the0nlyWyvern / seam-carving

Using Nvidia GPU to optimize algorithm and reduce time in image processing

Seam Carving - Image resizing with CUDA

Description

• A final project in Parallel Programming Class (University of Science).
• Using Nvidia GPU to optimize algorithm and reduce time in image processing. The main program was written in C and C++.
• All details can be found in colab file.
• Only 1 version was developed.
• Note: This is a demonstration, not a real-world application.

Table of contents

0. Introduction
1. Main idea
2. Convert from RGB to Grayscale
3. Convolution
4. Calculate the energy of each pixels using gradient magnitude
5. Find low energy seams
6. Remove low-energy seams
7. Demonstration with images

Usage

To view pnm images, install IrfanView

There are 2 ways to use:

• Easy way: go to this colab file, that's all!
• Hard way: Download ./seam-carving.ipynb only, it is the same as colab file in easy way. Upload it to your drive and you are good to go.

Screenshots

Before	After

Have not achieved

Small errors when resizing large image, don't know where is the bug

Contributors

• Tran Tuan Kiet (Group leader)
• Nguyen Duc Thang
• Le Thanh Loc

References

Slides from teacher Phạm Trọng Nghĩa

Wikipedia - Seam Carving

18.S191 MIT Fall 2020 - Seam Carving - Grant Sanderson

MIT 18.S191 Fall 2020 - Seam Carving - James Schloss

Convolution - 3Blue1Brown - Grant Sanderson