Tone Mapping of High Dynamic Range Images for testing square images
This article shows the result of applying different tone mapping operators on HDR images generated by direct volume rendering. We evaluated a total of 24 tone mapping functions.
We input HDR images (given in OpenEXR format) to the tone mapping functions, which generate LDR images in jpeg formats. The input images are generated using Vista, which does direct volumne rendering on data generated using the genraw script, which generates a square volume set which has the dimension 2048*2048*3.
Below are some images generated by different tone mapping functions. The original HDR image has color values ranged from -65504 to 65344.
More information about the individual tone mapping operators can be found in this document.
| Operator | Original Image (OpenEXR) |
Ashikhmin's operator |
 |
| Bilateral filtering |  |
| Chiu's spatially variant operator |  |
| Drago's logarithmic mapping |  |
| Ferschin's exponential mapping |  |
| Ferwerda's scale factor |  |
| Gradient domain compression |  |
| Histogram adjustment |  |
| Horn's lightness computation |  |
| iCAM |  |
| Linear scaling (included for comparison) |  |
| Logarithmic mapping |  |
| Miller's operator |  |
| Pattanaik's multi-scale observer model |  |
Oppenheim's frequency domain operator |
 |
| Pattanaik's visual adaptation for time-dependent display |  |
| Photographic operator |  |
| Photoreceptor-based operator |  |
| Retinex |  |
| Schlick's uniform rational quantization |  |
| Tumblin-Rushmeier operator |  |
| Trilateral filtering | |
| Ward's contrast-based scale factor |  |
| Yee's segmentation-based technique |  |
