View Synthesizing for a Large-Scale Object in a Scene

Abstract

A robust method for panorama view reconstruction of a scene is presented. The images of a scene are acquired by moving a camera to multiple viewpoints. We present a robust method of panorama synthesizing based on image mosaicing approach. The edge detection and feature points extraction are performed for each image. The corresponding feature points between two successive images are estimated and the between these images is computed. These two images are integrated based on the different of minimum threshold values between them. After that, the full-view of a scene is reconstructed by merging the successive integrated images by developed image mosaicing approach. This research describes how to establish feature correspondences between images accurately and effectively. Image registration technique provides an initial estimation for establishing feature correspondences of point features. The linear solution with the reliable correspondences makes the computation of the geometric transformation between two images.