Comment by 4ndrewl
17 hours ago
The great thing about TIF was it's extensibility. Flexible (data could be stored as tiles or in stripes), multiple compression options etc.
Well documented spec, easy to bolt on extras either as public tags - GeoTIFFs added projection metadata - or private, for your own needs.
Back in the day, to improve a desktop application's performance I found it was simple to create a custom reader and writer to handle cases where tiles were completely one single colour removing the need to decompress at run time.
Thank you TIFf!
Perhaps the greatest thing about TIFF, but also the most horrible things, and probably why TIFF is mostly historical. It was so extendable that no two programs ever accepted the exact same TIFF extensions. (omitting the war story)
edit: forgot about byte order...
But most of these variations were part of the spec (endianness with II or MM, later magic 43 for bigTIFF 64bit extension). I work with tiff and tiff-derived formats in digital microscopy where its very much not historical. And the alternatives (DICOM supp 145, vendor-specific garbage ... and thats it) are worse.
I quite like the format, the only thing I would change is to have the option not to store directory information in a linked list spread throughout the file but in a simple array. Duplicate it at the beginning and end of the file and you've got resilience too (important in the age of floppies)
We have hundreds of thousands of TIFF files where I work which are scans of questionnaires filled out by clinical trial participants. The one annoying thing is that web browsers don’t natively display them. I did some incredibly inefficient JavaScript bs to decode the pixel data, plop it in a canvas, get a PNG data url from it, and set that as the src for an img element xD (why not just display the canvas? because I was too lazy to manually handle resizes…) good times
This is why some people consider TIFF to stand for “Thousands of Incompatible File Formats”.
TIFF is still very much alive in certain circles, see for example https://cogeo.org/ The format is basically a TIFF file with attached georeferencing information and with the data organized by geographical sector, enabling fast downloads of regional subsets.
The Adobe DNG standard for raw camera images is based on TIFF as well. DNG is used in lots of places, including the raw capture support built into all modern iOS and Android smartphones.
I’ve been using both TIFF and DNG this very week in my work (https://filmlabapp.com), so I was happy to read this post and learn about Steve Carlsen aka Mr. TIFF, whose work we’re still building on 39 years later.
3 replies →
I'm not super knowledgeable about this stuff.. but out of curiosity, what advantage does it have over NetCDF?
I wrote a program processing GeoTIFF data. When I had started this project I chose GeoTIFF mostly b/c i wanted something simple. And I could load them in to Java's BufferImage class and manipulate them that way. But it seems all the pros exclusively use NetCDF and GeoTiffs are for noobs (working with atmospheric science data here)
GeoTIFF does extend "images" to cover more usecases, but a lot of stuff doesn't fit (like say a wind vector) and then you need some other container or metadata b/c you generally have many images. So I get the sense the complexity just ends up being moved elsewhere.
1 reply →
Unfortunately, it doesn't help.
Almost any digital camera RAW format is TIFF inside. And you can see how much kludges good metadata library needs to read all of them: offsets from the IFD, offsets from beginning of file with or without header, offsets from fields in IFD, etc, etc, etc. You take TIFF, you change header to make your format, and then you cannot implement this TIFF properly!
Even DNG (which is tiff inside) is mangled by camera firmware authors!
And the (early) availability of well made library, LibTIFF by Sam Leffler. I used it extensively from 1995 on, but only found out that according to Wikipedia is dates back to 1988!