Fuzzy impulse noise detector for efficient image restoration

The present article proposes an efficient restoration model for images corrupted with impulse noise of varying values that follow a random distribution over some dynamic range. The model extracts a set of informative features, uses a fuzzy detector based on product aggregation reasoning rule for noisy pixels detection and noise removal operator for filtration. The fuzzy set-based detector provides a better learning and generalization capability for improved detection. The model thus explores mutually the advantages of both fuzzy detector and noise removal operator. Superiority of the proposed model to other similar methods is established both visually and quantitatively in removing impulse noise from highly corrupted images. With experimental results, it is found that the proposed model performs better and at the same time takes less computational time than others.