Project Name : Image alignment and registration

Introduction

Image Alignment and Registration is a technique where two similar type images are taken into account (viewing angles may be different) and wrap together to align themselves using homography matrix. The applications of image alignment and registration includes:

• Medical Images - To better read different scan images such as MRI scans, SPECT scans and align them together to help specialist provide more accurate diagnosis
• Military - Multiple images of target and aligning them using Automatic Target Recognition (ATR) to improve target recognition
• OCR - Document images alignment in the context of feature based optical character recognition (OCR) to build automatic forms.

In this work, we will study feature based algorithms for OCR - Document alignment by stacking together and also overlaying them.

---

Algorithms

There are different image alignment and registration algorithms used:

• Feature Based Algorithms - Keypoint detectors (DoG, Harris,GFFT), local variant descriptors (SIFT,SURF,ORB) and keypoint matching (RANSAC) - [[OCR}}
• Similarity measure - Cross-correlation, Sum of squared intensity differences, and Mutual Information - [[Medical]]
• Deep Learning Algorithms (Current State of the Art)

---

Images/Program

The images used are (2 scan images and one template)

We will be using two programs

• - To define function for align images using openCV
• - to stack or overlay aligned and template images

---

Results

Scan_01

Scan_02