Statistics – Computation
Scientific paper
Sep 2011
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2011amos.confe..72p&link_type=abstract
Proceedings of the Advanced Maui Optical and Space Surveillance Technologies Conference, held in Wailea, Maui, Hawaii, September
Statistics
Computation
Scientific paper
Despite improvements in automatic processing, analysts are still faced with the task of evaluating gigapixel-scale mosaics or images acquired by telescopes such as Pan-STARRS. Displaying such images in ‘ideal’ form is a major challenge even today, and the amount of data will only increase as sensor resolutions improve.
In our opinion, the ideal viewer has several key characteristics. Lossless display - down to individual pixels - ensures all information can be extracted from the image. Support for all relevant pixel formats (integer or floating point) allows displaying data from different sensors. Smooth zooming and panning in the high-resolution data enables rapid screening and navigation in the image. High responsiveness to input commands avoids frustrating delays. Instantaneous image enhancement, e.g. contrast adjustment and image channel selection, helps with analysis tasks. Modest system requirements allow viewing on regular workstation computers or even laptops.
To the best of our knowledge, no such software product is currently available. Meeting these goals requires addressing certain realities of current computer architectures. GPU hardware accelerates rendering and allows smooth zooming without high CPU load. Programmable GPU shaders enable instant channel selection and contrast adjustment without any perceptible slowdown or changes to the input data. Relatively low disk transfer speeds suggest the use of compression to decrease the amount of data to transfer. Asynchronous I/O allows decompressing while waiting for previous I/O operations to complete. The slow seek times of magnetic disks motivate optimizing the order of the data on disk. Vectorization and parallelization allow significant increases in computational capacity. Limited memory requires streaming and caching of image regions.
We develop a viewer that takes the above issues into account. Its awareness of the computer architecture enables previously unattainable features such as smooth zooming and image enhancement within high-resolution data. We describe our implementation, disclosing its novel file format and lossless image codec whose decompression is faster than copying the raw data in memory. Both provide crucial performance boosts compared to conventional approaches. Usability tests demonstrate the suitability of our viewer for rapid analysis of large SAR datasets, multispectral satellite imagery and mosaics.
No associations
LandOfFree
Engineering the Ideal Gigapixel Image Viewer does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with Engineering the Ideal Gigapixel Image Viewer, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Engineering the Ideal Gigapixel Image Viewer will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1079287