In this paper, we autonomously define an optimal, efficient image-processing tree for extracting the cracks from concrete images using genetic programming (GP)-oriented evolutionary image processing known as Automatic Construction of Tree-structural Image Transformation (ACTIT). We propose the use of automatic finding feature from input and internal transformation images to optimize image-processing filters. These alternative solutions show significant improvements and can be performed by extracting small areas through our experimentation. This can possibly be used with an optimal image-processing system based on feature filters. © 2012 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.