid different image magnification between the two images obtained by regulating focus.