Image Compression

Image Compression


Suresh Murthy
HOPE Technologies Ltd., B-1 Geetanjali Enclave New Delhi
Email: [email protected]

It is feasible to work with large images and also serve high-resolution orthophoto mosaics for entire cities, real-estate companies and other non-technical markets using a combination of Compression Technologies.

As imagery becomes more affordable and easier to use, GIS and CAD users are increasingly demanding more powerful imagery capabilities. However, the integration required for mosaicing of a large number of imagery, airphotos, scanned topographical maps and other types of imagery, for various applications the size of image files can become very large. The key issue is to easily use, manage and disseminate an always growing image database with limited equipment and funding. This is an important issue with most organisations. It is now feasible to work with such large images and also serve high-resolution orthophoto mosaics for entire cities, real-estate companies and other non-technical markets using a combination of Compression Technologies like Wavelet compression and Image Web server (IWS) from ER Mapper.

What is a compressed image?
A compressed image is one that has been compressed to reduce the image file size, while still providing an almost perfect version of the original. With some compressed formats, such as the ER Mapper Compressed Wavelet (ECW) imagery format created using the ER Mapper Compression Wizard, you can use a compressed image almost immediatly. This is achieved by automatically selecting and decompressing the portion of the image, at the desired level of detail viewed or printed. There are various compression techniques. The most effective technique is wavelet based, which is a method of processing, quantising and then encoding the image in a way that results in very high levels of compression, with an almost perfect reconstruction of the original image.

Wavelet based compression: Performance gains
In addition to the advantages of smaller files, (which requires lower seek time to access data from CD ROM, DVD ROM or Hard disk drive) compression of digital imagery offers higher performance, because compressed imagery trades slower disk speeds against faster CPU speeds. Compression is “compute bound” rather than “disk bound”, meaning that decompression is limited by CPU speed rather than by disk IO or seek speed. A 10GB image that has been compressed down to a 500MB ECW, the compressed image file can be read as quickly from a CD ROM as the uncompressed imagery can be read from hard disk. This makes distribution and use of imagery directly from CD ROM a practical reality.

Working with large images
The table 1 demonstrates the reduction in file sizes when compressing imagery using the new ER Mapper wavelet based compression technology, which can compress very large images, at high compression rates, with high quality results.

Usage of Compression
Compression is very useful for the following requirements.

  • Easy distribution of large amounts (10MB to multi TB) of imagery
  • Fast access directly from CD ROM
  • Airphotos, orthophtos, satellite imagery or scanned topographic maps.

One can use a lower compression factor, when some of the following criteria are important to the application.

  • Hyperspectral signature analysis is being carried out
  • Important information is represented as suitable secondary changes to the date
  • Your application is “mission critical,” where you want to retain the original quality of Image, for example navigation system using DEM data.
Table 1: The reduction of file size using ECW
Image Dimensions Uncompressed Size Compression Ratio Compressed Size
Color airphoto 3×8,000×8,000 192 MB 50:1 4MB
Color airphoto mosaic 3×50,000×100,000 15,000 MB 25:1 600MB

Interactive image decompression
An important consideration when dealing with large images is the ability to:

  • Interactively decompress a subset region of a compressed image.
  • Interactively decompress to a Level of Detail (LOD) of a compressed image. For example, the compressed image might be 200,000 x 300,000 pixels in size, yet your viewing window might cover 800 x 600 pixels on screen. In this case, it is important to decompress only the top levels of detail, down to 800 x 600 pixels in resolution, as that is all the information that can be displayed at the current level of detail on screen.

RAM requirements during interactive viewing/ decompression
The ER Mapper Compressed Wavelet (ECW) format has a very low memory footprint during imagery decompression. About 2MB of RAM will be used during interactive decompression of imagery, thus compressed imagery can be used even on quite small computers with limited amounts of RAM – the ER Viewer will view compressed imagery on a machine with a total of 16MB of RAM.

