---
title: "A Computational Framework for the Visual Correspondence Problem"
authors:
  - "Michael Kass"
doi: ""
url: "https://ijcai.org/proceedings/1983/241"
pdf: "https://ijcai.org/proceedings/1983/0241.pdf"
venue: "Proceedings of the Eighth International Joint Conference on Artificial Intelligence"
year: 1983
pages: "1043-1044"
track: "Vision"
keywords:
---

# A Computational Framework for the Visual Correspondence Problem

Michael Kass

*Proceedings of the Eighth International Joint Conference on Artificial Intelligence*, pages 1043-1044.

## Abstract

;I <'0111pu totional(rame 11 Vrk fo r solving th e visuol correspondence problem is JW' . \e/1/ed a11d l'valua / cd by u si11g a stochas l ic image model. The fram e1 vork dij.f'rrs principal~)' .fTom pr evio us work in !hat it em· phasizcs I he cumhi11a1ion of a lar ge cullcclion of independe11t measure · llll'nts. l' an ial derivwivcs of i ma gfs smootlli'd 1rilh a fl'lv different-sized Gaussian filt ers are sugges t ed as sui/able m eas uremc 11ts. A specific conJputalioll is shr>w n lo reliably es /ab lis h whflher or not two pointscurrespuml, pr o1 • idcd thai th e signal lo correspo11dencc noise ratio in the imag es tu be mmchcd exceeds two. The computation has been applied to anijicial and //a/Ural ima ges with encouraging r es ults.