Using automatic greyscale image colourisation as a compression technique

Abstract

This project is about image compression. By using a compression algorithm and an automatic colourisation algorithm we aim to test whether while keeping a constant file size, the quality of our produced image will contain less artefacts at the expense of colour fidelity. JPEG (Joint Photographic Experts Group) image compression is applied to reduce image data by removing redundant information. This may then be stored or transmitted in a more efficient form due to its reduced size. Image colourisation is a process by which colour is added to grayscale images. This is used to add back the colour data which has been removed from the image for storage or transmission purposes. Our proposed method begins by extracting relevant colour information from the input image. This is done to make our proposed method as automatic as possible. Next compression on this image is carried out and the luminance data is stored. Upon reconstructing the image, the luminance file and the colour data are used by the colourisation method to generate the final image. Our results show that while producing a comparable image to JPEG's, it was our colour data that caused our file size to be larger than that of JPEG.