Re: 2 questions about using pfstmo(Mantiuk06)

[Rafal]

Hi,

On Friday, January 25, 2013 3:36:41 AM UTC, yuanch...@gmail.com wrote:

First,  when I finish Mantiuk06 tmo operator on my hdr image, I get  R,G,B,Y channels which have been tone mapped. How can I change the RGB channels( float ) to sRGB( 0-255 ) and display it? I can't find a way to generate the same pictures as what are in the paper. 

The images needs to be gamma corrected after tone mapping. Check the examples from the manual page:

pfsin memorial.hdr | pfstmo_mantiuk06 -f 0.5 | pfsgamma -g 2.2  | pfsout memorial.png

              Tone map an image using contrast mapping and save it in the PNG format.

       pfsin memorial.hdr | pfstmo_mantiuk06 -e 0.5 | pfsgamma -g 2.2  | pfsout memorial.png

              The same as above, but use the contrast equalization algorithm.

 

Second, the  Mantiuk06 tmo histogram equalization always can't converge for every hdr image when the tolerence is set to 0.001 (defalut), I do not know why?

The result is not good. Image for example is "Stanford Memorial Church".

I used the function in my program, and the contrast mapping can work.

 

The TMO does converge for me. Please add --verbose, send the command line that you use as well as all the messages printed by the tmo.

Cheers,

Rafal

[yuanch...@gmail.com]

Thank you so much, 

I made a mistake by using histogram contrast. I misunderstood your code. Now it works and the result is not bad. But the convergence still can not be reached when the iteration is set 1000. How large should the iteration be set enough for convergence?

Also, after gamma correction colors^(1/2.2), what should I do next to change float color channels to sRGB? 

color_int = 255*(col- min_col ) / (max_col - min_col )?  The luminance is lower than the image in the paper if I do so.

Can you tell me in which file the code is to change float color channels to sRGB. The tmo papers I have read do not explain this issue. Do I ignore something or not know something?

[Rafal]

On Friday, January 25, 2013 12:07:15 PM UTC, yuanch...@gmail.com wrote:

I made a mistake by using histogram contrast. I misunderstood your code. Now it works and the result is not bad. But the convergence still can not be reached when the iteration is set 1000. How large should the iteration be set enough for convergence?

The best is to post the example image for which the problem can be reproduced. 

 

Also, after gamma correction colors^(1/2.2), what should I do next to change float color channels to sRGB? 

color_int = 255*(col- min_col ) / (max_col - min_col )?  The luminance is lower than the image in the paper if I do so.

Can you tell me in which file the code is to change float color channels to sRGB. The tmo papers I have read do not explain this issue. Do I ignore something or not know something?

One way: 

 ... | pfsgamma -g 2.2 | pfsout output.png

Another way:

 ... | pfsgamma -g 2.2 | pfsview 

then adjust "dynamic range window" and crtl+S to save as an image. 

Rafal