User manual

Small Vision System User Manual 39
3.4 Area Correlation Window
Stereo analysis is the process of measuring range to an object based on a comparison of the object
projection on two or more images. The fundamental problem in stereo analysis is finding corresponding
elements between the images. Once the match is made, the range to the object can be computed using the
image geometry.
Area correlation compares small patches, or
windows, among images using correlation. The window
size is a compromise, since small windows are more likely to be similar in images with different
viewpoints, but larger windows increase the signal-to-noise ratio. Figure 3-7 shows a sequence of disparity
images using window sizes from 7x7 to 13x13. The texture filter was turned off to see the effects on less-
textured areas, but the left/right check was left turned on.
There are several interesting trends that appear in this side-by-side comparison. First, the effect of
better signal-to-noise ratios, especially for less-textured areas, is clearly seen as noise disparities are
eliminated in the larger window sizes. But there is a tradeoff in disparity image spatial resolution. Large
windows tend to “smear” foreground objects, so that the image of a close object appears larger in the
disparity image than in the original input image. The size of the subject’s head grows appreciably at the
end of the sequence. Also, in the 7x7 the nose can be seen protruding slightly; at 13x13, it has been
smeared out to cover most of the face.
One of the hardest problems with any stereo algorithm is to match very small objects in the image. If
an object does not subsume enough pixels to cover an appreciable portion of the area correlation window,
it will be invisible to stereo processing. If you want to match small objects , you have to use imagers with
good enough spatial resolution to put lots of pixels on the object.