Image Stitching Research Paper

3154 Words13 Pages

Abstract—Image stitching is used to integrate information from multiple images with overlapping fields of view in order to produce a panoramic view with all the contents fitted into a single frame.Image stitching literature shows that image stitching is still a challenging problem for single and panoramic images. In recent years many algorithms have been proposed widely to tackle image stitching problem. In this paper we present a detail review of all the recent approaches proposed to tackle the image stitching issue. In addition we also discuss the image stitching process for the understanding of the reader.
Keywords—Panorama, Image stitching, Multiple-Constraint Corner Mapping.
I. INTRODUCTION

Image stitching is a sub branch of computer …show more content…

IMAGE STITCHING APPROACHES

Image stitching is the process of combining two or more different images to form one single image. On a broader scale there are two main approaches for image stitching.

• Direct techniques
• Feature-based techniques

The direct techniques work by directly minimizing pixel to pixel dissimilarities. And, the feature-based techniques work by extracting a sparse set of features and then matching these to each other.

3.1. Direct Techniques

The direct technique depends on comparing all the pixel intensities of the images with each other. Direct techniques minimize the sum of absolute differences between overlapping pixels or use any other available cost functions. These methods are computationally complex as they compare each pixel window to others. They are not invariant to image scale and rotation.
The main advantage of direct methods is that they make optimal use of the information available in image alignment. They measure the contribution of every pixel in the image. However, the biggest disadvantage of direct techniques is that they have a limited range of convergence [5].

3.2. Feature-based …show more content…

Global alignment

The most relevant technique is bundle adjustment, which is a photogrammetric technique to combine multiple images of the same scene into an accurate 3D reconstruction. The aim of this step is to find a globally consistent set of alignment parameters that minimize the miss-registration between all pairs of images. Initial estimates of the 3D location of features in the scene must first be computed, as well as estimates of the camera locations. Then, bundle adjustment applies an iterative algorithm to compute optimal values for the 3D reconstruction of the scene and camera positions, by minimizing the log-likelihood of the overall feature projection errors using a least-squares algorithm [17].
In order to do this, we need to extend the pairwise matching criteria to a global energy function. Once we have computed the global alignment, we need to perform local adjustments such as parallax removal to reduce double images and blurring due to local mis-registration. Finally, if we are given an unordered set of images to register, we need to discover which images go together to form one or more panoramas [12].

4.5. Blending and

More about Image Stitching Research Paper

Open Document