Region-Based Video Quality Control for Multi-Object Recognition

The patent applies dynamics by making the image quality improvement strategy adaptive rather than static. The system dynamically determines whether to improve image quality based on real-time detection of object relationships, transitioning from a fixed registered object approach to a flexible, situation-based approach that adjusts quality control according to current video content.

The patent changes the parameter of image quality improvement from a binary state (always improve or never improve) to a conditional state based on detected object relationships. By monitoring spatial relationships between objects and changing the quality improvement parameter accordingly, the system achieves both high recognition accuracy and flexible video quality control.

The patent applies local quality by selectively improving image quality only for specific regions where objects with important spatial relationships are detected. Rather than uniformly improving quality for all registered objects, the system identifies local regions with critical object interactions and applies quality improvement only there, optimizing the balance between recognition accuracy and transmission efficiency.

The patent dynamically changes the image quality parameter based on the detected spatial relationships between objects. When objects are detected to have significant spatial relationships, the quality improvement parameter is activated for those specific regions; when relationships are minimal, the parameter is deactivated, thereby optimizing resource utilization while maintaining necessary recognition accuracy.