Depth from Defocusing

We have developed a new algorithm for "depth from defocusing" (or "depth from blurring") for recovering the depth of objects by measuring the amount of image blur. Compared to earlier algorithms, this method has the following advantages. ffl It gives a unique estimate of depth. ffl It has high accuracy at low computational costs. ffl Easy calibration (only one unknown constant) These advantages are achieved by introducing a "dynamic referencing technique", and taking the focusing position of the lens as a camera parameter rather than the aperture of the lens. The accuracy in the measurement of depth was at worst 8.4 percent and 5.8 percent on the average when the focus length of the lens was 16.7 mm and the distance to the object was varied from 1500 mm to 2500 mm. 2 1 Introduction Depth perception is a key component in low-level computer vision and in understanding the three dimensional relationships between the objects in the world. The availability of absolute depth cues has..