Research On Stitching Method Of Ink Writing Ink Image

The application of digital image processing in various fields makes people put higher requirements on the image.Image splicing technology is an effective technique to obtain high resolution and wide viewing angle images,instead of expensive panoramic cameras.In the field of virtual reality and panoramic images Widely used.Among them,the image registration is the key step of the image stitching,but also difficult.The feature-based image registration is a hotspot for Chinese and foreign scholars because of its robustness and good adaptability.This paper studies the splicing problem of writing ink.In the brush writing process,a camera is difficult to capture the entire process,so the need for three cameras to capture the entire process,the use of image stitching technology on the three cameras to capture the image want to stitch,you can restore the brush writing process.Since the three cameras are installed at the end of the pen of the brush,the writing process moves irregularly as the pen moves,there is a variety of changes in the images,such as translation,rotation,and zoom,which brings image registration difficult.SIFT operator has good scale invariance and rotation invariance,and the SURF operator is an improvement to the SIFT operator.Therefore,the SIFT operator and the SURF operator are chosen to carry out the feature Detection and matching.In this paper,the performance of SIFT operator and SURF operator is tested by rotating and scaling the brush image.The results show that the rotation invariance and scaling invariance of SIFT operator are better than SURF operator.In this paper,the adjacent frames in the writing video are matched,and the RANSAC algorithm is optimized.The algorithm can remove more mismatched pairs than the previous algorithm.