cvfind memory usage for large projects is enormous

[Bob-O-Rama]

cvfind pre-loads all images into RAM as UINT8 x 3 RGB images, then also creates a UINT8 monochrome image. These consume at least 4 bytes / pixel of image data. cvfind also produces significant feature, match, and control point data for each image pair. cvfind generally is fairly efficient in its use of swap space, so the solution of instructing the user to "add more swap space" will work in most cases. The retention of the RGB images may be unnecessary. opotions like --savediag and --looprepair require the retention of images. The effort to discard and reload images is fairly low compared to the image processing. Feature and matching data is not used once transcribed into pointPad structures.