Automatic Image and Video Colourisation using Deep Learning

This is the project repository for the research paper published in IEEE (in ICSCET 2018)

View Publication

Abstract:

A pivotal area of research among the machine learning and computer vision communities is the Colourisation of monochrome/black and white images. Colourisation is the computer-assisted process of adding colour to a greyscale image/movie. Traditionally, this process required significant user interaction, in the form of placing numerous colour scribbles, looking at related images, and performing segmentation. However, with advancements in technology, automated Colourisation systems have been created. Apart from the aesthetic appeal, such capability has broad practical applications ranging from video restoration to image enhancement for improved comprehensibility. However, these current systems face some major challenges, such as - Colour inconsistency within individual objects, under/over-saturation, and green tones in bright environments. The main purpose here is to eliminate the issues faced by current systems, and at the same time, efficiently colourise Black & White images and videos.