Image Retrieval Using Fuzzy and Neuro-fuzzy Approaches with Fuzzy Color Semantics

In this paper, two different approaches to color binning and subsequent JNS (Just Not the Same) color histogram computation are discussed. The first approach is based on a neural network color classifier trained using error back propagation training algorithm. The second approach is based on heuristically designed fuzzy classifier using fuzzy if-then rules for classifying color pixels into one of the eleven JNS colors. Color signatures for images in the database are obtained using both the methods. Further a fuzzy set theoretic approach is proposed to describe and extract the fuzzy color semantics that attempt to reduce the semantic gap between the low-level visual features and the high-level semantic features. Five linguistic variables are used to represent the image color semantics providing a flexible query scheme that is able to effectively represent vagueness in human color perception. The calibration images are inserted in the database for verifying correctness of the two approaches. The comparison of the image retrieval results obtained using fuzzy and fuzzy-neural approaches are presented at the end.