A Fusion Approach for Efficient Human Skin Detection (1410.3751v1)

Abstract: A reliable human skin detection method that is adaptable to different human skin colours and illu- mination conditions is essential for better human skin segmentation. Even though different human skin colour detection solutions have been successfully applied, they are prone to false skin detection and are not able to cope with the variety of human skin colours across different ethnic. Moreover, existing methods require high computational cost. In this paper, we propose a novel human skin de- tection approach that combines a smoothed 2D histogram and Gaussian model, for automatic human skin detection in colour image(s). In our approach an eye detector is used to refine the skin model for a specific person. The proposed approach reduces computational costs as no training is required; and it improves the accuracy of skin detection despite wide variation in ethnicity and illumination. To the best of our knowledge, this is the first method to employ fusion strategy for this purpose. Qualitative and quantitative results on three standard public datasets and a comparison with state-of-the-art methods have shown the effectiveness and robustness of the proposed approach.

• The paper proposes an efficient human skin detection method that fuses a smoothed 2D histogram with a Gaussian model, requiring no training.
• This fusion approach demonstrates improved accuracy and computational efficiency for skin detection across various illumination conditions and ethnicities.
• The training-free, low-cost nature of the method makes it highly suitable for real-time applications such as human-computer interaction and facial recognition.

Efficient Human Skin Detection Using a Fusion Approach

The paper, authored by Tan et al., presents a novel methodology for human skin detection that skillfully addresses the challenges associated with variations in human skin color across diverse ethnicities and fluctuating illumination conditions. This method, devoid of a training phase, ingeniously combines a smoothed 2D histogram with a Gaussian model, producing an effective skin detection algorithm. This paper underscores its adaptation to various conditions while maintaining computational efficiency, setting it apart from the traditionally high-cost, training-dependent methods.

Summary of Approach

The proposed approach begins with transforming images into the log opponent chromaticity color space. This decision is driven by the perceptual relevance of color opponency, which aligns with the natural encoding performed by the human visual system. The transformation simplifies illumination changes into coordinate translations, facilitating robust skin detection.

The methodology employs a dynamic thresholding mechanism utilizing a smoothed 2D histogram. This eliminates the need for pre-defined or learned thresholds, allowing for adaptable response to diverse illumination scenarios and ethnic variability in skin color. The Gaussian model complements this by offering a probabilistic framework for skin pixel classification, calculated based on elliptical Gaussian joint probability distribution functions.

Fusion Strategy

Central to this paper is the fusion strategy, which integrates the individual outputs from the smoothed 2D histogram and the Gaussian model. By applying a product rule on these two features, the fusion strategy enhances detection robustness, achieving a balance between precision and computational efficiency. This novel combination demonstrates superior performance against various background complexities and illumination changes.

Experimental Validation

The efficacy of the proposed method was tested against three public datasets: Pratheepan's dataset, ETHZ PASCAL dataset, and the Stottinger dataset. Qualitative analysis reveals a notable improvement in detection accuracy compared to existing methods, such as those proposed by Cheddad and Pratheepan. The method's effectiveness is further reinforced by quantitative analysis using the Stottinger dataset, where it attained a performance close to Random Forest classifiers without necessitating training.

Implications and Future Work

The implications of this research are substantial for fields requiring accurate skin segmentation, such as human-computer interaction, facial recognition, and content creation. The elimination of training phases translates to low computational costs, making this method particularly advantageous for real-time applications.

While the current dependency on an accurate eye detector is acknowledged as a limitation, the authors propose future work in optimizing the preprocessing steps and potentially implementing the method on FPGA-based hardware for enhanced performance. This direction suggests promising avenues for integrating the approach into more sophisticated systems, possibly extending its applicability beyond current domains.

Overall, Tan et al.'s fusion approach introduces a significant stride in skin detection technology, offering a balanced blend of theoretical innovation and practical utility while paving the way for future explorations in adaptive image processing algorithms.