We have presented various tone mapping techniques and image metrics used in computer graphics. This is still an open area for further research. There will be better and better display devices in the future, and tone mapping and perception based difference techniques will become more important.
Unfortunately, we can not say which tone mapping technique is the best. There are lots of good techniques for various purposes, but there is no one generally apply-able method. Everyone should choose the appropriate tone mapping method for a particular goal. If visibility and lighting design is crucial in some application, the visibility matching or TR mapping method should be used. Of course absolute units rendering is a necessity in this case. On the other hand, if the lighting conditions in the scene are unusual (e.g. back light) minimum loss or interactive calibration techniques could be just the right choice.
If the correct reproduction of the true colors is important, incident light metering is the clear winner. This method reproduces selected object colors faithfully, even if the average scene reflectance is very low, or very high (and this is not the case with other mapping methods).
Contributions of this thesis to tone mapping techniques are interactive calibration, minimum information and area loss techniques, and incident light metering.
The last part of this work presented a color image difference technique that operates in the original image space. It is more intuitive than metrics that operate in transform spaces, and the color is taken into account more properly. This technique describes also how image difference depends on the viewing distance. This metric is based on the contrast sensitivity function and CIE LUV space, and it corresponds to human perception.
Currently we are trying to find some results evaluation mechanism, that will correspond to human vision.
We are also trying to find a global contrast factor which shall describe the overall impression of an image, based on human vision. Such a factor can be used to evaluate tone mapping techniques, and images in general.