You are on page 1of 7

Introduction to Image Stitching

Stitching the image to generate panorama pictures

a
 Panoramic-Image-Creator

This project is about the development of an algorithm for


creating panoramic views from sequential images using
computer vision techniques such as key points extraction,
matching, and image alignment. The algorithm aims to merge
images while overcoming challenges in scale, rotation, and
lighting variations. It is implemented by using Python along
with OpenCV and NumPy.
 Requirements

 Python 3.9.6
 opencv-contrib-python 4.5.1.48
 numpy 1.26.1
 matplotlib-base 3.4.2 (only for
plotting the result)
 How to use it

• An example has been provided in the src/ folder, in which you


can fill it up with your own images paths and provide the
corresponding image list to the function.
• In order to use the implemented function, first, you will need
to read your images and store them in variables. Then, call
computePanoramicView, providing a list with the images read
ordered sequentially from left to right. You will need also to
import the panoramic.py file where the implementation is
located.
• The result variable will be the panoramic view image
constructed. You can pass a list of images it is not restricted to
five. For further information about the approach, optional
arguments and return value.
Algorithm constraints
Sequential Images: the images have to be sequential, otherwise probably the matches
found will be false matches.
Order importance: not only have to be sequential, but also go from left to right in the list
that has to be delivered to the function: list[0] will be the leftmost image (in the real
scenario) and list[len(list) - 1] will be the rightmost image.
Shared elements within adjacent images: between two adjacent images in the list must be
shared elements: a tree, a rock, a field, building, etc. If not, it will be impossible to find true
matches between the two adjacent images.
Planar transformation Assumption: the accuracy of image alignment using homography
depends on the assumption that the scene undergoes planar transformations. These
constraints the use of the algorithm in non-planar or complex 3D scenes. For instance, if
the user takes photos changing dramatically in position and orientation, the algorithm's
accuracy may be affected drastically.
Static scene: the algorithm success is tied on the assumption of a static scene. The presence
of moving elements such as people, cars, or big objects can significantly impact correct key
point matching. Dynamic elements introduce variations in the scene between consecutive
images, leading to mismatches in key points during feature extraction and compromising
the accuracy of subsequent steps. Nevertheless, it could work fine in some dynamic
scenarios.
Benefits of Using Log Transformation

1 Enhanced Visibility
Brings out details in dark areas, improves the visual appeal, and enhances the overall contrast.
Conclusion
Image enhancement techniques like negative transformation, log transformation, and power law transformation
play a vital role in improving the quality of images. Each technique offers unique benefits, making them valuable
tools in image processing and enhancement.

You might also like