Inverse Free-form Deformation for interactive UV map editing

Interactive texture editing for garment line drawings

“…Figure 5 shows some examples of texture editing with our system. The input textured models are generated by using Hashimoto et al 12 Since most existing editing methods, such as FFD, 1,2 do not support discontinuous region handling, such as wrinkles, their results are continuous across the discontinuous regions, giving a flat overall impression (see Figure 2). In addition, while other deformers based on the mesh graph structure 22 can be suitable for considering self‐closing wrinkles, they often create unnatural texture distortions around the wrinkles.…”

“…However, FFD constructs a grid of control points; thus, the number of grids must be increased to perform detailed operations, which is labor‐intensive (e.g., 100$$$ \times $$$100 grids). Therefore, inverse problems of estimating the positions of control points from the input UV coordinate data 2 and some researches to optimize UV coordinates with a few handles 16,17 have been thoroughly investigated. For example, Lévy 18 proposed a method to interactively build correspondence points between input drawings (or photos) and 2D texture patterns, and generated textured results.…”

“…Examples of textured results with FFD 1,2 (source images from Manga109 dataset 21 ). FFD‐based methods cannot handle discontinuities on self‐occluded wrinkles (red frame).…”

“…Although our system allows users to manually place curve/point handles on garment drawings one‐by‐one, it is still difficult to understand where to set the handles in garment drawings for appropriate texturing. This is a general problem in deformation research, but it might be interesting to incorporate an automatic (or semi‐automatic) placement of handles, such as the inverse FFD 2 …”

“…Against this background, the manipulation of 2D images (or UV coordinates) on 2D space based on user‐specified constraints (e.g., control points) has been well‐studied; 1,2 however, most of these methods are unsuitable for handling boundaries and discontinuous regions caused by 3D characteristics. Although a straightforward solution is to prepare 3D characteristics using 3D template models 3 or normal/depth estimation methods, 4,5 it remains difficult to automatically create natural‐looking results envisioned by the user.…”