DocumentCode :
744501
Title :
Efficient Image Stitching of Continuous Image Sequence With Image and Seam Selections
Author :
Cheng-Ming Huang ; Shu-Wei Lin ; Jyun-Hong Chen
Author_Institution :
Dept. of Electr. Eng., Nat. Taipei Univ. of Technol., Taipei, Taiwan
Volume :
15
Issue :
10
fYear :
2015
Firstpage :
5910
Lastpage :
5918
Abstract :
This paper presents an efficient image stitching algorithm with better image quality and less processing time for online stitching of a continuous image sequence. Stitching many frames with similar scenes is time-consuming and easily generates inconsistency in the overlap region between frames. Instead of stitching every captured frame, only the dominant frames with significant visual content dissimilar to that in the previously stitched result are worthy of being stitched and can contribute much new information. Although the nondominant frames are not included in the mosaicked image, their homography matrix with respect to the mosaicked image will be roughly evaluated to localize the camera view on the mosaicked image as the region of stitching interest. A particle filter with partitioned sampling is utilized to categorize the stitching label and estimate the homography matrix. Moreover, to substantially increase the quality of offline image mosaicking, a seam planning algorithm is designed to eliminate the unclear ghost effect in the overlap region between dominant frames. The aim is to maximally preserve the visual content while eliminating inconsistencies in the overlap region. The proposed stitching algorithm has been verified in several online and offline experiments to present the efficiency of the overall system.
Keywords :
image segmentation; image sequences; particle filtering (numerical methods); continuous image sequence; homography matrix; image mosaicking; image quality; image stitching algorithm; particle filter; seam planning algorithm; Cameras; Estimation; Image color analysis; Image edge detection; Image sequences; Planning; Visualization; Image stitching; blurred image; ghost elimination; importance sampling; particle filter;
fLanguage :
English
Journal_Title :
Sensors Journal, IEEE
Publisher :
ieee
ISSN :
1530-437X
Type :
jour
DOI :
10.1109/JSEN.2015.2449879
Filename :
7134706
Link To Document :
بازگشت