[PDFdev] L*a*b Color Space problem -- an empirical solution found

[Peter Persits]

Hi everyone,   I wanted to share with everyone a solution to an interesting Lab color space problem that I discovered empirically, but cannot explain why it works.   The Lab color space consists of three values: L, *a, and *b. L controls the intensity of a color, and (a, b) its hue. The range of L is 0 to 255 and (a, b) -128 to 127.   The trio (0xA0, 0, 0) should produce a gray color, however a 2x2 image defined below produces some sort of cyan hue:   4 0 obj <</Type /XObject /Subtype /Image /Width 2 /Height 2 /ColorSpace [/Lab <</WhitePoint [1 1 1] /Range [-128 127 -128 127] >>] /BitsPerComponent 8 >> stream 0xA0 0 0 0xA0 0 0 0xA0 0 0 0xA0 0 0 endstream endobj Through trial and error, I discovered that the *a and *b values have to be XOR'ed with 0x80, and then it works fine. I tested it on a large number of Lab tiffs. Here is a code fragment to do the job:     for( int i = 0; i < m_pImageObj->m_objStream.m_nCurrentSize / 3; i++ )   {    m_pImageObj->m_objStream.m_pBuffer[i * 3 + 1] ^= 0x80;    m_pImageObj->m_objStream.m_pBuffer[i * 3 + 2] ^= 0x80;   }   If someone could offer an explanation why this works, I would appreciated it very much.   (BTW My experiments show that the WhitePoint parameter does not affect anything).   Peter Persits Persits Software, Inc. http://www.persits.com