Colour transfer between images

Colour image processing is one the most crucial aspects of image processing, it serves as a method to enhance the visual quality of images. Traditional methods usually focus on altering an image's colour using predefined rules or statistical models. Additionally, it acts as a foundation for...

Full description

Saved in:
Bibliographic Details
Main Author: Huang, QiYuan
Other Authors: He Ying
Format: Final Year Project
Language:English
Published: Nanyang Technological University 2024
Subjects:
Online Access:https://hdl.handle.net/10356/175204
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Colour image processing is one the most crucial aspects of image processing, it serves as a method to enhance the visual quality of images. Traditional methods usually focus on altering an image's colour using predefined rules or statistical models. Additionally, it acts as a foundation for other applications such as medical imaging and object recognition. In this project, we delve into the concept of colour transfer which involves borrowing colour characteristics from reference images. Specifically, we study two well-known algorithms for colour transfer tasks, one by Reinhard et al. and the other by Pitie et al. Next, integrating advanced semantic segmentation techniques by leveraging state-of-the-art models such as DeepLab V3 developed by the Google Research team, along with LangSAM, originally developed by Meta AI as the Segment-Anything Model (SAM) and later adapted into LangSAM by Luca Medeiros and others, we propose a framework for applying colour transfer based on these models. By conducting a comprehensive comparative analysis, we assess the performance and efficacy of each approach, shedding light on their strengths and limitations in real-world applications. Our findings aim to provide valuable insights for future research in image processing.