Super-resolution Application In Tracking Camera Calibration

In the process of tracking camera calibration,the resolution if the target images is oneof the most important factor in determining the calibrationprecision.However,influenced by the motion blur,point spread blur,noise and undersampling in the process of acquiring target images,the images will degrade.Superresolution is reconstructing a high resolution image from a set of degrading images.This paper focuses on the theoretical and experimental analysis based on the spatialdomain super-resolution reconstruction technology,concretely on the motion estimationalgorithm and single frame image and the multi-frame images super-resolutionreconstruction technology.The main work include the following contents:1. On the basis of reading a large amount of literatures,an explanation is given on theconcept,principle and the research background. It is determined that motionestimation,single image and multi-images super resolution reconstruction to be as themain research content.2. we set up an observation model of the real scene to the low resolution images fromthe process of acquiring images,and then analyze the feasibility of super-resolutionreconstruction.3. In the part of motion information estimation keren registration and fast blockmatching algorithm are discussed,we adopt the latter as the motion estimation of thepaper.And we take the average signal-to-noise ratio PSNR and the average search pointsas the evaluation measurements,analyze and compare in the basic ideas,algorithmdescription,the performance of the algorithm of four step search algorithm,diamondalgorithm,rood patten search algorithm and the double cross rood search algorithm.4. The single-frame image reconstruction methods are discussed,and we mainlyanalyze algorithms based on interpolation and edge directed.In this paper,we adoptadaptive gradient magnitude self-interpolation based on edge directed algorithm toreconstruct a high resolution image from a single-frame image.5. The basic principle and Implementation steps of POCS,IBP and MAP method areanalyzed,comparing the advantages and disadvantage of the above algorithms in thetarget image reconstruction by experiments and analysis,and put forward the corresponding improved methods in view of the advantages and disadvantages ofeach method.