Research On Key Technology Of Panoramic Image Mosaic

Panoramic image mosaic technology is a set of computer vision,image processing and computer graphics and other disciplines,widely used in many fields of cross processing technology.It can mainly solve the problems that could not directly captured the 360-degrees full view image because of the limitation for image equipment perspective.It is the hot research field of Medical minimally invasive operation,Geologic survey,Aerospace and Military reconnaissance etc.Its application prospect is very extensive.So we research the panoramic image mosaic technology deeply,it has important practical significance and application prospect.In this paper,according to the panoramic image stitching technology based on feature,we analyze the problems of traditional algorithms.And according to the image preprocessing,image feature extraction and matching,image fusion and the global optimization of image stitching etc,we research and analysis of the algorithm realization deeply,and put forward the improvement plan,introduce the new algorithm and new technology,such as the BRAND descriptor and ORB descriptor based on feature matching algorithm,grid feature matching algorithm.And it seeks the optimal solutions to improve the panoramic image stitching algorithm efficiency and operating efficiency.In the image preprocessing,focus on the problem that the uneven of image brightness,high image dynamic range,the image feature information extraction is difficult in image dark portion.In this paper,we proposed the optimal region selection and the correction algorithm based on color constancy of the image information(OCCA algorithm)to solve image information recovery of uneven illumination and uneven color.The image is divided into blocks,and discrete wavelet transform and different scales in the division of the region,it gets the wavelet coefficients for estimating the illumination on different scales in the region.And it selects the minimum illumination estimation error as the final illumination estimation value of the area,so the cycle estimate the value of each illumination region.And according to the effective area of choice,it can estimate illumination value of effective region as illumination estimation value of the whole image.And it corrects image color using the diagonal model.In image registration,the slow speed and poor precision problems caused by a huge number of features points for the traditional image extraction method,this paper introduce two kinds of new feature points extraction method which are the BRAND(Binary Robust Appearance and Normal Descriptor)and improved ORB(Oriented Brief descriptor based on feature extraction)extraction method.Using binary character descriptors,the extracted feature points with scale invariance,rotation invariance,and it can effectively combine the geometric information of the feature points for feature extraction and less sensitive to noise.It solves the problems that the slow speed of more pixels are extracted of digital image,it has good stability.And it proposes cluster analysis based on the scale algorithm to filter mismatch feature points and feature matching algorithm based on Delaunay mesh,it is different from the traditional matching method that considering only geometric correlation between two points.It constructs global matching point pair,and eliminates the false matching points.By experimental verification,it not only has a strong robustness to translation,zoom,rotation,affine of image transform,but also has high accuracy of image matching,the matching accuracy is increased by about 30%.In image fusion,for the appearance of uneven of brightness and color,ghost artifact and seam in image stitching process,this paper proposes an improved method for optimal stitching line,that are the stitching line optimization method based on energy spectrum technology and based on graph cut optimization method.It solves the"incomplete","ghosting" problems of moving objects in stitching process.First of all,it does image bundle adjustment for matching images.Based on classic research on the stitching line optimization,we focus on introduce two kinds of improved stitching line optimization strategy,and use improved Possion fusion technology for the smooth transition of stitching image,eliminates the differences and fight seam brightness etc.Experimental comparison with commercial software AutoStitch,Panorama Maker show,the algorithm in this paper achieved good fusion effect in big scene,uneven illumination,and better in wide baseline static scenario.For the image mosaic cumulative error seriously affects the quality of the stitching image,in this paper,we put forward on the visual quality detection and optimization technology and image patching method based on global optimization.Although the panoramic image mosaic algorithm has improved,it minimizes the loss of image quality,lack of information,image brightness uniformity,ghosting artifact and stitching gap and other issues in image stitching process,there are still some vision quality problems such as lack of consistency of image information,stitching object missing or incomplete,the appearance of redundant objects etc.Through the research of several image quality parameters,it detects the panoramic mosaic images,and uses the line graph optimization algorithm for optimization and compensation of panoramic mosaic images.It uses the gap filling technology based on image pixels fusion to patch the missing pixels and make further efforts to image distortion correction of panoramic mosaic image.And through the reasonable construction of Markov random field,and set the definition of label set energy consumption function method,it completes the spatial-temporal image patching technology of panoramic mosaic image based on global optimization,and enhance the robustness of image patching.Through experiment and analysis,the key technology of panoramic mosaic image in this paper is accurate and effective.In addition,this paper through the contrasts experiments,shows the panoramic image mosaic method in this paper has obvious advantages than traditional mosaic technology in effect,time and quality.