Wavelet based image compression and decompression
Wavelet compression involves a way of analysing an uncompressed image in a recursive fashion, resulting in a series of higher resolution images, each “adding to” the information content in lower resolution images.

The primary step in wavelet compression are performing a Discrete Wavelet Transformation (DWT), quantisation of the wavelet-sace image subbands, and then encoding these sub bands.

Wavelet images by and of themselves are not compressed images, rather it is the quantisation and encoding stages that do the image compression. Image decompression, or reconstruction, is achieved by carrying out the above steps in reverse and inverse order. Thus, to restore the original image, the compressed image is decoded, dequantised, and then an inverse-DWT is performed. Because wavelet compression inherently results in a set of multi-resolution images, it is well suited to work with large imagery which needs to be selectively viewed at different resolutions, as only the levels containing the required level of detail need to be decompressed. Wavelet mathematics embraces an entire range of methods each offering different properties and advantages. Wavelet compression has not been widely used because the DWT operation takes a lot of compute power, and because historical techniques perform the DWT operation in memory or by storing intermediate results on hard disk.

The ER Mapper 6.1 ECW V2.0 wavelet compression uses a breakthrough new technique for performing the DWT and inverse-DWT operations (patent pending), which makes the use of wavelet based compression a practical reality.

Compressing Mosaics of input images to a single compressed image file
The ECW Decompression and Compression SDK’s are to the only libraries on the market that are provided on an open basis and enable you to both compress and decompress and is tightly linked with the Image processing software through the ER Mapper 6.1 Compression Wizard. This means that you can use all the other ER Mapper tools, such as the Ortho Wizard, the Mosaic Wizard and the Balance Wizard, to create a seamless mosaic which can then be compressed out to a single compressed ECW file.

CD ROM/DVD ROM versus Internet distribution
Internet distribution is likely to increase as Internet access rates increase. Currently, imagery distributed over the Internet is best suited for imagery under 5MB compressed in size, which means that the original image size would 100MB in size assuming colour imagery compressed at 25:1. CD ROM and DVD ROM are ideally suited for distribution of large amounts of imagery, as 15GB of imagery can be stored on a single DVD-ROM when compressed at 25:1.

The Image Web Server (IWS)
The Image Web Server has been designed by Earth Resource Mapping for the following:

  • Serve large GB or TB (1,000 GB) sized) images via the Internet or Intranet
  • Enable users to access the imagery via their existing applications including web browsers, MS Office, GIS and CAD Software Products.
  • Enable users to access the imagery via slow modem links.
  • E-commerce revenue by charging access subscription fees or charging for image sub-set downloads
  • Tightly integrate imagery with your existing web based HTML and database applications.

The Image Web Server can serve many different types of imagery including airphotos, high-resolution satellite images, digital terrain models, Radar etc.

The need for the Image Web Server
Currently the large size of image data and network bandwidth limitations preclude making imagery available over a network, in particular the Internet. Typically, sites with imagery would post thumbnail JPEG compressed images for viewing on-line. Distribution of larger images is done via CD. Apart from the time taken and inconvenience, other problems are

  • If the original image is updated, it has to be re-distributed via CD. Users can never be certain that they have the latest version. This is not very satisfactory for industries that rely on up to date information.
  • Users can generally not request subsets of the image. At best they can ask for set portions of the image that contain the part they require. This means that they often pay for data that they do not need.

The Image Server uses ER Mapper’s ECW V2.0 wavelet compression technology to minimize the size of images so that they can easily be propagated over a network with no discernable loss in quality. On the client side, the user can access the data via Internet Explorer and Netscape web browsers, or from within applications like ER Mapper, MapInfo, and Microsoft Office Tools.

Solutions using Image Web Server Technology
The Image Web Server solves the problems of serving large image on the Internet or Intranet by combining the powerful ECW compression with image sub setting. It not only compresses the images, but also only transfers the portion of the image currently being displayed by the client.