[Gdal-dev] 1 and 4 bits image

[Frank Warmerdam]

Sylvain Rioux wrote:
> Hi,
>  
> I am having trouble with 16-colors bitmaps (4 bits) and i was wondering 
> at how GDAL manages this format. I get a bitmap where the pixels values 
> seem to have been slighly shifted and the result is far from what i 
> expected. This problem does also occur with monochrome bitmaps. So far i 
> successfully read 256 colors bitmaps and, gifs, jpegs, and the other 
> raster formats supported by GDAL.It seems clear to me that the fact that 
> the pixels are defined on 1 or 4 bits rather than a byte causes this 
> misalignment.
>  
> I'm using Windows and my current procedure is to use a BITMAPINFO 
> structure, and fill the pixel values buffer with 
> GDALDataset::RasterIO(...). This works for all the raster images i 
> tried. So, to get to the point, i want to know how RasterIO actually 
> fills the buffer when reading 4 bits bitmap.

Sylvain,

Currently it looks as though GDAL will translate TIFF files with less than
8 bits per pixel, and more than 1 via the libtiff RGBA interface. That is
expand it to four 8bit bands.  There is a special case for 1 bit data which
just expands it to 8bit data.  I don't find alot of 2, or 4 bit TIFF files,
but if necessary the tiff driver could be modified to expand 2 and 4 bit
files to 8 bit, like the 1 bit case.

Note that 1 bit files are expanded to an 8bit band that happens to only
have the values 0 and 1.

Best reards,

-- 
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | Geospatial Programmer for Rent