Corresponding point searching method and three-dimensional position measuring method
A plurality of images (I, J) of an object (M) when viewed from different viewpoints are taken in. One of the images is set as a standard image (I), and the other image is set as a reference image (J). One-dimensional pixel data strings with a predetermined width (W) are cut out from the standard image (I) and the reference image (J) along epipolar lines (EP 1 , EP 2 ) calculated from a camera parameter (CPR) and the reference point (p). Calculating a phase-only correlation function from the cut one-dimensional pixel data strings will obtain a correlation peak position (Pa 1 ). A positional shift amount (d) from the correlation peak position (Pa 1 ) is obtained. A search is made for a corresponding point (q) corresponding to the reference point (p) based on this position shift amount (d).
1. A corresponding point searching method comprising the steps of:
taking in a plurality of images of an object when viewed from different viewpoints via image take-in means;
cutting out one-dimensional pixel data strings from the plurality of images along epipolar lines calculated from a parameter for the image take-in means and a reference point; and
searching for corresponding points in the plurality of images from a correlation between the one-dimensional pixel data strings cut out from the plurality of images;
wherein the step of cutting out comprises the steps of cutting out one-dimensional pixel data strings from the plurality of images along the epipolar lines, and cutting out one-dimensional pixel data strings near the one-dimensional pixel data strings along lines parallel to the epipolar lines,
and the step of searching comprises the step of searching for corresponding points in the plurality of images from a correlation obtained by synthesizing correlations taken between each of a plurality of corresponding pairs of one-dimensional pixel data strings cut out from the plurality of images.
2. A corresponding point searching method according to claim 1 , further comprising the step of extracting object regions as corresponding point search target regions from the plurality of images,
wherein the step of cutting out comprises the step of cutting out one-dimensional pixel data strings from the plurality of images along the epipolar lines, from the object regions extracted from the plurality of images, and cutting out one-dimensional pixel data strings near the one-dimensional pixel data strings along lines parallel to the epipolar lines.
3. A corresponding point searching method according to claim 1 , further comprising the step of measuring three-dimensional positions of corresponding points based on parallax between the searched corresponding points.
4. A three-dimensional position measuring method comprising the steps of:
taking in a plurality of images of an object when viewed from different viewpoints via image take-in means;
cutting out one-dimensional pixel data strings from the plurality of images along an epipolar line calculated from a parameter for the image take-in means and a reference point, and cutting out one-dimensional pixel data strings near the one-dimensional pixel data strings along a line parallel to the epipolar line;
searching for corresponding points in the plurality of images from a correlation obtained by synthesizing correlations taken between each of a plurality of corresponding pairs of the one-dimensional pixel data strings cut out from the plurality of images; and
measuring three-dimensional positions of corresponding points based on parallax between the searched corresponding points.
5. A corresponding point searching method according to claim 1 , wherein
the step of searching comprises searching for the corresponding points in the plurality of images from a correlation obtained by synthesizing, with a predetermined weight, correlations taken between each of a plurality of corresponding pairs of the one-dimensional pixel data strings cut out from the plurality of images.