Human Skin Detection with OpenCV: YCrCb and HSV Based Methods

The OpenCV library, released under a BSD license, provides cross‑platform computer‑vision functions for Linux, Windows, macOS, Android, and more, and can be accessed from Python, Ruby, MATLAB, etc. It is used here to implement a human skin‑color detection feature.

First, install the required Python packages from a fast mirror:

pip install opencv-python -i https://mirrors.ustc.edu.cn/pypi/web/simple

pip install numpy -i https://mirrors.ustc.edu.cn/pypi/web/simple

Basic image handling with OpenCV includes reading, displaying, and saving images:

import numpy as np
import cv2
imname = "6358772.jpg"  # path to the image
img = cv2.imread(imname, cv2.IMREAD_COLOR)
cv2.imshow("image", img)
cv2.waitKey(0)
cv2.destroyAllWindows()
cv2.imwrite(imname + "01.jpg", img)

Algorithm 1 – YCrCb Cr component + Otsu thresholding : Convert the image to YCrCb, extract the Cr channel, apply Gaussian blur, then use Otsu’s method to obtain a binary skin mask.

img = cv2.imread(imname, cv2.IMREAD_COLOR)
ycrcb = cv2.cvtColor(img, cv2.COLOR_BGR2YCrCb)
(y, cr, cb) = cv2.split(ycrcb)
cr1 = cv2.GaussianBlur(cr, (5, 5), 0)
_, skin1 = cv2.threshold(cr1, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
cv2.imshow("image CR", cr1)
cv2.imshow("Skin Cr+OSTU", skin1)

Algorithm 2 – YCrCb Cr and Cb range filtering : After converting to YCrCb, keep pixels whose Cr value lies between 140‑175 and Cb between 100‑120.

img = cv2.imread(imname, cv2.IMREAD_COLOR)
ycrcb = cv2.cvtColor(img, cv2.COLOR_BGR2YCrCb)
(y, cr, cb) = cv2.split(ycrcb)
skin2 = np.zeros(cr.shape, dtype=np.uint8)
(x, y) = cr.shape
for i in range(0, x):
    for j in range(0, y):
        if 140 < cr[i][j] < 175 and 100 < cb[i][j] < 120:
            skin2[i][j] = 255
        else:
            skin2[i][j] = 0
cv2.imshow(imname, img)
cv2.imshow(imname + " Skin2 Cr+Cb", skin2)

Algorithm 3 – HSV range filtering : Convert the image to HSV and keep pixels with H in 7‑20, S in 28‑255, and V in 50‑255.

img = cv2.imread(imname, cv2.IMREAD_COLOR)
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
(_h, _s, _v) = cv2.split(hsv)
skin3 = np.zeros(_h.shape, dtype=np.uint8)
(x, y) = _h.shape
for i in range(0, x):
    for j in range(0, y):
        if 7 < _h[i][j] < 20 and 28 < _s[i][j] < 255 and 50 < _v[i][j] < 255:
            skin3[i][j] = 255
        else:
            skin3[i][j] = 0
cv2.imshow(imname, img)
cv2.imshow(imname + " Skin3 HSV", skin3)

Running screenshots and comparative result images illustrate the effectiveness of each method. The first method (Cr+Otsu) produces a clean binary mask, the second (Cr/Cb range) offers a tighter skin region, and the third (HSV range) provides a balance between precision and recall.

The three detection results are further compared in a single figure, showing the trade‑offs among the approaches.

Reference: OpenCV documentation and various online tutorials on skin detection.