2005.10.26 22:06 "Re: [Tiff] Re: checking test images", by Frank Warmerdam

On 10/26/05, ryuki_dono <ryuki_dono@yahoo.de> wrote:

I think so too. In most cases there is no need for accurate data display. If you want to create a simple preview or thumbnails from tiff files the loss through downsampling doesn't matter.

Ryuki,

Well, it does if the image is floating point and the values are between -100 and 3000 and the default scaling assuming inputs in the 0.0 to 1.0 range. In this case default scaling rules will produce a useless image.

Nevertheless, I am not adverse to someone knowledgable implementing default rules for more image types in the RGBA interface. I was previously pointing out that I'm not going to do it, because it wouldn't be of much use in my community. I would also stress that the RGBA implementation code is pretty damn confusing and easy to screw up.

I tried to use the default errorhandler and under normal conditions it would be the best solution. Unfortunately I'm missing JPEG support. You can look

at the errormessages I get for images like zackthecat.tif. While TIFFReadRGBAImage reads unsupported TIFF files, recovers from errors and give back

a NULL, it crashes fatal here. For the moment I stick to the exception throwing.

If there is a fatal crash with zackthecat.tif then perhaps you should be filing a bug report. I just ran "tiff2rgba" on zackthecat.tif and it fails of course because it is old jpeg, but it doesn't crash. So crashing is not the normal behavior.

Best regards,

--

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