Summary

• In this study, we described twelve color interpolation algorithms used for single-detector digital cameras. We selectively implemented nine of them using Matlab and finally did comparative performance analysis from three aspects of the algorithm : subjective quality of reproduced images, objective quality (MSE) of reproduced images and computational cost.

• We made some observations for these algorithms based on our simulation results. In particular, we note that Nearest Neighbor Replication Interpolation and Interpolation using Threshold-based Variable Number of Gradients represent two extremes of interpolation algorithms. Nearest Neighbor Replication Interpolation is extremely computationally efficient but with the worst quality performance. Variable Number Gradients Method gives impressive reproduced image quality but imposes a huge computation penalty.

• As far as the future work is concerned, an algorithm that is both superior in image reproduction and computationaly efficient is still worth pursuing given the fact that none of the existing algorithm satisfies both criteria.