Year: 2023
Type: article
Abstract: Over the past ten years, there has been a lot of research done on image colorization using computer vision and machine learning approaches in order to suit the needs of its numerous applications, such as grayscale image colorization and old and faded image restoration.There are many papers/algorithms have shown to produce plausible colorization of black and white images but they lack the realism and semantics of the living and nonliving things in the images.The idea is that humans can differentiate between different colors and objects and tell what color that particular thing would be.Keeping this idea in mind, we propose a system where one part of the model learns what the objects are and the other learns the colors that can be used for the particular object (humans and animals, included).Next, we propose to use a feed-forward pass model in Convolutional Neural Network which is trained using a dataset of available colorful images.We plan to train the model with semantic segmentation labels, which in turns helps produce much more realistic and fine results for the colorization using deep learning.
Language: English
Topic: Image Enhancement Techniques
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W3137782727