Improved multi-scale image matching approach for monitoring Amery ice shelf velocity using Landsat 8

Abstract

The velocity of an ice shelf is an important characteristic to understand its dynamics and interaction with the internal ice sheet. Therefore, we develop an improved multi-scale image matching method for producing a complete and accurate Amery ice shelf velocity field from Landsat 8 images. First, we investigate the relationship between the template size and the image entropy and propose an image entropy-based preliminary operation for distinguishing the high-contrast regions from the low-contrast regions prior to iteratively determining the optimum template size. Second, a Gaussian pyramid image-based hierarchical data structure is designed to support a coarse-to-fine image matching strategy.The image entropy-based matching method is applied on the top layer of the image pyramid to guarantee the matching results have optimal completeness and high accuracy. Finally, a postprocess procedure is performed to derive a complete and accurate Amery ice shelf velocity field. Experimental results demonstrate that the proposed method can significantly improve computational efficiency. Moreover, the proposed method provides more accurate and robust matching results than other existing methods, particularly over the low-contrast surfaces. Additionally, the derived velocity field also shows good consistency with the one acquired from MEaSUREs Annual Antarctic Ice Velocity Maps 2016–2017.

Keywords

• image entropy
• multi-scale image matching
• normalized cross-correlation
• amery ice shelf velocity