Generally, to map the original image into

Generally, tone mapping algorithms canbe classified into twocategories by their functionalities during the imaging process.1) White Balancing:Because of the undesirable illuminanceor the physical limitations of inexpensive imaging sensors, thecaptured image may carry obvious color bias.1To calibratethe color bias of image, we need to estimate the value of lightsource, the problem of which called color constancy 16,18, 21, 40, 41. Using asuitable physical imaging model,one can get an approximated illuminance, and then a lineartransform can be applied to map the original image into anideal one.2) Contrast Enhancement: Contrast enhancement algo-rithms are widely used forthe restoration of degraded media,among which global histogram equalization is the most popularchoice. Other variants includes local histogram equalization42 and the spatial filtering type of methods 11, 14, 27,32, 39, 44. For example, in 32 the fractional filter isused to promote the variance of texture so as to enhance theimage. In 31, a texture synthesis based algorithm is proposedfor degraded media, such as old pictures orfilms. On the otherhand, transform based methods also exist, e.g. curvelet basedalgorithm in 35. In 44, an adaptive steering regression kernelis proposed tocombine image sharpening with denoising.Despite of the abundant literature on image enhancement,including those representatives listed above, two challengingproblems for image enhancement are still not solved. First,howto achieve contrast enhancement while preserving a good tone.The contrastand tone of an image have mutual influence. Be-cause of the complicated interaction,those algorithms merelyaiming towards contrast enhancement or white balancingcannot provide optimal visual effect. Most, if not all, of currentimage enhancement systems divide white balancing and con-trast enhancement into two separate and independent phases,as Fig.1(a) shows. This strategy has an obvious drawback:although tone has adjusted in the white balancing phase, con-trast enhancement may undesirably bias it again. This troublehas been observed in many applications, e.g. the de-hazingalgorithms in 26, 37, 38 achieve contrast enhancement by