Font Size: a A A

Relighting3D Capture And Visual Hull Reconstruction Based On Mapreduce

Posted on:2016-09-07Degree:MasterType:Thesis
Country:ChinaCandidate:H Q ChengFull Text:PDF
GTID:2308330467482347Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
The acquisition and reconstruction of real world scenes has always been theresearch focuses in computer vision and computational photography area. Usingdimensionality reduction, traditional camera gives a realization to the capture ofcomplex scenes by taking pictures. However dimensionality reduction led to theinformation loss, such as depth, time and other dimensions information, making itimpossible to fully capture the reality of the scene. In order to obtain a more completevisual information of the scene, the plenoptic function and many acquisitiontechniques which based on plenoptic function had been invented. The plenopticfunction which proprosed by Adelson in1991, recodes light field information such asthe incident light coordinates, the incident angle, wave length and incident time. Theacquisition technology based on plenoptic makes up for the shortcomings andlimitations of traditional visual acquisition, letting us achieve a more realistic result.This paper focuses on the researches on variable illuminated acquisition and visualhull reconstruction, rasing a series of solutions to the variable lighting acquisition andcontrols, besides it designed a visual hull reconstruction method base on MapReduce.First, Compared with traditional visual acquisition system, the variable lightingacquisition system can simulates more lights so that it can achieves a more efficientlight field acquisition. Second, parallel visual hull reconstruction which based onMapRedcue provides a more rapid reconstruction. The main contents of this paper aresummarized in the following three points:(1) Chapter2has designed a variable light acquisition system. The whole system isa7meters heigh and10meters wide hemispheroid. There three main part of thesystem:the light unit, the acquisition unit and the control unit. The light unit consisting2000LED lights, can simulate200different illumination angles. The acquisition unitcontains a annular camera array composed by20cameras to meet the requirements ofmulti-angle acquisition. The control unit is responsible for the precise control andsynchronous of the lighting unit and the acquisition unit.(2) Chapter3has designed a light field acquisition program based on illuminationmultiplexing. Different from single light source acquisition, illumination multiplexingacquisition gets the caculated single light acquisition by demultiplexing theillumination multiplexing acquisition results. Compared to single light source acquisition, the illumination multiplexing one can solve many problems under singlelight source acquisition, such as too dark and low signal to noise ratio,resulting inhigher quality image acquisition.(3) Chapter4shows the researches on parallel visual hull reconstruction whichbased on distributed computing.1)First, based on MapReduce architecture, adistributed parallel visual hull reconstruction system has been invented whichsignificantly reduce the visual hull reconstruction time.2)Second, a custom distribuedvisual hull reconstruction system has been designed and implemented. The system ishightly customized so that to suit the hardware, the algorithm and the softwarecharacteristics. Load balancing and fault tolerance mechanisms have also beenenrolled into the custom system. The most important is that the custom distribuedvisual hull reconstruction system can achieve a excellent performance.
Keywords/Search Tags:Variable Illumination Acquisition, Light Illumination, Visual Hull, Distributed computing, Big Data
PDF Full Text Request
Related items