Notes from "Histogram based prefiltering for luminance ......" Fecker et al.

Fecker, U.; Barkowsky, M. & Kaup, A. Histogram-based prefiltering for luminance and chrominance compensation of multiview video IEEE Transactions on Circuits and Systems for Video Technology, IEEE, 2008, 18, 1258-1267

From Introduction
Some definitions from the Introduction of the paper:
Multi-view video sequence = dynamic light filed [can record moving objects]
Static light field = rigid and non-moving objects are described by images from various positions.
Image based rendering = photo-realistic intermediate views(View Interpolations), where no camera has been positioned during recording.
Simulcast coding = each multi-view scene is coded independently using an existing video codec.(e.g. H.264/AVC)

Luminance and Chrominance differences (Color Differences) are coming from

• Calibration difference between cameras (Leading global change between views)
• Local illumination variations occur because the same object surface can appear differently from different angles. 

Joint Video Team (JVT) uses a block based illumination compensation technique which I think would be similar as I have read in "Learning OpenCV Book" chapter 12 (In preprocessing step)

Local illumination change should be preserved because after image based rendering (view interpolations) that preservation will help to visualize such complex surface characteristics.


Time Constant mapping function:
For each view a separate mapping function (separate histogram) is computed. This is independent from other view.

In Summary:
Mapping Model:[from literature[]Hekstra et al.] From the cumulative histogram of the reference and distorted image the mapping is as follows:
M[v]=u, Cr[u]<= Cd[v]<Cr[u+1]; d refers to distorted & r refers to reference image.
Which means for sum of occurrences in Cd[v] will be mapped to Cr[u] so that Cr[u]<= Cd[v]<Cr[u+1]
Improvement: time constant mapping function & and in RGB color space.
Some improvements are proposed;

•  Instead of taking only image lets take couple of frames for making the time constant cumulative histogram and doing the same mapping as before. 
• The use of RGB color space along with time constant mapping function improves coding efficiency.
• Global disparity compensation does not improve the algorithm
• Illumination compensation [lee et al.] works by modifying the coding process but histogram matching is done in pre-filtering step and thus they can be combined. Combined approach gives the best result!