Mouth Image Segmentation Using Ternary Pixel Classification

The patent segments the binary classification problem into multiple sub-classifications by introducing intermediate categories (mouth internal pixel points, mouth boundary pixel points, face region pixel points). This segmentation allows the system to handle the complexity of mouth edge recognition through staged classification, improving precision without requiring a single complex binary classifier.

The patent transitions from a one-dimensional binary classification (mouth vs. non-mouth) to a multi-dimensional classification system with multiple categories and probability scores. By adding dimensional depth through probability values and multiple classification levels, the system achieves higher recognition accuracy while managing complexity through structured dimensionality.

The classification system is segmented into distinct stages: first classifying pixel points into three categories (mouth internal, mouth boundary, face region), then applying different processing rules to each category. This segmentation reduces the apparent complexity by breaking down the complex ternary classification into manageable, rule-based decisions for each category.

The patent changes the parameter space by introducing probability values and category-specific processing rules. Instead of a simple multi-class classifier, the system uses probability thresholds and category-based decision rules, which simplifies the implementation complexity while maintaining the benefits of ternary classification for improved accuracy.