2003.11.27 15:48 "[Tiff] Re: Bug in LogLuv", by Antonio Scuri

Is there an option in libtiff to convert Your format to 32-bit floats or 16-bit ints other than the 8-bit output I optain currently with libtiffs fallback functionality?

This would help me supporting this format in cinepaint.

Use the pseudo tag TIFFTAG_SGILOGDATAFMT to force the returned data to be what you want:

• SGILOGDATAFMT_FLOAT = IEEE 32-bit float XYZ values
• SGILOGDATAFMT_16BIT = 16-bit integer encodings of logL, u and v
• SGILOGDATAFMT_RAW = single 32-bit unsigned integer with encoded pixel
• SGILOGDATAFMT_8BIT = 8-bit default RGB gamma-corrected values

Notice that you should set this tag before consulting BitsPerSample and SampleFormat, because they will be changed internally when the tag is set.

But in the TIFFRGBAImageBegin function it is hardcoded an SGILOGDATAFMT_8BIT conversion. If you use the TIFFRGBAImage interface you should be aware of that.

Best,
scuri