Related Experiment Video
Updated: Apr 20, 2026
11:57
Measuring Spatially- and Directionally-varying Light Scattering from Biological Material
Published on: May 20, 2013
14.1K
Standardizing the protocol for hemispherical photographs: accuracy assessment of binarization algorithms.
Jonas Glatthorn1, Philip Beckschäfer2
1Department of Plant Ecology, Albrecht von Haller Institute of Plant Sciences, Georg-August-Universität Göttingen, Göttingen, Germany.
Plos One
|November 25, 2014
Summary
This study evaluated seven algorithms for processing hemispherical photographs, finding that "Minimum," "Edge Detection," and "Minimum Histogram" are best for images without overexposure, ensuring accurate canopy analysis.
Area of Science:
- Ecology
- Forestry
- Remote Sensing
Background:
- Hemispherical photography is crucial for assessing plant canopy parameters like light and foliage distribution.
- Standardized analysis methods are lacking, hindering cross-study comparisons.
- Automatic thresholding algorithms vary, impacting result reliability.
Purpose of the Study:
- To evaluate the accuracy of seven binarization algorithms for hemispherical photograph analysis.
- To identify optimal threshold selection methods for consistent ecological data.
- To compare algorithm performance under different exposure settings.
Main Methods:
- Compared seven binarization algorithms against manually classified reference pixels.
- Assessed accuracy using Percentage Correct (Pc) and Kappa statistics (K).
- Evaluated performance on both auto-exposed and histogram-exposed (overexposure-avoiding) photographs.
Main Results:
- All tested algorithms were sensitive to overexposure in photographs.
- "Minimum," "Edge Detection," and "Minimum Histogram" algorithms showed high accuracy (Pc > 98%) for histogram-exposed images.
- "Minimum" algorithm had the lowest gap fraction overestimation (11%), while others significantly overestimated it.
Conclusions:
- "Minimum," "Edge Detection," and "Minimum Histogram" are recommended for processing histogram-exposed hemispherical photographs.
- Overexposed images significantly compromise the accuracy of most tested algorithms.
- Careful exposure control and algorithm selection are vital for reliable canopy analysis.
