At the moment, in the GeoSpatial context, a wide range of raster data format exists. However, GeoTools may successfully handle only a small subset of them (GeoTiff, ArcGrid, WorldImage and Gtopo30). Developing a native coverage reader is no small task, so it makes sense to try and leverage other libraries that already have an extensive set of supported formats.
GDAL is a well known open source library providing access access tens of different coverage formats with a common API.
Programming language wise, GDAL is a C++ api, but it provides bindings for integration with various other languages (python, VB6, java,...) based on the SWIG code generator.
License wise, GDAL is based on the X/MIT license.
The Image I/O-Ext project
The Image I/O-Ext project leverages existing native APIs to extend the set of formats that the JAI ImageIO API can read. At the moment, the project is building on top of GDAL java bindings and ImageMagick java bindings, and hopefully the list will grow in the future. The project is compatible with GeoTools both from the license point of view (LGPL) and the API point of view, since most of the current coverage readers are leveraging ImageIO already.
At the moment Image I/O Ext does not support all the formats GDAL provides, but adding a new format is usually a matter of few hours of work (for a complete list see of the currently supported formats see https://imageio-ext.dev.java.net/index.html#Supported_Formats).
Of course, in order to leverage the native readers provided by GDAL, the GDAL libraries and java bindings have to be built for the target platform.
Image I/O Ext provides two options in this regard:
- Step-by-step instructions provided in the [ImageI/O-Ext setup guide, available either as an OpenOffice document or as a PDF document (if you are not registered to java.net, specify "guest" as userID without PASSWORD when asked for login to download the documentation). This option suits people that have the build tools ready, are proficient with them, and need custom builds for some reasons (for example, license ones, the Kakadu JPEG 2000 readers are not open source so they cannot be redistributed along with Image I/O Ext).
- As nicer alternative for those that don't want to mess with native compilers the Image I/O-Ext project provides a self-deployment module which allows to deploy prebuilt native libs in the proper location. Detailed instructions about this are contained in the previous document (Basically, all you need is specifying the -Ddeploylibs property when building ImageI/O-Ext with maven). At the moment binaries are available for Windows and Linux, with plans to build the same for MacOSX as well.
The purpose of the proposed module is to extend the raster data access capabilities of the GeoTools library and, consequently, the GeoServer capabilities too, by building on top of Image I/O-Ext, and adding the necessary GIS features to the otherwise plain ImageIO API offered by Image I/O-Ext.
Given that most of the hard work is already done by the lower levels, the GeoTools readers only have to deal with format specific metadata that aren't already handled in a generic way by GDAL itself (informations such as the CRS, the bounds and the bands structure are already provided thru a generic API).
The only residual difficulty is the distribution one, which can be handled by providing users with install packages based on the pre-built binaries of the GDAL libraries, much like JAI ImageIO is alrady doing for its native counterparts.
The proposed module is composed of a main abstract BaseGDALGridCoverage2DReader (extending AbstractGridCoverage2DReader) which leverages on the core of the ImageI/O-EXT project: the gdalframework module. Since, the hard work of data access and management is performed by the GdalFramework, the Geotools reader simply needs to query it to obtain all the needed information (such as CRS, Envelope, raster properties) and delegate the read to the ImageI/O-Ext.
The GDAL Data model allows to represents any datasource (of any format) with a basic Dataset object.
ImageI/O-Ext wraps all the properties of this object within a GDALCommonIIOImageMetadata object extending the ImageIO IIOMetadata class, which can be leverage at two levels:
- The default implementation of the BaseGDALGridCoverage2DReader will query an instance of this object obtained by the framework to setup Envelope, CRS, GeoTransformation and GridRange.
- Some formats allows to specify/handle an additional set of metadata. In such a context, an extended implementation of the geotools GDAL reader will leverages on a proper GDAL metadata subclass.
The GDAL based plugin allows to customize the data access by means of a set of read parameters.
First of all, it leverages on the AbstractGridFormat USE_JAI_IMAGEREAD parameter/Hint which allows to specify whether the underlying ImageIO plugin should use a JAI ImageRead operation (and thus deferred execution model, tile caching/scheduling, ...) or a more simple direct call to the read method of the ImageReader. Note that when performing a JAI ImageRead operation, the internal computations get access to any involved tile using an ImageReader read operatin. In case some raster data are striped (where tiles are in the form Nx1 pixels) the tiles read process may be time consuming. To improve this, the parameter SUGGESTED_TILE_SIZE allows to specify a different layout to be used by JAI ImageRead operation when consuming tiles. Values to be specified are in the form "W,H" where W represent the suggested tile width and H represents the suggested tile height.
Moreover, ImageIO-Ext introduces a very useful JAI operation: ImageReadMT. This allows to perform multithreaded JAI ImageRead operations.
To tell the geotools reader to use multithreading, you simply need to specify the USE_MULTITHREADING parameter (Set as TRUE).
These parameters are declared within a BaseGDALGridFormat extending the AbstractGridFormat.
Actually supported formats on the gt2-imageio-ext-gdal module are:
- Erdas Imagine
- JPEG2000 (based either Kakadu or MrSid libraries)
Extending the raster data access capabilities with add for additional formats is a simple task due to the GDAL power capabilities integration in the ImageI/O-Ext.
Usually, writing a new ImageI/O-Ext plugin for a "not too much complex" format requires 10-15 minutes of work (you only need to setup a XXXImageReader as well as a XXXImageReaderSpi). The same (or minor time) is requested to setup the related GeoTools plugin in the gt2-imageio-ext-gdal module (You only need to setup a XXXFormat, a XXXFormatFactory and a XXXReader), where XXX is usually the name of the new format to be supported. With further developments it would be nice to extend GeoTools raster data access capabilities to support any format supported by GDAL (See: http://www.gdal.org/formats_list.html).
Not all the GDAL supported formats are freely available and freely redistributable. Notable exceptions are:
- ECW: the SDK license agreement prohibits the SDK use on a server (So, for the moment, we can't distribute ECW support with GeoServer).
- JPEG200 Kakadu reader: this is a high performance, commercial (and quite expensive) library that needs to be bought in order to be used at all (unlikely the ECW license, you have to buy a license even for desktop usage).
Some notes on old unsupported modules
Actually, in the unsupported section, there are an ECW and a MrSID modules. Since some GeoTools/Udig/GeoServer people use these modules, they wouldn't be removed until the GDAL based Geotools plugin will ready and fully tested.
Some screenshots produced running GeoServer leveraging on the proposed module:
In the left side of the image, the rendering of a NITF-CADRG data sample;
In the right side, a Landsat7 MrSID data sample
In the left side of the image (Sardegna), the rendering of a DTED data sample;
In the center, a Landsat7 MrSID data sample
Some more ImageIO Ext information
Here are a couple of class diagrams showing how ImageIOExt GDAL framework is set up
... and how the MrSid plugin is using it.