Research And Implementation On Video Scaling Algorithms Based On FPGA Technology

With the development of the information and multimedia technology, how to display images and videos in monitors with different resolutions become more and more important, so video scaling technology become one hot spot in the field of digital image processing. This paper studied the classical digital image scaling algorithms, compared their advantages and shortcomings in real time video processing systems, then combined the thought of the edge based scaling algorithms and the classical Lagrange interpolation algorithm together to get the improved EBL (edge based Lagrange) algorithm which could not only realize the real-time stepless image scaling but also keep better edge information.This research is mainly focus on the following aspects:1. Studied the traditional non edge based image scaling algorithms, and implemented these algorithms using MATLAB, then analyzed their performance. Through simulation and analyzation, knew that the traditional non-edge-based image scaling algorithms can realize stepless image scaling, meanwhile, due to its low complexity, it can realize real-time processing, so it is widely used in the real-time video scaling systems, but the shortcoming of this kind of algorithms is that it may produce image edge blur.2. Studied the edge adaptive image scaling algorithm based on the sobel operator, and implemented the simulation using MATLAB, then analyzed their performance. Through simulation and analyzation, I knew the features of this algorithm. This image scaling algorithm detects the edge of the image using the sobel edge detector to divide the image into edge area and non edge area before the scaling processing, and then resizes the two areas by different interpolation methods, so this algorithm can keep better edge information than the traditional non-edge-based image scaling algorithms. But, in order to realize2k times image scaling, this algorithm need to do k times iterative based on2and0.5times image scaling, so the algorithm can’t realize arbitrary ratio scaling, at the same time, due to its higher operational complexity, it can’t be able to do real-time video scaling.3. Combined the advantages of the above two kinds of algorithms, I presented the improved EBL (edge based Lagrange) algorithm, implemented the improved algorithm using MATLAB and then did the performance analysis. By comparing the non-edge-based and edge-based image scaling algorithms, I mastered the thought and the processing step of the two kinds of algorithms, and then combined the thought of the edge based scaling algorithms and the classic Lagrange interpolation algorithm together, this paper put forward a method which choosing the original pixel group from some neighboring groups according to their correlations instead of choosing the original pixel group by nearest neighbor principle, so this method not only took distance but also the edge information into consideration, and then realized the reservation of edge information except stepless image scaling, and finally improved the scaling result.4. Designed the FPGA image processing system to implement and verify the algorithm of this paper. By using Xilinx design suite ISE, I completed the design of the real-time video scaling system, included the design of modules such as the format conversion and the caching of the input video signals, clock control, dynamic configuration, video scaling processor and the caching and the format conversion of the output video signals, and at the same time, implemented the system based on Xilinx FPGA development board ML605, which validated the improved algorithm by this paper. The results showed that this algorithm could accomplish real-time video scaling, and also obtained a good scaling effects.