Image Edge Enhancement with Face-Aware Skin Tone Refinement

The patent segments the skin tone detection process into two independent stages: first detecting skin tone regions using color space analysis, then refining the results using face detection algorithms. This segmentation allows each stage to specialize in its function, with the first stage providing broad coverage and the second stage providing precise refinement, thereby reducing false detections while maintaining implementation simplicity.

The patent performs preliminary skin tone region detection using color space methods before applying face detection refinement. This preliminary action establishes a broad initial set of candidate regions that can be efficiently processed by the subsequent face detection algorithm, avoiding the need to process every pixel through complex algorithms while ensuring accurate identification.

The patent applies different edge enhancement parameters to different regions of the image based on their semantic meaning. Facial skin tone points receive one set of parameters optimized for preserving facial details, while non-facial skin tone points receive another set optimized for general edge enhancement. This local differentiation ensures that each region receives the appropriate treatment for its specific requirements.

The patent dynamically adjusts enhancement parameters based on the detected region type and characteristics. The system adapts parameter values such as edge strength, noise suppression levels, and detail preservation factors according to whether the region is facial or non-facial, allowing optimal enhancement quality while maintaining manageable processing complexity through automated parameter selection.

The patent applies different edge enhancement strengths to facial and non-facial regions based on their specific requirements. Non-facial skin tone points receive stronger enhancement to make details like textures and patterns more visible, while facial skin tone points receive reduced enhancement to preserve natural appearance. This local differentiation prevents unnatural-looking faces while enhancing important non-facial details.

The patent segments the image processing into distinct handling paths for facial and non-facial regions. By separating the processing flows, the system can apply aggressive enhancement parameters to non-facial areas without affecting the subtle enhancement applied to facial areas, ensuring both detail visibility and natural appearance are achieved simultaneously.