Overview / Usage

Image stitching is widely used in modern applications, such as the following:

Document scanning, Image stabilization in certain video recorders, High-resolution photomosaics in digital maps, and satellite imagery (e.g. Google Maps), Multiple-image super-resolution imaging, Medical imaging, Video stitching, Object insertion.

This project is a showcase of running image stitching using the most popular open-source image processing framework OpenCV.

Methodology / Approach

The algorithms used in Image Stitching generally do pixel-to-pixel comparisons along with gradient descent to estimate the joining parameters. Distinctive features are gathered from each image are gathered together and projected on a surface along with the stitching lines so that all the images are aligned properly. OpenCV uses the RANSAC algorithm along with keypoint detection and local invariant descriptors extraction to stitch the images.

Experience with OneAPI-DPC++

The Good:

The best thing about OneAPI is the cross-architecture development, which enables us to develop a single codebase and deploy workloads in CPU, Multi -CPU (DevCloud is awesome with 28-cores), GPU, and FPGAs.

The Bad:

The experience was otherwise good other than some glitches in DevCloud (lags, crashes).

Technologies Used

1. OpenCV [https://opencv.org/]
2. Intel OneAPI DevCloud [https://software.intel.com/content/www/us/en/develop/tools/oneapi.html]
3. C++

Repository

https://github.com/forkbomb-666/opencv-image-stitching-oneapi