Skin color classification of Koreans using clustering

Area of Science:

• Dermatology
• Image Analysis
• Computational Science

Background:

• Skin color is a critical indicator for diagnosing and predicting skin conditions like irritation and dermatitis.
• Existing methods for skin color evaluation have limitations in consistency and accuracy.
• Skin color is influenced by various intrinsic and extrinsic factors.

Purpose of the Study:

• To propose a novel, objective method for evaluating Korean skin color.
• To overcome the limitations of current subjective visual scoring methods.
• To demonstrate the usefulness of the proposed skin color classification technique.

Main Methods:

• Quantified Korean skin color from digital images using RGB, HSV, CIELab, and YCbCr color spaces.
• Classified skin color through clustering algorithms.
• Compared classification performance against existing methods using multinomial logistic regression, SVM, KNN, and random forest.

Main Results:

• Verified the skin color classification performance based on quantified features and chosen classifiers.
• Demonstrated the superiority of the proposed method by comparing its performance with existing techniques.
• Confirmed the usefulness of the novel objective skin color classification approach.

Conclusions:

• Developed an objective method to classify Korean skin color values from digital images using clustering.
• Presented an optimized classification strategy for Korean skin tones.
• The proposed method provides a foundation for objective skin color quantification and standardization in research.