Investigating the Development of Colour Interpolation Techniques for GIMP

Abstract

In this study we investigate recent automated greyscale image colourisation techniques using a custom developed GNU Image Manipulation Package (GIMP) plug-in. This process involves the transfer of colour between a compositionally similar source colour image and destination greyscale image. Emerging colourisation techniques offer an automated solution to the colourisation problem.

We have implemented a GIMP plug-in that was used to critically evaluate the colourisation process. A complete analysis was performed on the images by varying parameters to find the best balance between the neighbourhood size, sampling technique, weighted ratios of pixel luminance, standard deviation and mean. The plug-in developed offers an integrated greyscale image colourisation capability to the GIMP suite, and brings the power of automated colourisation to the GIMP user community.

We discovered that a trivial scan line uniform sampling technique produced comparable if not slightly better results than more computationally expensive techniques based on pixel neighbourhood statistical analysis. We conclude that it is important for images to correlate in composition and size.

Participants

Technical Reports

[1] Vera Kukic. Investigating the development of colour interpolation techniques for GIMP. Technical Report Honours Project Report, Virtual Reality Special Interest Group, Computer Science Department, Rhodes University, Grahamstown, South Africa, November 2003. [DOC] [PDF] [BibTeX]

[2] Vera Kukic. Image plug-ins for GIMP. Technical report, Virtual Reality Special Interest Group, Computer Science Department, Rhodes University, Grahamstown, South Africa, November 2003. [PDF] [BibTeX]

Images