Abstract:In the current Video-based Dynamic Mesh Coding (V-DMC) standard, inter-frame coding is restricted to mesh frames with constant topology. Consequently, temporal redundancy is not fully leveraged, resulting in suboptimal compression efficacy. To address this limitation, this paper introduces a novel coarse-to-fine scheme to generate anchor meshes for frames with time-varying topology. Initially, we generate a coarse anchor mesh using an octree-based nearest neighbor search. Motion estimation compensates for regions with significant motion changes during this process. However, the quality of the coarse mesh is low due to its suboptimal vertices. To enhance details, the fine anchor mesh is further optimized using the Quadric Error Metrics (QEM) algorithm to calculate more precise anchor points. The inter-frame anchor mesh generated herein retains the connectivity of the reference base mesh, while concurrently preserving superior quality. Experimental results show that our method achieves 7.2% ~ 10.3% BD-rate gain compared to the existing V-DMC test model version 7.
What problem does this paper attempt to address?
This paper attempts to solve the problem of inter - frame coding in dynamic mesh sequences, especially for meshes with time - varying topologies. Specifically, the existing video - based dynamic mesh coding (V - DMC) standard performs well when processing mesh frames with a constant topology, but when dealing with mesh frames with time - varying topologies, it fails to fully utilize temporal redundancy, resulting in poor compression efficiency.
### Core problems of the paper:
1. **Limitations of existing methods**: The existing V - DMC standard can only handle mesh frames with a constant topology and cannot effectively handle mesh frames with time - varying topologies. This leads to the under - utilization of temporal redundancy, which in turn affects the compression effect.
2. **Challenges of time - varying topologies**: For mesh sequences with time - varying topologies, traditional inter - frame coding methods cannot generate the target base mesh by tracking corresponding vertices because the correspondence of these vertices becomes complex over time.
### Solutions:
To solve the above problems, this paper proposes a new coarse - to - fine method to generate an anchor mesh to adapt to mesh frames with time - varying topologies. The specific steps are as follows:
1. **Generation of anchor mesh in the coarse stage**:
- Use octree nearest - neighbor search and motion estimation techniques to quickly generate a coarse anchor mesh, ensuring that it has the same topology as the reference base mesh.
- Formula representation:
\[
\arg \min_{V_j^t} \sum_i \| \text{Dist}(V_i^k, V_j^t) \|
\]
where \(\text{Dist}(A, B)=\sqrt{(A - B)^2}\) is the Euclidean distance between two points.
2. **Generation of anchor mesh in the fine stage**:
- Use the quadratic error metric (QEM) algorithm to optimize the coarse anchor mesh and generate a more accurate fine - grained anchor mesh.
- The QEM algorithm selects the optimal collapsing edge by minimizing the quadratic error matrix of the edge, thereby improving the quality of the mesh.
- Formula representation:
\[
\nabla(V)=\sum_{P\in\text{pl}(V)}(P^T V)^2 = V^T\left(\sum_{P\in\text{pl}(V)}P P^T\right)V
\]
\[
Q_E = q_V+q_{V'}
\]
\[
\arg \min_V V^T Q_E V
\]
3. **Adaptive displacement quantization**:
- Propose an adaptive quantization method based on the number of vertex neighbors to optimize the quantization of displacement, making resource allocation more reasonable.
- Formula representation:
\[
D_q = D*\alpha*A+\delta
\]
\[
A=\frac{N(V_D)}{\hbar}
\]
where \(D\) and \(D_q\) are the displacements before and after quantization respectively, \(\delta\) and \(\alpha\) are preset displacement offsets and quantization step sizes respectively, \(A\) is the adaptive weight, and \(\hbar\) is a hyper - parameter that controls the sensitivity of the weight to the number of neighbors.
### Experimental results:
The experimental results show that this method achieves significant BD - rate gains ranging from 7.2% to 10.3% on four challenging dynamic mesh sequences, and performs particularly well at lower bit rates. In addition, the adaptive quantization method further optimizes the bit - rate allocation and improves the compression quality.
In conclusion, this paper effectively solves the inter - frame coding problem of dynamic mesh sequences with time - varying topologies by introducing the coarse - to - fine anchor mesh generation method and adaptive quantization technology, significantly improving the compression efficiency and reconstruction quality.