how to fit gamma value?

i have 24 digitally created image with specific RGB values. i captured from this images and saved then as captured images. now i have 24 target images in 64*64 dimension and also 24 input images in size 64*64 too. i have this formula: Y=c*X^gamma-b now o used lsqnonlin to find values of gamma,c,b

but after finding these values when i do gamma correction my algorithm fails. here is the gamma corection equation: Ycor=((X-b)/c)^(1/gamma) but it doesnt working. all images corrapted after correction gamma of input images and arent like the target images. what should i do?

[Image Analyst]

"Digitally created" seems to be the opposite of "captured". Did you capture these with a digital color camera, or did you synthesize these images (made them up)? Also, what is X? Is it the mean R, G, B, or L, A, or B value of the color? Or do you think it's the whole image array? Where exactly did the "target" images come from, and where exactly did the "captured" images come from?

[Afsaneh]

digitally created means i knew RGB of specific color and created them in photoshop. and captured one means i captured from printed Image (print of that digitally created one). i am going to undrestand gamma value of my cammera.

Answer 1

[Walter Roberson]

I suggest you im2double(), work with the doubles, and im2uint8() afterwards. Otherwise you need to be careful about whether any given operating is dealing with uint8() and producing uint8 quantized values (which, for example, cannot go negative.)

[Afsaneh]

dear Walter, using im2double made better result. but not the desired result. can i send you my code?

[Walter Roberson]

My back is sore and I am going to sleep soon. Don't send me the code, post it.

The Gamma FAQ might help; see http://www.poynton.com/GammaFAQ.html