Exploring edge complexity in remote-sensing vegetation index imageries

Spatial heterogeneity is a fundamental characteristic of all landscapes. There is a large collection of methods on heterogeneity measurement accumulated in the past several decades, but most of them largely depend on categorical data as a primary input. However, the production of spatially and/or temporally extensive land-cover maps can be extremely time-consuming and sometimes prohibitively expensive. Edge number is closely related to spatial heterogeneity and could be identified by edge detection image processing techniques. The widely used edge detectors were tested to evaluate their performances on continuous remote-sensing vegetation index imageries. The results indicate edge features estimated by a Canny-Deriche filter that outperforms other detectors, as validated by a reference land-cover data set. This study uncovers the potential of Canny-Deriche edge detection to be a cost-efficient and time-saving method in remote-sensing applications, that is, to identify and measure edges in a landscape and so to support heterogeneity evaluation and assessment.