Agricultural production plays a vital role in a country’s economy, and leaf diseases are a common and natural occurrence. Early and accurate identification of these diseases is essential to prevent significant losses in crop yield and ensure sustainable agricultural output. This paper presents an efficient and intelligent framework for Pepper Leaf Disease (PLD) detection by integrating advanced image processing, segmentation, feature extraction, and classification techniques. The Black Pepper Leaf Blight and Yellow Mottle Virus dataset is applied to preprocessing stage using a Spatially Adaptive Difference of Gaussians (SADoG) filter, which enhances disease-affected regions by preserving essential edge information while effectively suppressing background noise. To accurately localize infected areas, a Noise-Resilient Fuzzy C-Means (NR-FCM) clustering algorithm is employed for segmentation. This method leverages fuzzy membership and spatial context to robustly handle noise and variation in leaf textures. Feature extraction is then carried out using the Stockwell Transform (S-Transform), which provides a rich time-frequency representation capable of capturing both global and localized disease patterns in pepper leaf images. Finally, the extracted features are fed into a Dense Convolutional Neural Network enhanced with Adaptive Batch Normalization (AdaBN-DenseNet) classifier. This classifier dynamically adjusts to variations in input data distributions, enabling precise and robust disease classification. Using Python software, the proposed framework achieved higher accuracy, F1-score and recall of 92%, precision of 93% is accomplished when compared to other techniques. Experimental results demonstrate that the proposed system achieves high accuracy and resilience in detecting various PLD, making it suitable for real-time agricultural diagnostics and smart farming applications.