A comparative study of grayscale conversion techniques applied to SIFT descriptors

Authors

  • Samuel Macêdo Universidade Federal de Pernambuco
  • Givânio Melo Universidade Federal de Pernambuco
  • Judith Kelner Universidade Federal de Pernambuco

Keywords:

grayscale conversion, SIFT descriptors, computer vision, RGB channels

Abstract

In computer vision, gradient-based tracking is usually performed from monochromatic inputs. However, a few research studies consider the influence of the chosen color-tograyscale conversion technique. This paper evaluates the impact of these conversion algorithms on tracking and homography calculation results, both being fundamental steps of augmented reality applications. Eighteen color-to-greyscale algorithms were investigated. These observations allowed the authors to conclude that the methods can cause significant discrepancies in the overall performance. As a related finding, experiments also showed that pure color channels (R, G, B) yielded more stability and precision when compared to other approaches.

Downloads

Download data is not yet available.

Downloads

Published

2015-11-17

Issue

Section

Special Issue - SVR 2015