Introduction

As SEP begins to address the problems of imaging 3-D prestack data and of estimating 3-D velocity models, it needs tools to analyze large 3-D prestack datasets, either computer generated or recorded in the field. For the visualization of 3-D datasets SEP has been relying on the Movie program developed by Rick Ottolini in 1984 and its more recent X-Window version. Over the years the Movie program has proven to be an invaluable tool for better understanding seismic datasets and for gaining an immediate intuition of the properties of seismic processing operators. However, the Movie program begins to show its age, and it has some important shortcomings for the task of displaying 3-D surveys. First, the program is limited to the display of three-dimensional datasets. Second, it is essentially a batch program in the sense that it displays data already converted to single-byte integers with predetermined clipping and gain factors. Finally, the effectiveness of the program in manipulating large datasets is hampered by the hardware characteristics of workstations (memory size, data handling capabilities, and computation speed). These limitations are important for visualizing 3-D prestack data, which are in the most general case five-dimensional, and may span several gigabytes, even for small ``research-size'' surveys.

To overcome Movie's limitations we developed a multi-dimensional visualization application running on SEP's CM-5 and based on the AVS (Application Visualization System) graphics system. We chose the CM-5 as the developing platform because of its data-handling capabilities. In particular, the CM-5 provides fast I/O for reading the data from files on disks that can span multiple gigabytes and it has up to one gigabyte of memory to store large subsets of the survey. In addition, the multiple data-paths to memory that are typical of a parallel computer allow fast access to data stored in memory according to a wide variety of data-access patterns. Finally, the processing capabilities of the CM-5 allow the user to perform some simple image processing (i.e. clipping and gaining) in an interactive fashion, improving the information content of the display.

We developed the visualizer using AVS and CM/AVS (an adaptation of AVS to the CM-5 developed by Thinking Machines) mainly for two reasons. First, AVS enabled us to quickly develop modular interactive applications. More importantly, AVS allows the visualization of many different types of geophysical data integrated in the same display. For example, our data-slicer can be integrated with a geological model (e.g. GOCAD Mallet (1993)) and velocity model visualizer. The final goal is to make available to SEP researchers a fully integrated tool for understanding 3-D seismic data and its relationship with geological and velocity models.

The visualization application that we have developed has some limitations. Some of them are intrinsic to the design, and thus it would be difficult to remove them without overhauling the whole application. The most important restriction is that the data to be visualized must be in SEP format, i.e. it must be regularly sampled along all the axes. This restriction does not necessarily limit the handling of marine surveys, but it is definitely a problem for land data. As for performance, presently the bottleneck is the in last step of the visualization process. Once images are extracted from the data and processed on the CM-5, they need be sent to an X-server. The efficiency of this communication step should improve with future developments of the CM-5 software and hardware, but it is presently unsatisfactory.