hunt@spar.SPAR.SLB.COM (Neil Hunt) (02/25/88)
A question for you vision researchers and graphics wizards: what file formats are you using for transferring and storing your image data ? We are using a format called `picpac' which I believe originated from CalTech; we are looking for something more comprehensive to use. Some of the features to be considered are: Storage of image data in 1D (histograms etc), 2D (simple images), 3D (image sequences, sets of images, etc) and higher dimensionalities (sequences of sets of multispectral images...). Typed pixels: integer and floating point; unsigned and signed, packed and unpacked, RGB, CMY, CMYB, IR-vis-UV, 1, 8, 16 and 32 bits, etc. etc. Storage of image data in non array format (list of points of interest in sparse images, compressed formats, etc.) Recording of arbitrary additional data: colour maps (both predefined and locally specified), camera pixel size, aspect ratio, date, time, condition, title, subject, pre- and post-processing, etc. Efficient storage and transfer (ie: the PostScript format is not ideal, with ASCII pixel values). Indirection: the ability to make a new image being some subset of another image, without having to copy the actual data. Please let me know what format you are using at the moment, what you are doing with it, and what other features we should consider if we decide to invent our own format ? Neil/. hunt@spar.slb.com ...{amdahl|decwrl|hplabs}!spar!hunt (415) 496 4708
schwrtze@acf8.UUCP (E. Schwartz group) (03/02/88)
On image file formats: We are using the HIPS system, of Mike Landy and colleagues (NYU Dept. Psych.), which was described in Comp. Graphics and Image Processing (1985?..). The original version of this system consists of a large number of image processing routines ( written in C) and features a well thought out image header, which includes an executable history of the operations which have been performed on the given file. The basic system supports typed images ( BYTE, SHORT, INT, FLOAT, DOUBLE, COMPLEX), and image sequences. In regard to some of the other requirements mentioned, we (Computational Neuroscience, NYU MedCtr) have extened the basic HIPS system to include color maps, non-array image formats (histograms, sparse images), arbitrary additional data ( pixel size aspect ratio, and user defined info that needs to accopmany the image), and interactive window based tools for the SUN environment. My impression is that essentially all of the requirements listed are met by the original HIPS implementation and/or our extensions of it. We have used this system for a number of years in a large computer aided neuro-anatomy project, and various computational vision applications: since the basic system is well constructed, it is easy to extend it to handle new problems, although we have found that the current implementation has stabilized, and there is little that we need add... Eric Schwartz schwrtze@acf8.nyu.edu Computational Neuroscience Labs NYU Med. Center
sims@stsci.EDU (Jim Sims) (03/03/88)
I tried to reply but our mailer is brain-dead... Most astronomers use FITS, which allows multiple data types, multiple groups and types of data in the same file, and extension to the format. Any observatory or astronomy program at a university can get you the full scoop. -- Jim Sims Space Telescope Science Institute Baltimore, MD 21218 UUCP: {arizona,decvax,hao,ihnp4}!noao!stsci!sims SPAN: {SCIVAX,KEPLER}::SIMS ARPA: sims@stsci.edu