Study On The Estimate Of Depth Based On Defocus Image

In the computer vision system, measurement of the depth information of objects has been a hot issue .It is the key to three-dimensional reconstruction. In recent years, the depth of object estimation based on the defocus image by more and more attention and research . It belongs to the monocular vision ,It avoided feature point matching question in binocular vision which has not been able to be solved effectively at present .Therefore it has been favored by many areas and it has practical values.Now, the algorithm of the depth from defocusing(Depth from defocus, referred to DFD) requires two or more images ,and it identifies the depth of object information according to the difference of same object under different levels of image defocusing .But the constraints of secondary imaging increases the complexity of calculation ,and reduces the application of real-time. The main work of this thesis is research the estimates of depth of the defocus image based on monocular, according to the application characteristics in practice to reduce the accuracy of the depth information of objects for calculating the speed and real-time.The main contents of this article is based on single defocus image, use the method of image depth stratification to estimate the depth of the object. This article has mainly carried on the following works:Firstly, Analysis of defocus image optics from the theoretical, the relationship between the radius of the circular fuzzy spot on the imaging surface and the optical system parameters .Described two classic defocus image location algorithm: Pentland algorithm and Subbarao algorithm.Secondly, introduce the clarity evaluation function, and use it as the standard when restore the image, strike the point spread function of image. By comparing the experimental data obtained image clarity evaluation function by Laplacian in a small regional. Thirdly, analysis the degradation model of image and its mathematical model, then study the two major problems of defocus image blind restoration process: the selection of point spread function model and filter. Finally, by the simulation experiments, confirmed the feasibility of recover defocus image by point spread functionFourthly ,proposedone new methods that image depth layering, through layering the depth of images,and according to the relationship between the point spread function parameters and object distance, strike the corresponding point spread function parameters under different object distance. Restore the image, then obtained the object distance when the restored image is most clear, finally, according the different object distances to determine the relative position between two points. Finally, by a simulation experiment, we obtained depth information of objects.