Colour transfer between images

Abstract

Colour transfer plays a crucial role in image processing, serving key functions such as the automatic colourization of grayscale images and modifying the mood of images or videos through colour alterations. This report delves into and critically evaluates two colour transfer techniques: the method developed by Reinhard et al., which conducts a statistical analysis of colour properties to facilitate colour transfer based on mean and variance, and the N-dimensional probability density function (PDF) transfer method. A thorough assessment of both algorithms is conducted, examining image outcomes, and employing various metrics to gauge performance. Additionally, the report investigates the application of neural style transfer, utilizing the pre-trained Convolutional Neural Network (CNN) model VGG19, to accomplish style-based colour transfer. This is achieved by synthesizing artistic images that blend content with specific styles, showcasing the potential of neural style transfer in artistic image generation.