Subject: Review of a paper From: "Richard Crutcher" Date: Thu, 21 Dec 2000 13:33:54 -0600 To: "Geoffrey Fox" I have reviewed the paper C487: Real-time Multi-spectral Image Fusion. With the caveat that the material only marginally overlaps my area of expertise, I recommend that it be published essentially unchanged. It reports a technique for and experience with real-time detection of image signals using spectral imaging information about the signals. The paper is generally clearly written, although it was not carefully proof-read; there are a number of typographical and grammatical errors which the authors or an editor should look for and correct. I do have a few minor suggestions. The authors misuse several technical terms. (1) In the abstract they say "...with spectral resolution between 400 and 1000 nm." They mean spectral coverage, not spectral resolution, which could refer to the bandwidth of the filters employed but not to the total wavelength range sampled. (2) Starting in the introduction and continuing in several places, they use the word "spectra" when they really mean "images obtained at different wavelengths". The set of these images is not a spectrum. A spectrum would be what is produced by the dispersion of light by a prism, grating, or other optical device, with all wavelengths represented continuously. They do not obtain spectra, they obtain a set of medium spectral-resolution (probably, although they do not state the bandwidth of their filters) images at different wavelengths. Every use of the word "spectral" or "spectra" in the paper should be examined with this in mind and appropriately corrected. (3) Just before section 2 they use the word "lenses" when they mean "filters". The discussion of Spectral Angle Classification should be improved. They use the phrase "associated pixel vectors X and Y" with no definition of what X and Y are. Is X the intensity in an image obtained at one wavelength and Y the intensity at a second wavelength? Also, they do not mention having to deal with the fact that the transmission of the individual filters must vary from filter to filter and the sensitivity of the CCD detectors must vary as a function of wavelength and pixel in the detector array. Do they calibrate this or otherwise allow for it? Is this a significant source of error/uncertainty? ------------------- Richard M. Crutcher Professor & Chair, Astronomy Department Chief Application Scientist, NCSA University of Illinois 1002 W. Green Street Urbana, IL 61801 Voice: 217/333-9581 Fax: 217/244-7638