loc@yrloc.ipsa.reuter.COM (Leigh Clayton) (04/12/91)
----------Forwarded from Eugene McDonnell via IPSA mailbox----------------- no. 5092717 filed 17.20.18 wed 10 apr 1991 from eem to clapl Reply to William Ricker The histogram algorithm discussed recently does not derive from the work reported on at APL80, but rather results from recent developments in the J dialect. I'd be surprised if you were able to find a copy of the APL80 proceedings, since the organizers refused to be persuaded by my arguments and published them using the commercial house North-Holland. I wanted them to use ACM, as earlier and later meetings did. One of the organizers worked for North-Holland. The proceedings cost about $100, I believe. In any event, the histogram algorithm H referred to in the APL90 paper you cite is as follows: H:+/((_1+min/min/M)drop iota max/max/M)jot.=,M,O take quadio is 1 where: Let M(x,y) be a digitized image, 1 le y le n, 1 le x le m, O le M(x,y) le k, where k is the maximum grey level. The pixels are classified into classes of 'similar' values, which have the same physical behaviour in the scene, obtained by the interactive exploration of the numerical matrix. ----------------------------------------------------------- loc@tmsoft.UUCP uunet!mnetor!tmsoft!loc loc@yrloc.ipsa.reuter.COM (Leigh Clayton)