A METHOD FOR ESTIMATING THREE-DIMENSIONAL DEPTH VALUE FROM UN-CALIBRATED IMAGE SEQUENCES

The present invention relates to a method (100) for estimating three-dimensional depth value from un-calibrated image sequences, characterised in that the method comprises, a step (101) of un-calibrated image acquisition including steps of: placing an object on a platform with a distance from an image capture apparatus; acquiring a minimum of two images of the object, whereby a first image is obtained by capturing a first view of the object whereas a second image is obtained by capturing a second view of the object after moving the image capture apparatus; a step (102) of extracting two-dimensional feature point coordinates from the first image and the second image using a feature detector; a step (103) of applying an optical flow technique to obtain displacement magnitude, translation, and rotation of the image capture apparatus and to find matching two-dimensional feature point coordinates from the first image and the second image; a step (104) of auto-calibrating the first image and the second image based on an optical flow feature match; a step (105) of converting the two-dimensional feature point coordinates of the first image and the second image into world coordinates using the scaling factor; a step (106) of normalizing vectors from the world coordinates of the first image and the second image; and a step (107) of estimating the three-dimensional depth value of the acquired images using the focal length value, object distance value, and rotation value between the two vectors by utilizing equation (1). (Figure 1)