🤖 AI Summary
This work addresses the limitations of existing monocular 3D Morphable Model (3DMM) regression methods in reconstructing near-range facial images—such as those captured by head-mounted cameras—where the use of orthographic projection fails to account for perspective distortion, leading to geometric inaccuracies. To mitigate this issue, the authors propose an extended orthographic projection model augmented with a shrinkage parameter that effectively approximates pseudo-perspective effects while preserving training stability. The proposed formulation is readily integrated into existing 3DMM regression frameworks through direct fine-tuning, without requiring architectural overhauls. Experimental results on a newly collected head-mounted camera dataset demonstrate that the method significantly outperforms conventional orthographic projection in both quantitative metrics and qualitative visual fidelity, yielding more accurate and realistic 3D facial reconstructions at close range.
📝 Abstract
We introduce a novel camera model for monocular 3D Morphable Model (3DMM) regression methods that effectively captures the perspective distortion effect commonly seen in close-up facial images. Fitting 3D morphable models to video is a key technique in content creation. In particular, regression-based approaches have produced fast and accurate results by matching the rendered output of the morphable model to the target image. These methods typically achieve stable performance with orthographic projection, which eliminates the ambiguity between focal length and object distance. However, this simplification makes them unsuitable for close-up footage, such as that captured with head-mounted cameras. We extend orthographic projection with a new shrinkage parameter, incorporating a pseudo-perspective effect while preserving the stability of the original projection. We present several techniques that allow finetuning of existing models, and demonstrate the effectiveness of our modification through both quantitative and qualitative comparisons using a custom dataset recorded with head-mounted cameras.