Research of bird's-eye panoramic view for vehicle parking

This paper proposed a method to integrate multiple images from a video captured by the camera sticked on the back of a vehicle. The integration can be performed in real time. The image integration consists of two steps: Perspective transform and stitching. First, do camera calibration to get the intrinsic parameters and extrinsic parameters with camera calibration and get the birds-eye view of each image. Then detect the line of each image using Hough Transform. After gained the line histogram, gets the rotation angle between two images using minimum cross-correlation of line histogram. Use minimum cross-correlation of binary images to get the transform pixel in horizontal direction and vertical direction. Do spatial transform for one image to gain the images' coordinate in the same coordinate system. Add two images with different weight to get stitching image.