Free Viewpoint Video Capture Using Selective Sensor Readout

The patent divides the camera network into multiple zones with different sampling rates. Regions of interest (ROI) are segmented from the rest of the scene, allowing high-resolution capture in specific areas while using lower sampling rates in other areas. This enables high resolution in ROIs without requiring the entire coverage area to be captured at the same high resolution.

The patent implements local quality by applying different sampling strategies to different spatial regions. High sampling rates are applied locally to regions of interest where high resolution is needed, while lower sampling rates are used in regions where high resolution is not critical. This creates a spatially varying quality profile that optimizes the balance between resolution and coverage.

Instead of deploying multiple complete camera networks, the patent segments the single large camera network into multiple operational zones with different sampling characteristics. This virtual segmentation allows the system to achieve multi-region high-resolution coverage using one physical network, reducing hardware costs while maintaining the capability to serve multiple areas at high resolution.

The patent makes a single camera network universal by enabling it to serve multiple functions and cover multiple regions of interest simultaneously. Through software-controlled selective sampling, the same physical infrastructure can adapt to cover different areas at different resolutions, replacing the need for multiple dedicated camera networks for different regions.

The patent replaces the mechanical PTZ stage system with a software-based selective sampling system. Instead of physically moving cameras to retarget them, the system uses software to select which pixels from fixed cameras should be captured and processed. This substitution eliminates mechanical instability while maintaining the ability to retarget regions of interest through digital pixel selection and processing.

The patent creates virtual copies of camera viewpoints through software processing of pixel data. Instead of physically moving cameras to different positions, the system processes pixel data to synthesize views from different effective positions and angles. This digital copying approach provides retargeting capability without the mechanical instability of physical camera movement.