Convexity recognition using multi-scale autoconvolution

This paper introduces a novel measure for object convexity using the recently introduced multi-scale autoconvolution transform. The proposed measure is computationally efficient and recognizes even small errors in a convex domain. We also consider its implementation and give a complete Matlab algorithm for computing this measure for digital images. Finally, we give examples to verify its applicability and accuracy. The examples also consider convexity as a measure for complexity.