On Sat, 2009-12-19 at 15:23 +0300, ??????? ?????? wrote: > > First, without HDR input data you won't get a bloom effect but some silly > > "oh-it's-shiny-effect" that has little to do with it. > > Why not? - Bloom - it is effect that simulates moving signal from cells in eyes facets, exposed to bright light, to nearby cells. HDR - it simply more wide color space - not [ 0, 1 ], but [ 0, max float ] - it then tone mapped to [ 0, 1 ] - so why not to implement it on [ 0, 1 ] color space, instead of [ 0, max float ] - ofcourse the result will be not so precise as with HDR data, but it will work. The effect is supposed to produce halos around particularly bright areas, depending on just how bright they are. Once the brightness has been clipped into the [0, 1] range you don't know whether a value of 1 originally represented a brightness that should produce any noticeable halo at all, or if it should then how bright. The mapping to [0, 1] range loses too much information for it to be possible to calculate the effect after it.
