3D Depth Map Inpainting for Vietnamese Historical Printing Woodblocks: A Gated Convolution Approach
Nam le Viet, Viet Cuong Ta, Thi Duyen Ngo, An Nguyen Thai, Seung‐Won Jung, Thanh Ha Le · 2023
This research paper explores the application of inpainting techniques to reconstruct missing information in 3D depth maps of Vietnamese printing woodblock, a heritage object in need of preservation and reconstruction in Vietnam. In the paper, we provide a comprehensive review of related work encompassing various 3D formats, with a particular emphasis on depth map. The discussion delves into inpainting methods in the domain of depth images, notably including the deepfill with gated convolution approach [1] and Mask-Aware Transformer (MAT) inpainting [2] with transformer networks. Experimental results demonstrate the effectiveness of a gated convolution approach in restoring the missing data in woodblock images. We evaluate our technique using quantitative measures and visual comparisons, highlighting its ability to produce accurate and coherent results. We also discuss the strengths and limitations of our method, along with potential areas for future research. Our research contributes by evaluating the performance of popular 2D inpainting methods on 3D depth data. Moreover, our study contributes to the preservation and analysis of historical artifacts, specifically Vietnamese printing woodblock, by leveraging inpainting techniques to reconstruct their missing information.