Pipelines hold a crucial role in the effective transportation of liquids and gases over extensive distances. Detecting leaks through Acoustic Emission (AE) events is intricate due to the interference of noise. The conventional strategy of utilizing AE scalogram images from continuous wavelet transforms (CWT) faces challenges in precisely identifying leaks amidst noise. To surmount these complexities, we introduce an inventive technique. Our Enhanced Leak-Induced Scalograms are generated by processing CWT images with Laplacian filters, Non-Local Means (NLM) noise reduction, and adaptive histogram equalization (AHE) to enhance contrast. This unique approach ensures scalogram image quality while reducing background noise. We incorporate the Grey Level Run Length Matrix (GLRLM) model for feature extraction and deploy K-Nearest Neighbors (KNN) and t-Distributed Stochastic Neighbor Embedding (t-SNE) for accurate classification of leak and normal conditions. The proposed method excels across diverse metrics using real-world datasets. This methodology elevates image quality, bolsters classification precision, and marks a significant advancement in pipeline leak detection. Validation through experimentation on an industrial-scale testbed involving real pipelines affirms the efficacy of our proposed approach.