Soapbox Number 1

Image Compression

Soapbox Number 1
Soapbox Number 2
Soapbox Number 3
Soapbox Number 4
Soapbox Number 5


My first soapbox topic is work related.  Its image compression.  Increasingly over the years, people have been pushing image compression as "a good thing".  Well, that depends on if you know what you are doing, and what the purpose of your image is.  Image compression of just about any kind is ok if you are going to use it strictly for visual type images.  I.e.  Pictures of landscapes, people, pets, trees, things like that... 

The game changes when you are going to use the images for something else, like I do.  I.e.  Landcover classification, biology stress indicators, etc... 

Lossless image compressions like TIFF are just fine, I have no problem with them.  These compression algorithms compress data by looking for patterns in the data that can be compressed.  A simplified example would be if you had an image with 1000 pixels (dots) across, and one line was completely black (DN value 0), an uncompressed image would use 1000 bytes (assuming 8 bit) to store the data, while if you compressed it you could represent it by saying "begin here, and fill with black for the next 1000 spaces", basically which could be represented in as little as 5 bytes.  Trust me, its more complicated than that, but you get the idea, a very high compression ratio.  Of course images rarely consist of large strings of the same color, but anything occurrence of 5 or more bytes in a row would result in a space savings.

My soapbox issue is the lossy compression algorithms like JPG and MrSid.  These algorithms basically achieve their phenomenal compression be throwing away data.  They do this in various ways that is way to complicated to go into here.  Trust me, with lossy compression once you save something and it throws away the data, you will NEVER be able to recover the lost information.  Sometimes its color information, sometimes its texture, sometimes it creates strange little artifacts all over the data when you view it full resolution.

So the rule is simple, if you are going to use the data for anything "scientific", don't save it in a lossy compression.  Obviously, lots of people don't get this, because I keep seeing lots of agencies/companies distributing satellite and aerial photographic data in lossy compression schemes, with no option to request a lossless version.  Ok, people, if you don't understand what the heck you are doing, you shouldn't be in the business.   Does your opinion differ from mine?  If so, to put it bluntly, you are wrong.  Ask me sometime and I will show you the evidence mathematically, graphically, visually, any way you want.  Lossy compression has no place in the "scientific" community.

Home | Soapbox Number 1 | Soapbox Number 2 | Soapbox Number 3 | Soapbox Number 4 | Soapbox Number 5

This site was last updated 02/28/04