Perception-Based Image Processing for Adaptive Video Encoding

The patent applies different encoding qualities to different regions of the video frame based on predicted user attention. High-quality encoding is applied only to attention regions where users are likely to look, while non-attention regions use lower quality settings. This resolves the contradiction by maintaining perceived visual quality through selective high-quality encoding rather than uniform high-quality encoding across the entire frame.

The system dynamically changes encoding parameters (such as bitrate, resolution, or compression level) based on the predicted user attention regions. By adjusting these parameters locally according to attention predictions, the system achieves high perceived quality where needed while reducing overall power consumption through selective parameter optimization.

The system performs auto-focus and auto-exposure automatically based on predicted user attention regions without requiring manual user selection. The attention prediction model identifies what the user is likely to look at, and the camera automatically adjusts focus and exposure for those regions, making the system self-sufficient and eliminating the need for user intervention while maintaining accuracy.

The system uses feedback from the attention prediction model to automatically adjust camera settings. The model continuously predicts user attention regions, and this information feeds back to the auto-focus and auto-exposure algorithms, enabling them to make accurate adjustments without user input. This creates a closed-loop system that maintains measurement precision while improving ease of operation.