An Effective Pixel-Wise Approach for Skin Colour Segmentation Using Pixel Neighbourhood Technique

Tejas Dastane,Varun Rao,Kartik Shenoy,Devendra Vyavaharkar
DOI: https://doi.org/10.48550/arXiv.2108.10971
2021-08-25
Abstract:This paper presents a novel technique for skin colour segmentation that overcomes the limitations faced by existing techniques such as Colour Range Thresholding. Skin colour segmentation is affected by the varied skin colours and surrounding lighting conditions, leading to poorskin segmentation for many techniques. We propose a new two stage Pixel Neighbourhood technique that classifies any pixel as skin or non-skin based on its neighbourhood pixels. The first step calculates the probability of each pixel being skin by passing HSV values of the pixel to a Deep Neural Network model. In the next step, it calculates the likeliness of pixel being skin using these probabilities of neighbouring pixels. This technique performs skin colour segmentation better than the existing techniques.
Computer Vision and Pattern Recognition
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the poor performance of existing skin color segmentation techniques under different lighting conditions and skin color variations. Specifically, traditional methods based on color - range thresholds often lead to unsatisfactory skin segmentation results when dealing with different skin colors and complex lighting environments. To overcome these limitations, the author proposes a new two - stage pixel - neighborhood technique, which improves the accuracy of skin color segmentation by considering a pixel and other pixels within its neighborhood. The paper mentions that existing skin color segmentation methods are mainly divided into two categories: 1. **Pixel - level segmentation**: Each pixel is individually evaluated according to certain conditions to determine whether it belongs to the skin. 2. **Region - level segmentation**: The spatial relationship of pixels is considered, and skin regions are identified by expanding the initial skin regions. However, these methods have the following problems in practical applications: - **Skin color variation**: There are significant differences in skin colors among different people. - **Background interference**: There may be objects in the image with similar skin colors. - **Lighting influence**: Changes in lighting conditions will affect the identification of skin colors. - **Difficulty in determining thresholds**: Thresholds are usually determined empirically and lack theoretical basis. - **Poor performance in low - light conditions**: The segmentation effect is poor in darker environments. To solve these problems, the author proposes a new technique based on pixel neighborhoods. The main steps of this technique are as follows: 1. **Probability calculation**: A deep neural network model is used to calculate the probability that each pixel belongs to the skin. 2. **Neighborhood evaluation**: Based on the skin probabilities of other pixels in the neighborhood, further evaluate whether the current pixel belongs to the skin. Through this method, the paper aims to provide a more robust and accurate skin color segmentation technique, which is suitable for different lighting conditions and skin colors.