Research And Implementation Of Panorama Image Mosaic Using An Android Phone

Image mosaic refers to stitching multiple images with overlapped regions into a seamless image.Panorama image stitching is one of the hottest research topics in the fields of digital image processing,computer vision and virtual reality.It can solve the contradition that the hardware constrains of imaging device can not balance between wide scene view and high spatial resolution.Up to now,panorama image stitching has been widely used in many applications such as intelligent security,medical image processing and military.Panorama Image stitching is made up of four steps including image rectification,preprocessing,image calibration and image fusion.Among them,image calibration and image fusion are the key steps that determine the success of image stitching.In this thesis,panorama image stitching is investigated from a practical engineering point of view.That is,we research on the panorama image stitching on the basis of Android OS and open computer vision library(OpenCV).Specifically,the main works and contributions are summarized as follows.Firstly,a ORB and RANSAC based panorama image stitching approach is proposed.ORB is a abbreviation of Oriented FAST and Rotated BRIEF,which is a new feature with local invariance.It is an efficient alternative to SIFT or SURF.Thus,ORB-based features are exploited for image calibration.Compared with SIFT and SURF,ORB can be more efficient.In addition,Random Sample Consensus(RANSAC)is exploited to remove those invalid matched feature points,and obtain the overall translation between current image and the previous image.The homography matrix for projection is obtained by computing the feature point pairs.Second,the proposed panorama image stitching approach,which is based on ORB and RANSAC,is implemented by exploiting the OpenCV for Android SDK.That is,it exploits the Android in-built APIs and the OpenCV APIs(Android version).There are two main threads running including the Android GUI thread and the Computer vision thread.The main functionality of the GUI Thread is to capture input images and orientation from the mobile camera and sensor and display the captured images onto the screen.The computer vision thread gets the input image and does all the computations for image stitching.Finally,we verify the proposed approach with Huwei honor 6x,and the scenary images within the campus of Hunan University are used for test.Experimental results show that the proposed Android-based panorama image stitching APP has achieved desirable functionalities and performance.