The Research And Achievement On The Technology Of Image Real-time Deblurring

Image restoration technology as a cross subject in image processing, in the field of image processing, has been the most important and one of the most basic research topics with a strong theoretical value and engineering application value. In this paper, the theory of image restoration is closely combined with the engineering practice. The aim is to satisfy the need of image restoration in real time processing as far as possible. Restoration algorithm is mainly aimed at the contradiction between the protection of the details and the noise in the image restoration process, as well as the rapid improvement of the image restoration processing, the performance of the algorithm and the practicability of the algorithm are improved. In this paper, the research work has two directions, mainly based on the fast recovery algorithm and high-speed implementation. The research mainly contains the fast recovery algorithm and the current mainstream platform.The common degradation model and fast recovery algorithm are studied, and selectively analysis the out-of-foucs the deblurring model of G. Then demonstrate the effectiveness of the algorithm of SeDDaRA. On the basis of this, the SeDDaRA algorithm is used to restore the degraded image effectively. Research Based on probability model of blind image restoration algorithms, analysis of the principles and characteristics of representative algorithms MIA(multiplicative iterative algorithm), then the paper propose the improved iterative model, and prove the effective of the algorithm according to the mathematical model. Aiming at the problem of slow convergence and insufficient estimation of SeDDaRA, a multiplicative iterative method based on SeDDaRA algorithm is proposed. Based on the short time invariant of the target image in the image sequence, a two step fast recovery method for video sequences is proposed. By choosing a better iterative path to achieve fast iterative recovery, the experimental results show that only a small amount of computation is needed to reduce the number of iterations.Based on the fast method of the dark channel, from a broad sense, the image defog also belongs to the problem of image restoration. In view of the existing algorithm’s high computational complexity, the recovery time is long. the paper proposed a restoration algorithm based on double scale dark channel, the high scale dark channel for computing illumination coefficient, low scale dark channel used to calculate the transmission coefficients. On the basis of this, the circuit structure and implementation method of the FPGA resource constraints are proposed for fast recovery processing. Research with the high-level FPGA tool, such as the Impluse C tool. Utilizing the tool to solve the floating-point arithmetic and timing design problems, and effectively making use of the resources of the FPGA. On the platform of Xilinxspartan3, the paper can obtain real-time restoration results. On the basis of the research of the real time image restoration algorithm, the image de fog circuit is realized by using FPGA platform.Based on the research of the OpenCL, the paperthe utilize GPU platform to improve the speed of image restoration algorithm. For a long time, the complexity of the restoration algorithm is an important aspect of the application of the algorithm, so the real-time performance of the algorithm is an important direction in the research of this paper. For the fast recovery algorithm, the paper analyse its computational characteristics, for the two-dimensional FFT and other special functions, to study the low complexity of the calculation method. After the complexity analysis of the algorithm, the computation of nearly 70% of the restoration algorithm is derived from the 2D-FFT algorithm. The Fourier transform of periodic, and the conjugate symmetry of the frequency domain data to reduce the amount of calculation of the redundant data in the 2D-FFT, on this basis, designed for rapid restoration of FFT calculation module for restoration algorithm of GPU accelerated. For the same floating point processing capacity of the platform, the speedup can reach more than 4 times. On this basis, the GPU on the storage hierarchy is analyzed, to explore the bottleneck of transmission bandwidth of PCIE and GPU, as well as the way of memory access, and then find the key to affect the performance of OpenCL program. Then in the analysis of two step rapid iterative restoration algorithm based on, and summarized the characteristics of the algorithm, calculation module of the algorithm of dismantling, found in AMD 7400 platform and the universal miniaturization of GPU platform realization of fast and effective. By taking full advantage of GPU’s computing resources, the algorithm can effectively migrate to the GPU platform, the two platforms to accelerate the algorithm, respectively, can get 6 times and 30 times the significant speedup.