meaning of nonzero near_lossless

[arbenede]

What is the meaning of a nonzero near_lossless input param? From some tests I ran it look to be related in a linear way the RMSE in lossless mode. Could you please put a reference to a document where this is explained? More information about the other parameters (like t1, t2, t3) would be very useful as well, thanks.

[arbenede]

I ran more tests and verified that this parameter is actually the Peak Absolute Error (in MATLAB syntax: max(reshape(abs(test_img - ref_img),1,[])). I also noticed that it is limited to [0,255]. If you are compressing uint16 images, this range might be too small. Is this is a limitation of the Charls JPEG-LS library?