Comparison: JPEG, MrSid, TIFF and EVS World Map (Repost)
Isle of Wight - Landsat Image N-30-50_2000
I recently had a visitor to this site request Landsat ETM+ imagery so they might digitize shorelines to create a map. Global Mapper gives me an option to save a screen shot, along with JGW and PRJ files which allows the image to be loaded into it's proper lat/lon grid. I took the image of Isle of Wight and zoomed to a scale of 1:150,000. I knew there would be image degradation, but I hoped it would be of small significance for my digitizing purposes.
Isle of Wight - Image Detail (JPEG, MrSid and TIFF)
I am spoiled. I work with Landsat ETM+ images for much of my Enhanced Vector Shoreline (EVS) digitizing. All three images started at a scale of 1:150,000. The detailed images are zoomed to a scale of 1:12,500, where I typically do most of my digitizing. Anti-alias pixel smoothing was applied to all three images. I know that this algorithm is a "black box" smoothing technique. I happen to prefer it over the clunky pixel format. But that is another comparison.
The pristine central image is MrSid format. This is a proprietary image compression format, developed by LizardTech, that shrinks Landsat TIFF images into manageable file sizes. MrSid does to TIFF format what MP3 does to WAV format. In other words a tremendous reduction in file size. The top image is JPEG. I use this format for all of my EVS-Islands images. It shrinks the image typically to about 200kb or less. The bottom image is TIFF. This format, reputed to give the "best" image, I seldom work with.
It looks as if both JPEG and TIFF images degrade (blurry). If my work did not require extremely precise digitizing, these images, at this scale (1:12,500), would be fine. Between the JPEG and TIFF images, the TIFF image seems to offer the greatest clarity.
Should my visitor use either image (JPEG or TIFF) to do their digitizing? That is their choice. Both formats give one "okay" clarity. On the other hand, when you have used MrSid format, it is tough to recommend anything else. For the digitizing work that I do, MrSid format is the best available.
I am not a photogrammetrist. If you are, I am certain that there are many strong arguments for using TIFF imagery (original and uncompressed) versus MrSid imagery (2nd-generation and compressed). Any time you manipulate imagery you risk degradation and MrSid formatting manipulates imagery. However, I can see that the superior image for my work is MrSid format. I work with limited storage and MrSid formatted imagery fits easily onto my hard drives. My preferred mapping program, Global Mapper, effortlessly handles these MrSid formatted images. Once loaded, I can work at 1:12,500 scale throughout the world using MrSid format. In short, my EVS World Map will be made using MrSid formatted Landsat images.
10/20/2006 - More and more 1-meter resolution or better imagery is freely available. The Isle of Wight, used in the above example, is imaged at 1-meter or better resolution on both Google Earth and Windows Local Live. However, my last line in the previous paragraph still holds true, the Landsat ETM+ 2000 imagery, which is worldwide in scope, is still my choice for constructing an EVS precision World Shoreline Map. I am keenly aware of NGA's Prototype Global Shoreline (PGA) which used Landsat ETM+ 2000 imagery to construct a Global Shoreline. I have also demonstrated in a number of posts that EVS offers a more precise depiction of the world's shorelines
Enjoy!
Hmm, is this the whole comparison?
Posted by
Mateusz Loskot |
5/13/2006 09:03:00 AM
Interesting post. You are working with a slightly different goal than most GIS types, and although I don't feel that same as you do about the LizardTech format, I can see why you would like to use it. Keep in mind however, if you are ever using the reflectance values from the landsat for anything other than heads-up digitizing you will probably encounter serious differences between calculations derived from the different formats.
cheers,
Posted by
dylan beaudette |
10/20/2006 08:36:00 AM
I always like to see people do real tests on their own real data.. A few comments, though --
* Maybe I'm missing something, but if your bottom image is TIFF, it appears to be a lossy compressed TIFF; "normal" TIFF is lossless, and so should look much better than what you posted..?
* Which generation of MrSID are you using, the older MG2 or the newer MG3? And at what compression ratio?
* For the JPEG image also, what compression ratio are you using?
* Finally, you can also consider JPEG 2000. Although not nearly as widely supported in the GIS arena yet, it gives image quality comparable to MrSID (MG3) and is an ISO standard too.
[DISCLAIMER: employed by, but not speaking for, LizardTech.]
Posted by
mpg |
10/20/2006 03:30:00 PM
To mpg,
You got me! I am working with MrSID Landsat imagery downloaded from NASA (http://zulu.ssc.nasa.gov/mrsid/mrsid.pl). I am not sure if I have MG2 or MG3. Seeing that it is a government run website and I have been tapping into this imagery for about 2-years plus, it is probably the older compression ratio? Now, I must admit, I don't know MG2 from MG3. Also, the TIFF formatted image is probably not a true TIFF image. That would explain why it is blurry. A true TIFF formatted image should be sharper than MrSID. Right?
I am aware of JPEG 2000 imagery. The files tend to be quite large for my purposes. I truely love working with MrSID formatted files. My mapping program, Global Mapper, handles them easily. To make the maps I make, MrSID formatted Landsat imagery works great. Plus I can load nearly all of the Landsat Mosaics I might use onto my 160-GB hard drive.
mpg, did this make sense to you? Thanks for the questions and comments.
Mr Minton
San Diego CA
Posted by
Mr Minton |
10/20/2006 04:24:00 PM