image-stitching-paper - An Implementation on Recognizing...

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon
Abstract — I implemented a panorama creation tool using [2] and [3]. Although this implementation gives excellent result on narrow panoramas of few images, it has some drawbacks when it comes to wider panoramas of multiple images. Causes for it will be discussed. I. INTRODUCTION HIS report is a detailed presentation of the conducted research on Recognizing Panoramas . Fig. 1 shows some results of implemented Panorama Creator. After presenting the details of the method used for image stitching, the obtained results will be explained. II. OVERVIEW OF THE METHOD The implemented application is a simplified version of the method proposed by Lowe [1] for creating panoramas. Similar to Lowe ’s approach, the application first matches SIFT features between two given images. Then using those SIFT feature pairs , a homography matrix is calculated via RANSAC . This homography matrix is then used to warp one of the images onto other one. Section III below explains the Feature Matching process in detail. Similarly, Section IV talks about the details of Image Matching technique. III. FEATURE MATCHING Feature matching process consists of two steps. First, SIFT features are extracted for two input images. Then, using those SIFT features some interest points are matched between those two images. A. Compute SIFT Features Scale-invariant feature transform (or SIFT ) is an algorithm to detect local features in an image. SIFT features are very well-suited for image stitching problem because they are invariant to scale, orientation and affine distortion. To compute SIFT features of each input images, I have directly used a MATLAB function sift(imageName) from Lowe ’s SIFT Keypoint Detector . Demo software is provided at [3]. B. Match Interest Points After SIFT features of two images are calculated, they must be matched. Those matched SIFT pairs will be used to compute homography later on. To match SIFT features, I have adapted some code from a MATLAB function match(imageName1,imageName2) , similarly from Lowe ’s work. In that code, they accepted two SIFT features to be a pair, if the angle between those features are less than a threshold. They used the dot products of SIFT features’ coordinates to compute the angle. IV.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 04/22/2010 for the course MI IP taught by Professor Vladbalan during the Spring '10 term at Universidad del Rosario.

Page1 / 4

image-stitching-paper - An Implementation on Recognizing...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online