Revealing Disappearance Patterns in Murals Heavily Contaminated by Soot Based on Hyperspectral Imaging

Over time, many murals have become heavily contaminated with soot, obscuring patterns and hindering the recognition of these significant cultural relics. This study utilises the spectral discrimination and subsurface detection capabilities of hyperspectral imaging to nondestructively reveal pattern information concealed by soot. After preprocessing the hyperspectral data, independent component analysis (ICA) is used to preliminarily separate soot from the underlying patterns. Several independent components containing the most distinct pattern information are selected for a two‐part enhancement strategy. The first part introduces a novel feature extraction method, EMP‐PCA, to highlight and extract the patterns, followed by contrast stretching to enhance the pattern–background distinction. The second part uses an inverse ICA transformation on selected components to synthesise a true colour image, which is then enhanced using a weight map–based method. Finally, the results from both parts are combined using Laplacian fusion, comprehensively revealing the patterns hidden beneath the heavy soot. The results demonstrate the method's effectiveness and provide valuable information for the archaeological study and conservation of cultural heritage.