Reseach On Capturing And Processing System For High Frame Rate Videos

More recently, the video capture and process technology is widely used in both military and civilian fields. In many cases, the object to be captured moves fast; it is hard for ordinary cameras to get details of high-speed objects within such a short time.In contrast, High Frame Rate（HFR） camera is a possible solution. Then this paper proposes a HFR video capture and process system by employing KAI₀340 as the high speed image sensor. To make the system work correctly, there are two types of critical issues. One is designing the timing sequence and driver circuit for KAI₀340; the other is making the system support real-time image processes, like Bayer interpolation and white balance, for HFR videos.First of all, this paper designs the entire hardware system architecture, which is made up of the five modules of Optical Imaging, CCD Driver, Video Analog-Digital Convertor, Video Pre-processor, and Power Supply. Among them, the design of CCD Driver is the key of the system. This is not a trivial task, especially for the design of KAI₀340’s driver circuit, and its relevant control sequences. The hardware-based implementation indicates that the CCD Driver works correctly; KAI₀340 can output200 frames per second under 640*480 resolution, or 1400 frames per second under228*164 resolution alternatively.Secondly, as the original image captured by KAI₀340 in our system is a Bayer color image, each of whose pixels has only one kind of RGB color, not a really color image. This paper designs a novel interpolation algorithm for color recovering.Simulations on Matlab show that it outperforms today’s well-known interpolation schemes significantly. Then this algorithm is implemented on FPGA with Verilog.Thirdly, because the image captured in camera may appear different colors under different color temperatures, this paper illustrates and simulates the well-known white balance algorithms with Matlab. Motivated by the evaluation results, this paper presents an improved white balance algorithm based on the white point detection theory.Simulation results imply that it performs more stable. Finally, this paper implements the improved algorithm on FPGA as well.