AWARE SYSTEMS
TIFF and LibTiff Mail List Archive

Thread

2002.04.29 05:40 "Software: Tiff2Pdf, Pdf being Searchable text", by Subhajyoti Banerjee
2002.04.29 06:02 "Re: Software: Tiff2Pdf, Pdf being Searchable text", by Michael Still
2002.04.29 17:40 "Re: Software: Tiff2Pdf, Pdf being Searchable text", by Leonard Rosenthol
2002.05.13 20:43 "Fwd: Re: tsmAPI and LibTIFF", by Antonio E. Scuri
2002.05.13 21:36 "Re: Fwd: Re: tsmAPI and LibTIFF", by
2002.04.30 01:21 "RE: Software: Tiff2Pdf, Pdf being Searchable text", by Kiriakos Georgiou

2002.05.13 21:36 "Re: Fwd: Re: tsmAPI and LibTIFF", by

They definitely don't do this right. The TIFF orientation is for the image as a whole. It is NOT strip or tile specific. If it was, then this concept would also apply to multi-strip images (!) and we would need strip/tile specific processing.

"JPEG assumes a top-left origin and so any TIFF viewer will flip each tile"

This is not correct thinking. This statement applies to a JFIF image as a whole. We must not apply this thinking on partial ("raw") JPEG data.

Note that the statement also implies that the image is always stored consequtivelty from row 0 to row N. This part of the statement is true and is also quite relevant for strip/tile storage! Unfortunately, their implementation does not follow this. They have implemented an ordering where the rows no longer appear consequtivelty (4321|8765|...). The row and column order ordering for strips and tiles should always be (1234|5678|...) or (...|8765|4321).

Think of it like this: The TIFF data stream defines a two dimensional image [0..X, 0..Y]. The TIFF orientation only tells where the origin [0,0] is placed! That's all it does. It should not cause any inter-strip or inter- tile changes.

Best regards,
Peter Nielsen

    Here is their answer. Any comments?

>X-Authentication-Warning: Lomond.AI.SRI.COM: reddy owned process doing

>>-bs Date: Mon, 13 May 2002 13:02:25 -0700

>From: Martin Reddy <reddy@AI.SRI.COM>
>To: "Antonio E. Scuri" <scuri@tecgraf.puc-rio.br>
>Subject: Re: tsmAPI and LibTIFF
>Organization: SRI International - Artificial Intelligence Center
>
>
>Hi Antonio,
>

>In the tsmApi, we use a left->right, bottom->top ordering of pixels

>>which is common in geospatial applications, i.e. the origin is in the >>bottom-left corner not the top-left corner. We correctly set the TIFF

>>flags that specify this so any reader that supports TIFF fully will be >>able to view the raw (uncompressed) tiled files and flip them

>>correctly. However, an issue does arise when using JPEG compression for >>the data in each tile. JPEG assumes a top-left origin and so any TIFF

>>viewer will flip each tile. Because we use TIFF as a storage format for >>our own terrain data, this is not a big concern for us and the data is

>>read back correctly by the tsmApi library anyway (and ultimately by our >>TerraVision terrain visualization system).

>
>Cheers,
>
>Martin.
>
>
>On Mon, 13 May 2002, Antonio E. Scuri wrote:
> > Martin,
> >
> > I noticed that you use the LibTIFF for reading the TIFF images.
> >
> > I don't know if you guys are aware of a discussion going on in

>> > the

> > LibTIFF news about the image tiles used by the tsmAPI. In
> > http://www.libtiff.org) Check for the "Orientation and Tiled

> > or go direct to: images?" thread
> >
> >
> http://www.remotesensing.org/lists/libtiff_archive/thrd17.html#01633
> >
> >
> > The images seem to have the tiles in a reversed order.
> >
> > Best,
> > scuri

>>--------------------------------------------------------------------------
-

>Martin Reddy SRI International, AI

>>Center
>> Menlo Park, CA
>> 94025-3493

>reddy@ai.sri.com                               Tel : (650) 859-6468
>http://www.martinreddy.net/                    Fax : (650) 859-3735