stuff

TODO

@@ -1,28 +1,15 @@
-- im_lrmosaic
-
-	#       halfcorrelation is of type "integer"
-	#       halfarea is of type "integer"
-
-  try
-
-	vips im_lrmosaic t1.v t2.v t3.v 0 372 0 0 90 0 0 0 1
-
-  set them to zero (which you'd think would search for 1 pixel window) gives a
-  divide by zero error
-
-  maybe better to clip windows by image overlap? then we could specify
-  non-zero halfcorrelation etc. and get them to search the 1 pixel overlap we
-  have
-
-  we probably assume square windows in several places, argh
-
-  
 
 
 - memory.c
 
 - more stuff from util.c? too much to do it all now
 
+- how about an indexed histogram?
+
+  have two images, a 1-band index image and an n-band image to analyze
+
+  for each pixel, add the n-band pixel to the bin indicated by the index image
+