Unified framework for colourising corrupted image

Abstract

Colourisation is a process of adding colours to greyscale image. It is noticed that the colourisation process for corrupted image usually needs image to be completed and then colourised. In this study the authors present a novel scribble-based colourisation algorithm of corrupted greyscale image. The authors construct a block-based bilateral filter (BBF) framework for image completion and colourisation. Distance transform and an adaptive weight selection scheme are introduced to the BBF framework in order to achieve improvements of accuracy and speed. The rationale behind the authors’ algorithm is that image completion and colourisation can be both regarded as missing data recovery problem. Unlike existing sequential image completion method and colourisation method, the authors'algorithm can accomplish image completion and colourisation work under a proposed novel unified framework. Experiments results show the benefits of the authors'method.