Bin Chen 2008
The image rendered is high dynamic range, meaning that in order to display the image on a low dynmiac range display (with intensity value from 0 to 255), we will need to perform a mapping.
Tone mapping may not be critical for scenes with uniform lighting (the pixel values are close to each other) since we can do some simple scaling to map the pixel value to display range. However, for my scene, where the distribution of the light is very non-uniform, this is extremely important. If the pixel values that are greater than 255 are simply set to 255, then we will see parts of the scene become completely white, and the geomtry features of those parts will disapear. This is visible in the previous rendered images.
After each render, miro exports the high dynamic range image to a binary file, and then I can import using miro try to map the HDR imge to the display range.
For this reason, I implemented Photographic Tonemapping described in . The following pictures demonstrate the result:
Naïve tone mapping vs. Photographic tone mapping
The final image now looks a lot better:
 REINHARD, E., STARK, M., SHIRLEY, P., and FERWERDA, J. 2002. Photographic tone reproduction for digital images. ACM Transactions on Graphics, 21(3), (SIGGRAPH 2002): 267-276.