Locally edge-adapted distance for image interpolation based on genetic fuzzy system

This study presents a new adaptive scheme for developing kernel-based interpolation methods that simultaneously enhance spatial image resolution and preserve locally detailed edges. A new edge-adapted distance is first estimated according to local gradients information by combining fuzzy theory with genetic learning algorithm. This estimated distance is then employed in place of the original Euclidean distance in various interpolation methods. Additionally, a learning procedure based on genetic algorithm is presented to obtain crucial parameters of the fuzzy system automatically. Experimental results presented in numerical comparisons and in visual observations verify the effectiveness of the proposed adaptive framework for kernel-based interpolation methods.