I begin to work on the next operation ImageBufAlgo::contrast.
It is dead simple and yet brings different things than "over".
It has one input image, one float parameter, does not
care about alpha channels, works with both grayscale
and color images, etc.
I am working on input validation and I need your feedback on
what is a grayscale image and what is a color image. I use
the variables nchannels, alpha_channel and z_channel from
the spec to decide.
1. What is a grayscale image?
These are the possible cases: {G, GA, GZ, GAZ}, depending
on alpha and z channel being specified. G stands for grayscale.
I suggest we handle them all, we only care that there is one channel
that is non-alpha and non-z, we treat it as grayscale and we assume
it is the first channel.
2. What is a color image?
I think a color image has exactly 3 non-alpha and non-z
channels. I think we should not care if alpha or z channel
are specified, and we should assume RGB channels hold
the first three positions, 0, 1 and 2.
Summary: I propose this check in ImageBufAlgo::contrast
to remove the bad cases. A is the one input image.
int has_alpha_A = (specA.alpha_channel >= 0);
int has_z_A = (specA.z_channel >= 0);
int non_alpha_z_channels_A = specA.nchannels - has_alpha_A - has_z_A;
if (non_alpha_z_channels_A != 1
|| non_alpha_z_channels_A != 3) {
return false;
}
Then in the function where the algorithm is, and all types are
specialized, contrast_RA, I can have this:
if (non_alpha_z_channels_A == 1){
// algorithm for grayscale image
} else if (non_alpha_z_channels_A == 3) {
// algorithm for color image
}
Stefan
_______________________________________________
Oiio-dev mailing list
[email protected]
http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org
