Digitization and Enhancement of Ancient Marathi Inscription Images Using Wavelet-Based Techniques

Ancient stone inscriptions are important cultural and historical artifacts, but their readability often suffers from environmental damage, erosion, and wear over time. This study introduces a wavelet-based image enhance ment algorithm designed to improve the clarity and legibility of ancient Marathi inscriptions. The proposed method includes converting inscription images to grayscale, adding Gaussian noise for performance testing, and applying multi-level 2D Discrete Wavelet Transform (DWT) followed by soft or hard thresholding to re duce noise while keeping the structural features intact. Experimental validation took place on a dataset made up of images from archaeological texts, expert consultations, and field visits to historical sites in Maharashtra. A quantitative assessment using PSNR, MSE, SSIM, and FSIM showed that the method works well, with an average PSNR of 30.99 dB, SSIM of 0.72, and FSIM of 0.81. These results confirm that the algorithm effectively reduces noise while keeping important details of the inscriptions, making it a useful tool for digital preservation and epigraphic analysis. The study also points out that soft thresholding is better at achieving visually smoother and structurally consistent improvements. Overall, this approach provides a practical and strong solution for restoring damaged inscription images in archaeological and heritage contexts.