Research On Fast Carrier Interference Reconstruction Method With GPU

Carrier interference reconstruction is one of the most important technologies in optical analysis system combining computer general calculation technique and optical interference processing.Conventional approach is using the optical CCD camera to get the target image which will be dealt with later.Then we achieve a lot of algorithms on CPU.The system is simple,but the calculation process is complex with large amount of computation and non-real-time.Recently,domestic and foreign researchers focus on the technology of parallel computing on GPU.Using GPU to accelerate image processing is important for building the optical information processing system with real-tine,high efficiency,and precision.The aim of this project “Research on fast carrier interference reconstruction method with GPU” is to composite the advantage of parallel processing method on computer.In parallel way,we achieve a lot of algorithms on carrier interference image processing.Besides,we provide a theoretical basis and preparatory work for efficient and real-time optical processing system.At first,this paper introduces advantages and disadvantages of measuring object surface information using carrier interferometry,and discusses GPU parallel computing power.Then we introduce the basic theory of carrier interference image,including the form of the carrier interference image,phase recovery and so on.On this basis,the paper further studies that using CUDA platform,how to implement the fast optical image reconstruction algorithms with GPU,including accelerating space and frequency domain of carrier interference image pre-processing algorithms and phase unwrapping algorithms.The main topics of this project are as follows:(1)Carrier interference image preprocessing algorithm based on GPU.We use GPU to accelerate image processing algorithms in the space and frequency domain,including binarization,histogram equalization,median filtering,frequency domain high-pass and low-pass filtering.At last,we use CUDA to implement these algorithms,and verify the ability of GPU acceleration through experiments.(2)Phase unwrapping algorithm based on GPU.Phase unwrapping is the key technology in optical information processing,which has significant value in the phase recovery and surface topography measurements.This paperpresents a phase unwrapping algorithm based on GPU,and implement the phase unwrapping algorithm using CUDA.The experimental results show that this method can effectively improve the speed of the phase unwrapping.(3)Design of the optical image processing system.The author analyzes the actual problems in the optical image processing algorithms.Using Windows XP system and Visual Studio 2008 software development platform,the application of commonly used optical image processing algorithm is designed.In addition,user interface is made so that it is easier for user to use this software.this technique can be widely applied in industrial production,military,medical imaging and many other fields.