Researcn On Digital Image Filtering In Frequency Domain With High Performance

With the advent of the4th generation commucation times, the information transmission speed is faster, and the information gathering needs to be more quick and accurate. There’s an increasing demanding for the real-time image processing. Traditional computer has bottlenecks on high-speed signal processing mainly because of its serial execution and instruction stream driving, and these bottlenecks can only be solved by hardware. FPGA is one of the widely used programmable logic devices, which is rich in logic resources and convenient to adopt parallel and pipeline operation, in order to achieve the speed enhancement with the expense of resources. In sum, FPGA can provide a useful solution on real-time image processing.Compared with the space-domain filtering, the frequency-domain filtering is more intuitive and there is no edge problem, which may be more effective for some image characteristics. However, the frequency-domain filtering also has some weaknesses, such as complex calculation, long processing time, and integrated filtering. In order to perform real-time frequency-domain image filtering method and improve processing speed, we design a high-performance digital image frequency-domain filtering system.This paper optimizes system structure and FPGA implement based on traditional image frequency-domain filtering solution.In terms of system structure, different from the traditional image frequency-domain filtering based on FPGA, we firstly integrate2D FFT, image frequency-domain data and filter dot product,2D IFFT together. In particular, the previous reverse transformation and the shift of zero-frequency point inside FPGA can be completed outside FPGA, in this way, we can reduce the amount of calculation, and save FPGA resources as well as processing time. Meanwhile, we add the zero-padding operation, which makes the system be able to handle image with unregular sizes flexibility.In terms of FPGA implement, by analyzing data dependency and resource bearing ability, we build an efficient implementation architecture, in order to increase the2D FFT processing speed, dot product and2D IFFT processing speed, and thus decreasing the total processing time of system.Compared with the traditional image frequency-domain filtering system, the digital image frequency-domain filtering system designed and implemented in this paper has some distinctive features, such as short processing time, easy handling image with different sizes. We provide a better solution for the application of digital image frequency-domain filtering in real-time digital image processing.