Combining invariance, robustness, and stability in computer vision
Boris Kovalerchuk · 2008
This paper analyzes the theoretical foundations of the invariant, robust, and stable methods in computer vision applications. Some studies had shown that many known invariants used in pattern recognition algorithms are not robust to small changes in images and robust parameters of these algorithms are not invariant. We provide a conceptual framework for new studies on invariance, robustness, and stability of computer vision algorithms critical for applications. Based on this theoretical analysis new invariant, robust and stable methods suited for the complexity of computer vision tasks, can be designed.