Depth From Defocus Algorithms Research

With the development of computer vision, the two-dimensional information of image can not meet the requirements of research, now one of the most important research goals is to restore the three-dimensional information of images. To restore the depth information from the defocused images is becoming more and more important. Predecessors in this field have already done a lot of interesting research work, and achieved fruitful results. This method requires only two different defocused images captured by controlling the camera parameters, without the use of binocular vision method. This method has no point matching problem, different from the binocular vision method, so it is a valuable alternative.This paper focuses on the issue of recovering the depth information, that is how to get three-dimensional depth information from two-dimensional images. We describe the background information, meaning, advantages, status of depth from defocus. We first make a mathematical model for the imaging system, study the relationship among the parameters of this mathematical model. Then we analyze the imaging process of defocused image, explore the intrinsic relationship between defocus and depth, discuss some of the common depth from defocus methods, such as some methods based on frequency domain, some methods based on spatial domain, and on coded lens. On this basis, we focus on depth from defocus using the heat diffusion equation, proposed DFD method based on the relative blur and edge strengthening, and avoid calibrating camera parameters repeatly by measure depth difference. Experiment become more easy to operate by use of these methods, experimental results also would be a great improvement. Finally based on the study and discuss of this paper, we completed a system of depth from defocus(DFD).The main work of this paper are as follows:We analyze the imaging model of defocused images, study the method of simulating defocus with the thermal diffusion method, presents a depth restore method based on relative blur and edge enhancement. We acquire two defocused images under different camera parameters, simulate the process of defocusing by using thermal diffusion equation, find the relative blur and diffusion parameters of different parts in the images, where the blur degree is different, then calculate the depth information. We study the influence of edge information in blur region division, then acquire better experimental result through the edge enhancement.Depth from defocus is a kind of passive method, the camera parameters must be changed while acquire the defocused image in currently available DFD method, such as change the distance between CCD and lens. It is difficult to change the camera parameters in the actual experimental operation. In this paper we proposed a method to get image and calculate the depth information without changing the camera parameters. With a fixed camera parameters and through thermal diffusion theory this paper proposed a method that greatly facilitates the depth calculation process and simplifies the division of diffusion region.