Correspondence problem (image processing)

from Wikipedia, the free encyclopedia

The correspondence problem relates as the basis for calculation of the optical flow ( optical flow ) and the stereo vision ( stereo matching ) is a fundamental problem in image processing .

In doing so, those pixels are searched in two digital images which each represent the projection of the same element of the same scene. The result is usually a disparity map , in which a displacement vector to the corresponding pixel of the other image is determined for each pixel of one image.

Algorithms

The correspondence problem is usually solved using local filters or global algorithms.

Local filters

With local filters, only the local area around a pixel is usually considered. The Rank Filter and the Census Transformation were proposed by Zabih and Woodfill .

Global algorithms

Applications

Optical flow

The correspondence problem also arises when calculating the optical flow , which represents an estimate of the movements of objects in an image (" approximation to image motion defined as the projection of velocities of 3D surface points onto the image plane of a visual sensor ").

The continuous determination of the optical flow in optical images can be used to track moving objects and to automatically track the corresponding distance setting.

Stereo correspondence

Using the stereo correspondence ( stereo matching ), depth images in the viewing direction can be determined with two cameras, which is a basis for fully automatic driving or can be used for the computational reduction of the depth of field. The stereo correspondence can also be determined by means of a neural network .

Individual evidence

  1. ↑ Ramin Zabih, John Woodfill: Non-parametric local transforms for computing visual correspondence . In: Computer Vision - ECCV '94 (=  Lecture Notes in Computer Science ). Springer, Berlin, Heidelberg, 1994, ISBN 978-3-540-57957-1 , pp. 151–158 , doi : 10.1007 / bfb0028345 ( springer.com [accessed October 2, 2017]).
  2. ↑ BEAUCHEMIN, Steven S.; BARRON, John L.. The computation of optical flow. ACM computing surveys (CSUR), 1995, Vol. 27, No. 3, pp. 433-466. doi : 10.1145 / 212094.212141
  3. ↑ Advanced Depth From Defocus Autofocus - Lumix DC-GH5 - Technical Director Interview , Panasonic Newsroom July 12, 2017, accessed October 5, 2017
  4. ↑ SCHARSTEIN, Daniel; SZELISKI, Richard. A taxonomy and evaluation of dense two-frame stereo correspondence algorithms. International journal of computer vision, 2002, Volume 47, No. 1–3, pp. 7–42. doi : 10.1023 / A: 1014573219977
  5. ↑ KENDALL, Alex, et al. End-to-end learning of geometry and context for deep stereo regression. CoRR, vol. abs / 1703.04309, 2017. doi : 10.1109 / ICCV.2017.17