Be it noted that u;.3 and u;._3, which are useful for filtering, should be applied at rank 2 or lower, to avoid the construction of an expensive intermediate table.
Henry Rich On 5/30/2012 5:40 PM, Devon McCormick wrote: > Hi - > > this is much clearer but I'm still not quite sure about the meaning - > or maybe it's the notation - of the shift. In your example, it looks > like the "_3" part of the shift is what pads the 1s on the left with > zeros but I don't see how the overlap w/average of the bottom row of > 1s with the first row of 3s is specified by the "2" in the shift. I'm > also unsure why this shift pads a column on the result. > > I guess it would help to see what the "0 0 shift" case looks like - > where the anchor point is before the arrays get combined. > > I suspect part of my difficulty is that "shift" has a certain meaning > already in J and it looks like you mean something subtly different > here, maybe even two different things, something like "shift-extend" > (on the columns) and "shift-overlap-average-where-not-extended" (on > the rows). > > I'm also assuming the shift is applied only to the 1s in your example, > relative to the 3s. > > Regards, > > Devon > > On Wed, May 30, 2012 at 4:52 PM, Alexander Mikhailov<avm...@yahoo.com> wrote: >> >> >> >> The specific problem which I'm working on right now is stitching. Also known >> as "making panoramas from several independent shots". I'm having a >> simplified version of the problem - I control the environment where pictures >> are taken. In my case the pictures are taken with a camera which images >> samples under an optical system, and the camera shifts linearly between >> shots - but of course I wouldn't mind a more general solution. >> >> I don't think it's very unusual task in the image processing area... and I >> don't think, for example, that this is the only task which enlarges images - >> making from two nice rectangles, for example, a convex octagon. >> >> Unfortunately, averaging is just a simple option... I'd like to be able to >> "fade out" one image and "fade in" another along, say, a gaussian curve. But >> let's not have this as a requirement know. >> >> My problems seems to me to be geometrical - making a combined array. It's >> not about "merging" individual points per se. For example, let's have 3 4$1 >> and then 5 6$3 and then "combine" them, overlaying the second array with >> shift 2 _3 on top of the first with averaging... >> >> 1 1 1 1 3 3 3 3 3 3 0 0 0 1 1 1 1 >> 1 1 1 1 "operation" 3 3 3 3 3 3 -> 0 0 0 1 1 1 1 >> 1 1 1 1 3 3 3 3 3 3 3 3 3 2 2 2 1 >> 3 3 3 3 3 3 3 3 3 3 3 3 0 >> 3 3 3 3 3 3 3 3 3 3 3 3 0 >> 3 3 3 3 3 3 0 >> 3 3 3 3 3 3 0 >> ---------------------------------------------------------------------- >> For information about J forums see http://www.jsoftware.com/forums.htm > > > ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
