Chromasens, a global leader in line scan machine vision technology, has published a new whitepaper that provides a critical overview of the block-matching approach in height reconstruction for passive 3D stereoscopy.
The paper, titled “Limitations of block-matching algorithms on depth reconstruction” is available for free download from www.chromasens.com.
Block-matching has proven useful in 3D imaging height measurement when surfaces feature both a distinctive texture and a smooth height change. However, the algorithm intrinsically assumes that the object height relative to the camera is constant within the template window size.
The similarity measure can only find the best match if the image content actually appears similar in left and right perspective views of the stereo image scene. In practice, there are often areas in images where these strong requirements are not met, resulting in certain effects on the reconstructed height image. The aim of the whitepaper is to describe the cause of different height estimation artifacts that are visible at distinctive image features.