H Shih1,2. 1. Human-Computer Interaction Multimedia Laboratory, Department of Electrical Engineering, Yuan Ze University, Taoyuan, Taiwan. 2. The Innovation Center for Big Data and Digital Convergence, Yuan Ze University, Taoyuan, Taiwan.
Abstract
BACKGROUND/ PURPOSE: One emerging subject in medical image processing is to quantitatively assess the health and the properties of cranial hairs, including density, diameter, length, level of oiliness, and others. This information helps hair specialists with making a more accurate diagnosis and the therapy required. We develop a practical hair counting algorithm. This analytic system calculates the number of hairs on a scalp using a digital microscope camera, providing accurate information for both the hair specialist and the patient. Our proposed hair counting algorithm is substantially more accurate than the Hough-based one, and is robust to curls, oily scalp, noise-corruption, and overlapping hairs, under various levels of illumination. Rather than manually counting the hairs on a person's scalp, the proposed system determines the density, diameter, length, and level of oiliness of the hairs. METHODS: We propose an automated system for counting the amount of hairs in the microscopy images. To reduce the effect of bright spots, we develop a robust morphological algorithm for color to smooth out the color and preserve the fidelity of the hair. Then, we utilize a modified Hough transform algorithm to detect the different hair lengths and to reduce any false detection due to noise. Our proposed system enables us to look at curved hairs as multiple pieces of straight lines. To avoid missing hairs when the thinning process is applied, we use edge information to discover any hidden or overlapping hairs. Finally, we employ a mutually associative regression method to label a group of line segments into a meaningful 'hair'. RESULTS: We demonstrated a novel approach for accurately computing the number of hairs, and successfully solved the three main obstacles in automated hair counting, including (i) oily and moist hairs, (ii) wavy and curly hairs, and (iii) under-estimation of the number of hairs occurs when hairs cross and occlude each other. The framework of this paper can be seen as the first step toward intelligent computer-aided medical image processing for cosmetic treatment applications. CONCLUSIONS: The goal of this study was to develop an automated hair counting system for clinical application using the microscope image from the hairs. The proposed method reduces the frequent errors and variances encountered when hairs are manually counted by human assessors. This clinical intelligent system can diagnose the health condition of a person's hair and can be applied in therapy recommendations by doctors for their patients.
BACKGROUND/ PURPOSE: One emerging subject in medical image processing is to quantitatively assess the health and the properties of cranial hairs, including density, diameter, length, level of oiliness, and others. This information helps hair specialists with making a more accurate diagnosis and the therapy required. We develop a practical hair counting algorithm. This analytic system calculates the number of hairs on a scalp using a digital microscope camera, providing accurate information for both the hair specialist and the patient. Our proposed hair counting algorithm is substantially more accurate than the Hough-based one, and is robust to curls, oily scalp, noise-corruption, and overlapping hairs, under various levels of illumination. Rather than manually counting the hairs on a person's scalp, the proposed system determines the density, diameter, length, and level of oiliness of the hairs. METHODS: We propose an automated system for counting the amount of hairs in the microscopy images. To reduce the effect of bright spots, we develop a robust morphological algorithm for color to smooth out the color and preserve the fidelity of the hair. Then, we utilize a modified Hough transform algorithm to detect the different hair lengths and to reduce any false detection due to noise. Our proposed system enables us to look at curved hairs as multiple pieces of straight lines. To avoid missing hairs when the thinning process is applied, we use edge information to discover any hidden or overlapping hairs. Finally, we employ a mutually associative regression method to label a group of line segments into a meaningful 'hair'. RESULTS: We demonstrated a novel approach for accurately computing the number of hairs, and successfully solved the three main obstacles in automated hair counting, including (i) oily and moist hairs, (ii) wavy and curly hairs, and (iii) under-estimation of the number of hairs occurs when hairs cross and occlude each other. The framework of this paper can be seen as the first step toward intelligent computer-aided medical image processing for cosmetic treatment applications. CONCLUSIONS: The goal of this study was to develop an automated hair counting system for clinical application using the microscope image from the hairs. The proposed method reduces the frequent errors and variances encountered when hairs are manually counted by human assessors. This clinical intelligent system can diagnose the health condition of a person's hair and can be applied in therapy recommendations by doctors for their patients.