Font Size:  Small  Medium  Large

Using computer vision metrics for spatial pattern comparison

Karim Malik, Colin Robertson

Abstract


Spatial pattern comparison is a research area encompassing algorithms and metrics from disparate disciplines with the aim of detecting and quantifying changes in underlying spatial patterns and processes. In this paper, two computer-vision metrics, the structural similarity (SSIM) index and the complex wavelet structural similarity (CW-SSIM) index were examined for their utility in the comparison of snow water equivalence (SWE) maps as well as simulated maps from moving average Gaussian Markov random fields. Separate simulations were performed with different spatial dependency parameter settings. Comparisons were made with varying window sizes. A case study into comparison of spatial patterns of SWE data over northern Canada is used to explore the properties of these indices on real-world environmental data. Using rank tests, map ranks on CW-SSIM and SSIM scales showed both agreements and discrepancies, with agreement in the similarity of the spatial patterns more common in the SWE dataset. CW-SSIM was more robust to changes in window size than SSIM. In both simulated and SWE data, the CW-SSIM values were more variable than SSIM; likely owing to its translation and rotation invariance properties. The CW-SSIM may have potential in assessing or validating spatial model outputs, and detecting goodness-of-fit in spatial models. The index can potentially distinguish between model outputs or map pairs that possess subtle structural difference. Further research is required to explore the utility of these approaches for empirical comparison cases of different forms of landscape change and in comparison to human judgments of spatial pattern difference.



Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 License.