PixVOD: Pixel-Distributed Direct Visual Odometry and Depth Estimation
- arXiv: 2606.03989
- Authors: Shinjeong Kim, Ignacio Alzugaray, Callum Rhodes, Paul H. J. Kelly, Andrew J. Davison
- Field: Computer Vision
- Posted: 2026-06-02
- A fully parallelizable formulation of visual odometry and depth estimation distributed across pixels.
- Sensor-processors exchange information through Gaussian Belief Propagation (GBP) to achieve consensus about camera motion.
- Depth is inferred from per-pixel photometric observations combined with surface normal priors.
- To preserve geometric stability during optimization, the method introduces a keyframe-like anchoring mechanism that regulates the effective baseline between frames, enabling consistent joint motion and depth updates.
Summary
Images composed of 2D pixel arrays are the standard input to computer vision algorithms, yet many underlying computations can be distributed across pixels. Transmitting raw, redundant, and noisy pixel data off the sensor remains inefficient, motivating a shift toward focal-plane sensor-processors that perform a significant part of the computation directly within each pixel. The authors envision pixels synthesizing higher-level signals locally, reducing downstream load, and providing richer inputs for higher-level vision tasks.
Key contributions
*Auto-collected on 2026-06-04.*