Serious problem with quantization during lossy encoding

[boxerab]

Subband stepsize is always less than one in OpenJPEG. This means that quantization is not taking place: i.e. DWT coefficients are being mapped to a larger dynamic range, rather than a smaller one,
since quantization take place by multiplying coefficients by the reciprocal of the step size.

This is very inefficient. I don't think this is a conformance issue, because encoders are free to choose the quantization step. But, I would guess this situation creates a serious hit in performance - perhaps at least two times slower than with a proper step size.

[boxerab]

errrm this turns out NOT to be a problem, in fact. Yes, there is a performance penalty for using smaller step sizes: extra bit planes are coded and then possibly discarded in lossy mode. However, the advantage is giving EBCOT more bit planes to work with, resulting in higher PSNR. Closing.