Research And Implement Of Fisheye Correction

Fisheye image is captured with fisheye lens. comparing with the normal image, fisheye image has a larger viewing angle and more amount of image information. The device with a fisheye lens is able to obtain larger viewing angle. So larger field can be obtained with less fisheye cameras than the normal camera. And the image acquisition process is simplified, and the hardware resources is saved as well. Because of these features, the fisheye lens in the video conferencing, security surveillance, intelligent transportation and other fields have broad market prospects. However, images taken with a fisheye lens have distortion, which not only affects the effect of the human eye, makes image recognition results poor, and reduces the usefulness of the fish-eye lens. Therefore, in addition to the field of photographic art, the fish-eye images must be after correction for the human eye to observe or to use image recognition technology to expand the usefulness of the fish-eye lens. However, the market has many excellent fish-eye lens, but for the fish-eye image correction, there are many defects. As the fisheye correction must occupy much computation resources, though accurate correction can be achieved with many software programs, usual processor run the correction at a low speed. Excellent processors have a high cost and high power consumption. And hardware solutions may achieve a high speed. But they lacks flexibility, or the system structure is irrational. Existing hardware and software solutions have their problems, so this paper focuses on a reasonable fisheye correction system program. the program improves system flexibility with software resources, improves the speed of fisheye correction with the speed advantage of hardware, save the system hardware resources and lower power consumption.This paper analyzes the basic principles of the fisheye image correction and various characteristics of calibration models. And on that basis, this paper compared various model constraints and posed the process of fisheye correction. And system solution is posed in the end. There are two key technologies in the solution: 1) the use of the repeatability of the coordinates calculated by the software pre-calculated calibration coordinates, and the relevance of the use of coordinate data, compress coordinates 2,500 times with a variety of methods, and then coordinates are stored in the hardware module's internal memory, not only use less storage hardware instead of a more complex calculation, and data within the memory module in the hardware also accelerates data access speed. 2) During coordinate calculation it's easy to predict the sequence of loading image blocks in a fisheye image. In order to reduce the bandwidth loading pressure on the system buffer, this paper presents a loading sequence optimization algorithm, making the optimized loading sequence reduced about 90%. This paper analyzes the effect of loading sequence optimization and load the image memory block size and the relationship between the number of pages to determine the optimal size of loading image blocks and memory pages. And o that basis, the hardware architecture has also been simplified. Finally this paper presents the system simulation and test results, verified the effectiveness of the system.