Advanced Fuzzy Denoising Technique for Agricultural Remote Sensing: Modified Partition Filter for Suppressing Impulsive

Image denoising is a critical challenge in digital image processing, particularly when dealing with random-value impulse noise (RVIN). Existing methods for RVIN detection and removal often struggle with poor generalization performance, requiring manual adjustment of detection thresholds or local window information. These approaches have difficulty handling severely damaged images or those with high background noise. To overcome these limitations, a novel fuzzy-based approach has been developed for RVIN detection and denoising in digital photographs. This method combines the power of K-means clustering with fuzzy logic to locate noisy pixels by identifying their closest neighbors. By effectively separating actual signal points from misleading noise signals during the detection phase, the proposed technique ensures precise identification of RVIN. Furthermore, a robust partition decision filter is employed in the elimination phase, effectively removing the identified noise while preserving the underlying signal. The integration of fuzzy techniques enhances the robustness and adaptability of this method, allowing it to handle various noise types and challenging image conditions. Extensive simulations using diverse remote-sensing datasets corrupted with RVIN demonstrated the superior performance of the proposed fuzzy denoising technique. It achieved an impressive 90% success rate in noise detection and maintained high accuracy even at increased noise levels, outperforming other commonly reported methods. This innovative fuzzy-based approach offers a promising solution to the problem of RVIN detection and denoising in digital images. By leveraging the advantages of fuzzy logic and K-means clustering, it provides improved generalization, increased adaptability, and enhanced noise removal capabilities, making it a significant advancement in the field of image processing.