Converting ECW Files

[CMcDonald]

Hi, I have hundreds of ECW files and they are huge.  Can I convert to another format which would reduce the file size but not the quality i.e. SID, BMP, TIF, JPG etc.?

Thanks 😉

[RobertBerger]

Greetings,

ECW files generally are already highly compressed. You could try other high quality compression algorithms such as JPEG 2000, JPEG, or if you have the license for it SID. Note that these compressions are lossy, meaning you will lose some pixel information (usually not visually distinguishable). Changing the compression quality (e.g. jpeg 75 to 35) will result in a smaller file on disk but also degrades the visual image quality.
When you say the files are huge, is that the dataset properties size or the actual size on disk? The dataset properties size will be uncompressed size and does not always correlate to the actual size on disk.
Hope this helps.

Robert

[CMcDonald]

Hi Robert, thanks for the response.

Basically I'm not really sure what to do here!  I have 25 raster datasets in our other GIS which we are migrating to ESRI.  The files in these datasets are mainly ECW format but there are also GBM files too (the GBM�??s and ECW�??s aren�??t mixed but make up separate datasets).
My idea to got us up and running in the interim was to convert the GBM�??s to ECW so that way everything would sit in the same format that ESRI can read but the files that make up a dataset (usually 50-100 files) are quite big usually 2-3Mb and by the time you add these up it�??s over 10Gb and we don�??t have enough space on our GIS server left.
I�??m also a bit worried that ECW files are a bit big to sit on the network as we are using Arc 9.2 for desktop clients and a web browser for our Intranet so a bit concerned performance could be affected.
So ideally I would like to look into the possibility of reducing the files to SID/JPG/BMP/TIF format if of course these would result in improved traffic performance and no/little loss to image quality.

Thanks

Chris .

[JohnSobetzer]

You shouldn't expect sids or jpeg2000s to be any better than ecws in terms of quality and file size.  In my experience ecws are the fastest of the three.  Converting them to tifs or bmps would produce huge jumps in file sizes.  I don't know anything about GBMs.

So if you have to reduce file sizes, the only obvious solution I can see is that you will have to increase your compression and there will be some loss of quality, but it may be acceptable.  To do this, however, you need an encoder.  Sids require GeoViewer (which can also produce jp2s) and ecws above 500 MB (that is misleading its a pixel count number) need software from ERDAS.  Both are relatively expensive.  ArcGIS can produce jpeg 2000s although in my experience at least through 9.3.1 they were very slow compared to those produced by GeoViewer.

[GordonSumerling]

Chris,

I would be moving away from ECW's to JP2 format or GeoTIFF format imagery if you have the space. ECW's have two distinct issues.
1) As they are proprietry you need ERDAS software to create them. ArcGIS does not create these.
2) If you want to pass these images to your organization through ArcGIS Server then you need a special license from ERDAS. See the following technical article. http://resources.arcgis.com/content/kbase?fa=articleShow&d=36261

So what should you do? If I had the disk space then I would convert the imagery to GeoTIFF with 75% JPEG compression. If Disk space is at a premium then JP2 with 75% compression. GeoTIFF is pefferred as it is the fastest format to be read by ArcGIS. JP2 is computationally expensive and therefore slower to read. See the following help page: http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//009s0000008s000000.htm

How you convert these is a little more tricky. Smaller ECW's, less than 1gb, can be converted using ArcGIS or FWTools (http://fwtools.maptools.org/). The ECW libraries delivered with these products are free but contain internal size limit which prevents conversion of larger files. To convert the larger files you will need conversion tools from ERDAS.

Cheers
Gordon