Stéphane Dalbera’s Post

Transcript

Pixar is a leader in the creation of tools that allow people to understand data through the technology of visualization. Our hardware and software products for image computing provide the means to interpret complex multi dimensional data. The availability of public domain satellite images is representative of the growth in data gathering technology. The data generated by remote sensing, medical imaging and supercomputer systems is orders of magnitude greater than the resources available for people to process and interpret it. Pixar provides the visualization tools to solve this problem. These tools combine aspects of image processing, where existing images are analyzed, and computer graphics, where images are synthesized from models and data. Pixar creates general purpose solutions for the processing and synthesis of visual information. This is image computing. The Pixar family of image computers consists of the original Pixar Image Computer. The new lower priced Pixar 2 and includes AC compiler and software toolkits for developing specific applications. Pixar's image computers are being used in medical imaging, remote sensing. Graphic arts, including 3D design and animation and scientific visualization. An emerging field that requires the display of data from computational and experimental scientific disciplines. We will now demonstrate some examples of image computing drawn from applications in these markets. The first example demonstrates the power of Pixar's image computers applied to image processing. Using this Noah satellite image of the Midwestern US, we will enhance the data using some standard image processing techniques. First, we'll improve the dynamic range of the image using its histogram. Next we'll sharpen the image in 1/3 of a second using a three by three convolution kernel. And finally, we'll correct for the spherical curvature of the image using a bicubic warp. In addition to these standard software routines, the system can be programmed to perform higher level operations such as feature extraction and pattern recognition. Our next example is surface rendering. For product design and 3D broadcast animation, fast, high quality surface rendering is a required. Chap Raise features a full set of image synthesis capabilities including anti aliasing, advanced reflectance models and multiple light sources with the added advantage of high performance. This image of a jet was rendered in 18 seconds from a model of over 11,000 polygons. The same image would require more than 15 minutes of computation. Unconventional mini computers. By rendering a sequence of frames, we can animate this 3D model. Chap. Raise renders 50,000 to 100,000 complex polygons per minute. Its speed and realism allows designers to make informed decisions about the look and styling of their work. This sequence from Pixar's film Reds Dream demonstrates surface rendering applied to film animation. Each frame is composed of roughly 11,000 surface patches, which were subdivided by chap rays into a half million polygons and then rendered with texture maps, multiple spotlights, and shadows. Each frame was computed in eight to 10 minutes on the Pixar Image computer, roughly 50 times faster than many computer rendering speeds. The next example demonstrates the visualization of volumetric data, where essential information is internal to an object, not just on its surface. The CHAMP Volume Software Toolkit is the result of Pixar's pioneering work in volume visualization. Here we see the results of the stress analysis of the design for a machine part. It has been rendered as a volume of information, with areas of highest stress shown in red. And lesser stresses in green and blue. We can also view this three-dimensional volume interactively. The part is now being displayed as the contents of a cube of data. These views of the object are being constructed from over 16,000,000 data points, all loaded into the image computers memo. The ability to view the internal structure of complex 3D objects is essential to scientific applications such as astrophysical and geophysical analysis. Let's look at a few more examples of volume rendering. Here we see the kinetic energy around a cylinder flying at roughly twice the speed of sound. These images were generated directly from data samples of the energy distribution around the cylinder. Chap Volumes preserves the visual characteristics of three-dimensional information without reducing the data to geometric models. This is an illustration of fluid flow analysis. A puff of smoke was scanned by successive sweeps of a laser. The resulting data was used to generate these views of the smoke cloud. By calculating and studying multiple views of turbulence, scientists gain a better understanding of the three dimensionality of such complex structures. Volume imaging is a valuable medical research tool. Here we see a three-dimensional reconstruction from some 70 CT scans of a patient with fibrous dysplasia. When the same data is rendered with the skin transparent, fine details of the bone deformity are revealed that would not have been evident in the original CT scans. This use of volume imaging is now available in an end user system for radiologists, The Phillips picks 2000 this image. The rocket nozzle is the reconstruction of a series of images from an industrial CT scan. The Pixar Image computer is being used to locate flaws in the materials that make up this complex object. For applications in non destructive testing, 3D volume rendering is a valuable tool for quality assurance. Our next example is terrain rendering, the reconstruction of three-dimensional perspective scenes from digital terrain elevation data. Starting with this image, Chapter Reigns is combining it with digital elevation data to produce a realistic animated sequence simulating a flyby over these mountains. The speed of Chapter Ains enables users of geographic information systems and mission planning and training to quickly prepare high quality views of terrain elevation data. Both the Pixar image computer and the Pixar 2 feature high resolution memories storing from 12 to 192 megabytes of full color pixels at 48 bits per pixel, the largest image memory found on any computer. This storage capacity is matched by the systems computational horsepower. The general purpose programmable CHAP processor delivers high speed computational power efficiently and affordable. Pixar image computers are used as special purpose graphic systems connected to a wide range of hosts from personal to supercomputers. This advanced hardware is further supported by more than 200,000 lines of software in the form of libraries, AC compiler and other development tools, application toolkits and 3rd party software designed to serve your image computing needs. The Pixar family of image computers offers solutions to a range of applications in image processing and computer graphics. With the new Pixar 2 there is an entry level system priced under $30,000. They are supported by the people at Pixar through documentation, training, field support, third party software and most importantly, a commitment to high quality visualization. Through these efforts, Pixar has become the leader. And the new and growing field of image computing. I'm Ed Catmull, president and cofounder of Pixar. In the last seven years, we've assembled a team that has developed many of the innovations in computer generated imagery. As Pixar, we've established ourselves as leaders in the technology for image computing. We are very excited by the many uses of this technology and by our partnership for the companies that also make it available. Our goal is to provide high performance visualization at low cost. We look forward to discussing how these products can be applied to your need for the interpretation of visual data.