Abstract
Image analysis of tree crown images may provide more objective measures of crown condition which has been assessed visually in the forest health monitoring of many countries, but the commonly used interactive thresholding approach is too time-consuming to apply to large numbers of images. This study examined the usefulness of three automatic thresholding algorithms that are based on the between-class variance (VARIANCE), the classification error and the total entropy, respectively. VARIANCE had no significant difference from, and good correlation (r=0.992) with, interactive thresholding for the black pixel percentage within a region of interest, but the other two did not. VARIANCE also had high correlation (r=0.992) of the black pixel percentage between cloudy and blue sky conditions. It was concluded that VARIANCE can be successfully used for thresholding for tree crown images and can greatly reduce time for image processing, while maintaining a high level of accuracy and reproducibility. However, we need to pay attention to the quality of the original color images.
