Source-linked AI summary
Human Skin Detection Using RGB, HSV and YCbCr Color Models
S. Kolkur, D. Kalbande, P. Shimpi, C. Bapat, J. Jatakia
TL;DR
Human skin detection must handle variation from illumination, individual characteristics, race, backgrounds, shadows, and motion blur. This paper proposes a threshold-based RGB-HSV-YCbCr algorithm using individual and combinational parameter ranges, obtaining 89.33% precision and 94.43% accuracy on a Pratheepan-dataset subset.
Problem
Skin detection supports applications including face recognition, skin disease detection, gesture tracking, and human-computer interaction, but illumination, skin-tone, individual, and scene variations complicate color-based recognition.
Method
The proposed algorithm compares each pixel's ARGB, HSV, and YCbCr values with predefined skin-pixel thresholds, including combined RGB and HSV conditions.
Results
89.33% precision and 94.43% accuracy were obtained on a subset of images from the Pratheepan dataset.
Takeaways & Limitations
The algorithm recognizes skin pixels under different light conditions and provides promising precision and accuracy against the baseline dataset.
Takeaways & Limitations
Recognition is constrained by illumination, individual characteristics, racial variation in skin tone, background colors, shadows, and motion blur.
Abstract
from arXiv · showhide
Human Skin detection deals with the recognition of skin-colored pixels and regions in a given image. Skin color is often used in human skin detection because it is invariant to orientation and size and is fast to process. A new human skin detection algorithm is proposed in this paper. The three main parameters for recognizing a skin pixel are RGB (Red, Green, Blue), HSV (Hue, Saturation, Value) and YCbCr (Luminance, Chrominance) color models. The objective of proposed algorithm is to improve the recognition of skin pixels in given images. The algorithm not only considers individual ranges of the three color parameters but also takes into ac- count combinational ranges which provide greater accuracy in recognizing the skin area in a given image.
1 Introduction
Skin detection identifies skin-colored pixels and regions, but reliable recognition is challenged by illumination, skin-tone variation, and other image conditions. The paper proposes combining RGB, HSV, and YCbCr thresholds.
- Skin detection finds skin-colored pixels and regions in images or video, often as preprocessing for locating potential faces and limbs.
- Skin tone varies across races, while illumination, background colors, shadows, and motion blur influence detection results.
- Detection is optimized by considering combinations of brightness, contrast, transparency, illumination, and saturation within ideal ranges.
2 Literature Review
The literature describes pixel-based skin detection as classifying individual pixels according to color conditions, motivated by applications including early skin-disease diagnosis.
- Pixel-based skin detection classifies each pixel as skin or non-skin according to specified conditions, including color values.
- Accurate human-skin detection is presented as relevant to identifying skin diseases before they reach critical stages.
3 Color Spaces
The paper reviews RGB, YCbCr, and HSV as color models for skin-pixel recognition. RGB represents primary color components, YCbCr separates luminance from chrominance, and HSV organizes hue, saturation, and brightness.
- Color spaces represent color information through three or four components, and selecting an appropriate space is central to skin-color modeling and threshold classification.
- RGB Color Model: RGB uses red, green, and blue components to represent colors and is commonly the default format for storing digital images.
- RGB Color Model: Normalized RGB is described as invariant to changes in surface orientation relative to the light source for matte surfaces under specified assumptions.
- YCbCr Color Model: YCbCr represents luma as a weighted RGB sum and stores chrominance as blue- and red-difference components, Cb and Cr.
- HSV Color Model: HSV describes color using hue, saturation, and value, with value corresponding to brightness and hue spanning 0° to 360°.
4 Proposed Skin Detection Algorithm
The proposed algorithm processes every image pixel, extracts ARGB components, converts values to HSV and YCbCr, and classifies pixels using individual and combinational threshold conditions.
- The algorithm represents the image as a two-dimensional matrix and processes each entry as an individual pixel.
- ARGB values are extracted from each 32-bit pixel using bit shifting and bitwise AND operations for alpha, red, green, and blue components.
- The method converts ARGB values to HSV and YCbCr to make skin recognition more precise.
- A pixel is classified as skin when its parameters satisfy predefined thresholds across RGB, HSV, alpha, and YCbCr values.
- The YCbCr conditions include linear combinational bounds relating Cr and Cb in addition to individual chrominance and luminance thresholds.
5 Experimental Results
The algorithm was evaluated on the Pratheepan human-skin-detection dataset and additional internet-sourced sample images. It achieved reported precision of 89.33% and accuracy of 94.43%, while RGB, HSV, and YCbCr each contributed to skin-pixel identification.
- Pratheepan Dataset: The Pratheepan dataset provided the baseline, including images captured with different cameras, color enhancements, illuminations, and corresponding ground-truth images.Results compared original, ground-truth, and algorithm-result images.
- Sample Images: Additional internet-sourced sample images were evaluated, with results shown for the algorithm’s skin-pixel detections.The implementation was written in JAVA.
- Color-Space Contributions: RGB, HSV, and YCbCr were reported to contribute almost equally to skin-pixel identification across three sample images.A bar chart represents the number of detected skin pixels for each color space.
6 Conclusion
The paper presents a threshold-based skin-recognition algorithm using RGB-HSV-YCbCr color models. It reports promising precision and accuracy and identifies face, hand, gesture, and skin-disease detection as future applications.
- Conclusion: The proposed threshold-based algorithm recognizes skin images using the combined RGB-HSV-YCbCr color model.The conclusion describes the algorithm as capable of processing images under different light conditions, including brightness variation.
- Conclusion: The algorithm gives promising precision and accuracy results when compared with the baseline dataset.The conclusion refers to the experimental results shown in the paper’s figures.
- Future Scope: Future applications include face, hand, and hand-gesture detection for security, assistance to deaf individuals, and skin-disease detection.