Re: [Lcms-user] over white color clipping

[Marti M. <ma...@li...>]

>>The *_DBL format specifiers are just a way
>>to make things easier for programmer, and does NOT mean that internal computations 
>>are done in 16 bit.

>Marti,
>I guess, you mean, "... done in floating point" :-)

Right. My mistake. The computations are done in 16 bits of precision, unless
of course, you recompile the engine with USE_FLOAT. But even in this case
temporary results are stored in 16 bits. This is enough for most imaging operations
were visual accuracy is about 11 bits at most.

>I realized that I have a problem with the integer computation, when I 
>attempted to use cmsDoTransform() in the objective function, which I 
>passed to an optimizer for numerical minimization. Since the output of 
>the transform is not continuous, but a quantized staircase, and since 
>the optimizer did compute numerical derivatives with small deltas, it 
>always got stuck in a local minimum, often far away from the actual 
>minimum which it should find.

Yep, quantization is one of the major problems. One have to be very 
careful with errors induced by that, as they tend to propagate easily.  
But  limited precision is necessary for both size and speed.

Regards,
Marti.