偶像刘德华，我爱你30年，我要和你用python换个脸

1 说明：

=====

1.1 首先：刘德华是我的偶像，一直很喜欢他，向他致敬，无意冒犯，仅供学习。

1.2 技术要点：python+OpenCV（cv2）+dlib+numpy，代码详细，简单通俗，小白秒懂。

1.3 图片来源：今日头条免费正版图库。

2 准备：

=====

2.1 参看文章来源：

#https://github.com/Mister5ive/changeFaceImg  #下载，提取shape_predictor_68_face_landmarks.dat
#上面github下载的代码多，弃用。
#https://blog.csdn.net/ugly_scarecrow/article/details/77449576  #代码来源，并对代码进行修改，注释和删减

2.2 环境：

python3.8+OpenCV4.2.0+dlib19.19+微软编辑器vscode+深度操作系统deepin-linux。

3 代码分析：

=========

3.1 第1步：导入模块

import cv2
import dlib
import sys
import numpy as np

3.2 第2步：参数设定

SCALE_FACTOR = 1 
FEATHER_AMOUNT = 11
# 代表各个区域的关键点标号
FACE_POINTS = list(range(17, 68))
MOUTH_POINTS = list(range(48, 61))
RIGHT_BROW_POINTS = list(range(17, 22))
LEFT_BROW_POINTS = list(range(22, 27))
RIGHT_EYE_POINTS = list(range(36, 42))
LEFT_EYE_POINTS = list(range(42, 48))
NOSE_POINTS = list(range(27, 35))
JAW_POINTS = list(range(0, 17))

# pupillary distance. 瞳孔距离
COLOUR_CORRECT_BLUR_FRAC = 0.6

3.3 第3步：加载模型，注意路径。模型需要在dlib包或者上面提到的github中获取。

PREDICTOR_PATH = "/home/xgj/Desktop/change-face/model/shape_predictor_68_face_landmarks.dat"
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(PREDICTOR_PATH)

3.4 第4步：函数定义

# 获取关键点坐标位置，只获取一张人脸
# input：代表一张图片的numpy array
# output：68*2的关键点坐标位置matrix
def get_landmarks(im):
    rects = detector(im, 1)
    return np.matrix([[p.x, p.y] for p in predictor(im, rects[0]).parts()])

def read_im_and_landmarks(fname):
    im = cv2.imread(fname, cv2.IMREAD_COLOR)
    im = cv2.resize(im, (im.shape[1] * SCALE_FACTOR, im.shape[0] * SCALE_FACTOR))
    s = get_landmarks(im)
    return im, s

def draw_convex_hull(im, points, color):
    points = cv2.convexHull(points)  # 检测凸包函数
    cv2.fillConvexPoly(im, points, color=color)  # 绘制好多边形后并填充     点的顺序不同绘制出来的凸包也不同


def get_face_mask(im, landmarks):
    im = np.zeros(im.shape[:2], dtype=np.float64)
    draw_convex_hull(im,landmarks,color=1)
    im = np.array([im, im, im]).transpose((1, 2, 0))  # 得到一个类似于3通道的图片

    return im
 
# 用普氏分析(Procrustes analysis)调整脸部
def transformation_from_points(points1, points2):
    points1 = points1.astype(np.float64)
    points2 = points2.astype(np.float64)
    c1 = np.mean(points1, axis=0)
    c2 = np.mean(points2, axis=0)
    points1 -= c1
    points2 -= c2
    # 计算标准差
    s1 = np.std(points1)
    s2 = np.std(points2)
    points1 /= s1
    points2 /= s2
    # 通过奇异值分解求得旋转矩阵R
    U, S, Vt = np.linalg.svd(points1.T * points2)
 
    R = (U * Vt).T   # 维度:2*2
    # 仿射变换矩阵3*3 #  numpy.hstack用来在第1个维度上拼接tup  numpy.vstack在第0个维度上拼接tup
    return np.vstack([np.hstack(((s2 / s1) * R,
                                       c2.T - (s2 / s1) * R * c1.T)),
                         np.matrix([0., 0., 1.])])
 
def warp_im(im, M, dshape):
    output_im = np.zeros(dshape, dtype=im.dtype)
    cv2.warpAffine(im,M[:2],(dshape[1], dshape[0]),dst=output_im,borderMode=cv2.BORDER_TRANSPARENT,flags=cv2.WARP_INVERSE_MAP)
    return output_im
 
# 颜色校正
def correct_colours(im1, im2, landmarks1):
    blur_amount = COLOUR_CORRECT_BLUR_FRAC * np.linalg.norm(np.mean(landmarks1[LEFT_EYE_POINTS], axis=0)-np.mean(landmarks1[RIGHT_EYE_POINTS], axis=0))
    blur_amount = int(blur_amount)
    if blur_amount % 2 == 0:
        blur_amount += 1
    im1_blur = cv2.GaussianBlur(im1, (blur_amount, blur_amount), 0)
    im2_blur = cv2.GaussianBlur(im2, (blur_amount, blur_amount), 0)
    # Avoid divide-by-zero errors.
    im2_blur += (128 * (im2_blur <= 1.0)).astype(im2_blur.dtype)
    return (im2.astype(np.float64) * im1_blur.astype(np.float64)/im2_blur.astype(np.float64))

3.5 第5步：加载图片和开始换脸

im1, landmarks1 = read_im_and_landmarks('/home/xgj/Desktop/change-face/me.jpeg')  #需要换脸：me的脸
im2, landmarks2 = read_im_and_landmarks('/home/xgj/Desktop/change-face/ldh.jpeg')  #偶像的脸：ldh的脸

#换脸点
M = transformation_from_points(landmarks1,landmarks2)
 
# get_face_mask()的定义是为一张图像和一个标记矩阵生成一个掩膜
mask = get_face_mask(im2, landmarks2)
warped_mask = warp_im(mask, M, im1.shape)
# 33. 用min函数取掩膜区域效果更好
combined_mask = np.min([get_face_mask(im1, landmarks1), warped_mask],axis=0) 
# 将图像2的掩膜转换到图像1的坐标空间
warped_im2 = warp_im(im2, M, im1.shape)
warped_corrected_im2 = correct_colours(im1, warped_im2, landmarks1)
output_im = im1 * (1.0 - combined_mask) + warped_corrected_im2 * combined_mask
#保存换脸后的图片
#cv2.imwrite('/home/xgj/Desktop/change-face/output.jpg', output_im)
output_im = output_im.astype(np.uint8)
#展示图片
cv2.imshow('outputface', output_im)

cv2.waitKey()

4 操作和效果：

华仔，爱你一万年。