| Image mosaic technology has been widely applied in medicine, remote sensing, mili-tary, virtual reality and digital image processing and so on, because it has effectively solv-ed the contradiction between the wide field of vision and high-resolution. In order to realize a panoramic mosaic image quickly and effectively, this paper relies on the Northern Night Vision Research and Development Center in Nanjing. On the basis of having systematic studied and investigated SIFT algorithm, firstly, the differential gaussian scale-space was constructed by gaussian filter. The detection and the number control of feature points in the scale space were completed with the method of adaptive thresholding. The gradient magnitude and direction of the feature points were calculated by the use of CORDIC. The method of nearest and second neighbor ratio was used to strike matching points. The SIFT algorithm was achieved in the FPGA hardware platform. Then the image data acquisition and high-definition video output were achieved through the I2C bus configuration to Ov9715 and TFP410. Finally a panoramic image based on SIFT registration algorithm was completed in the FPGA hardware platform.According to the the analysis of the results, By inputing two images with changes of ranslation, rotation, scaling and illumination to match, the accurate rate is over 85%. The image mosaic system has good robustness. For an image with resolution of 256 * 256 and the number of feature points controlled within the range of 80 to 200, the system completes the match in less than 5 ms. The image mosaic system has good real-time performance. In addition, the utilization of resources has greatly reduced as a result of the design with Top-down modular structure and pipeline method.The image mosaic system meets the requirements and has strong application value. |
