Edge-preserving image enhancement using a morphological guided filter for texture smoothing and structural detail preservation

Abstract

Image filtering techniques play a critical role in modern image processing applications, particularly in tasks requiring a fine balance between noise suppression and detail preservation. Traditional smoothing and sharpening methods, such as Gaussian, Median, and Laplacian filters, often introduce undesirable artefacts or degrade important image features. In this study, a morphological guided filter (MGF) is proposed by integrating the principles of mathematical morphology with guided filter (GF) to achieve adaptive, edge-aware image enhancement. The proposed method computes a local similarity measure between the guidance image and the input image, which dynamically controls the contribution of morphological operations, such as opening and closing, together with the original pixel values. This adaptive blending mechanism allows the filter to selectively smooth noise-prone regions while preserving critical structures and edges. The proposed MGF is designed as a compact edge-aware filtering framework that supports structural preservation and denoising performance. Experimental evaluations using two image datasets, namely the University of Southern California Signal and Image Processing Institute (USC-SIPI) dataset and Berkeley Segmentation Dataset and Benchmark 500 (BSDS500), showed that the proposed method achieved favourable average mean squared error (MSE) and structural similarity index (SSIM) performance under controlled-noise conditions. The results indicate that the proposed filter can support structure-preserving image denoising and enhancement across the evaluated datasets. Runtime analysis also showed that the proposed method requires higher computational cost than lightweight baseline filters, indicating the need for further optimisation before real-time or embedded deployment can be considered.