This open source release of the Core RHSEG Software Package is intended to facilitate the investigation of methods for fine grained parallel implementations of the RHSEG software package as well as to facilitate the investigation of approaches to improve the segmentation results produced by RHSEG through algorithmic modifications.
The Core RHSEG Software Package differs from the licensable RHSEG Software Package in three key ways, to avoid licensing requirements and patent issues:
- The Core RHSEG Software Package does not include the source code necessary for parallel implementation,
- The Core RHSEG Software Package does not include the source code necessary for three-dimensional data analysis, and
- The Core RHSEG Software Package does not include the processing window artifact elimination source code.
Otherwise the Core RHSEG Software Package source code is identical to the licensable RHSEG Software Package source code.
Recursive Hierarchical Segmentation (RHSEG) Project
The RHSEG software package has evolved over the years from an early proceedings paper (Image Segmentation by Region Growing and Spectral Clustering with a Natural Convergence Criterion, by James C. Tilton, Proceedings of the 1998 International Geoscience and Remote Sensing Symposium, Seattle, WA, pp. 1766-1768, July 6-10, 1998) to an initial NASA New Technology Report (James C. Tilton, "Method for Recursive Hierarchical Segmentation by Region Growing and Spectral Clustering with a Natural Convergence Criterion", Disclosures of Invention and New Technology (Including Software): NASA Case No. GSC 14,328-1, February 28, 2000) to the most recent release (version 1.47) on December 2, 2009.
Currently available image analysis techniques do not effectively extract the information content from increasingly available high spatial resolution remotely sensed imagery data. High Spatial resolution imagery can resolve individual objects such as man- made structures and even individual large trees. However, several studies have shown that most currently available pixel based analysis techniques do not perform well on this type of data. The HSEG algorithm pre-process image or image- like data into region classes or objects, enabling a region- based analysis of the data.
However, the HSEG algorithm is computationally intensive for large data sets. The computational requirements of the HSEG algorithm can be significantly reduced through a recursive approximation to HSEG, called RHSEG, which recursively subdivides the imagery data into smaller sections. The recursive nature of RHSEG leads to a very efficient coarse grained parallel implementation.
RHSEG Project Goals
The RHSEG project is currently focusing on two main goals. (i) Infusing RHSEG technology into other NASA projects wherever this may be beneficial, and (ii) the application of RHSEG to the general field of object-based image analysis (OBIA). The RHSEG project was selected for NASA GSFC Internal Research and Development (IRAD) funding for FY 2010.
The field of object-based image analysis (OBIA) has arisen in recent years to address the need to move beyond pixel-based analysis. The RHSEG software package was developed specifically to facilitate moving from pixel-based image analysis to OBIA. RHSEG provides an excellent starting point for OBIA because of three key factors: (i) the high spatial fidelity of image segmentations produced by RHSEG, (ii) RHSEG's automatic grouping of spatially connected region objects into region classes, and (iii) RHSEG's automatic production of a hierarchical set of image segmentations.
The key unique aspect of RHSEG is that it tightly intertwines region growing segmentation, which produces spatially connected region objects, with region object classification, which groups sets of region objects together into region classes. No other practical, operational image segmentation approach has this tight integration of region growing object finding with region classification. This integration is made possible by the recursive, divide-and-conquer implementation utilized by RHSEG, in which the input image data is recursively subdivided until the image data sections are small enough to successfully mitigate the combinatorial explosion caused by the need to compute the dissimilarity between each pair of image pixels. RHSEG includes a NASA patent pending approach for blending together the results from neighboring data sections as the recursive subdivision is retraced back up to the full image. While the recursive approach itself makes practical the processing of moderately large data sets, RHSEG has a NASA patented parallel implementation that makes it practical to process very large data sets, such a full Landsat TM scenes or full MODIS swath scenes, in operational settings. For example, the currently available parallel implementation can process a full Landsat TM scene in less than 10 minutes utilizing 64 CPUs on the Discover Linux Cluster at the NASA Center for Computational Sciences (NCCS).
Functional Goals of the Core RHSEG Software Package Open Source Release
This project seeks to investigate additional fine grained parallelism for improve the processing speed of the HSEG or RHSEG algorithm implementation. This project also seeks to improve segmentation results provide by HSEG or RHSEG through algorithmic modifications.