Fuzzy Connectives for Efficient Image Reduction and Speeding Up Image Analysis

Abstract

We discuss non-monotone fuzzy connectives in large-scale image processing. We present an image reduction algorithm capable of differentiating between fine image details and noise in the image, particularly salt and pepper noise. The reduction algorithm is based on mode-like averaging functions. We compare the performance of the proposed method to the alternative reduction methods on artificial images and on two case studies: content-based image retrieval and pedestrian detection. Our algorithm improves the speed of the subsequently applied image analysis methods and allows efficient filtering of salt and pepper noise. Applications to on-board image recognition in autonomous robotic devices are envisaged.