Research On The Key Technique Of Panoramic Image Mosaics

Panoramic image has a wide variety of requirements in the widespreadapplications such as ground mapping, medical image synthesizing and virtual scenereconstruction. Thus, the research on acquiring panoramic image has been a hotspotin the image processing field. The common method in the existing panoramic imageacquisition is that the image acquired by a common camera is processed on thecomputer, but there are some defects in the method such as low matching accuracyand unsatisfactory fused effect in the transition effect. Nowadays how to improve thematching accuracy and ensure to not to increase the process time in image mosaics,is an important direction in panoramic image mosaics and also is what we will focuson.After making comparison for the classic feature extraction algorithms, wechoose SURF algorithm as the basic algorithm for the improved method. For the lowmatching ratio in the feature matching of SURF algorithm, we introduce RANSACalgorithm to solve the problem. In the algorithm, we calculate the threshold withmathematics methods, filter the feature points with a moving window for keeping thegood and remove the bad points, and finally obtained the points to be matched.Experimental result demonstrates that the improved method can effectively improvethe matching success ratio and the matching speed.To improve the visual effect of image mosaics, we process the transition areaafter acquiring the feature points. By comparing several classic transition extractionmethods, we choose the wavelet transform algorithm which can well restrain therandom noise in the background. Considering the problem that the extraction resultfor transition area is too small to form obvious shade in the wavelet, we introduce thedistance transform and process the transition area with the wavelet transform afterdetermining the boundaries of the transition area. Experimental result demonstratesthat the image mosaics effect with our algorithm is seamless and nature for the visual effect.Finally, we conduct the panoramic image mosaics experiment on a set of realimages. Experimental result demonstrates that the accuracy of image mosaics is veryhigh within little time.