opencv+face_recognition+python实现换脸（face swap）操作（3）——基于普氏分析法（Procrustes Analysis）的代码实现

#获得特征点
def ladmask(img):
    faces_loaction=face_recognition.face_locations(img,number_of_times_to_upsample = 0,model ='cnn')
    face_feature=face_recognition.face_landmarks(img,face_locations=faces_loaction)
    face_feature1=pd.DataFrame(face_feature)
    def listexpend(X):
        a=[]
        for i in list(X):
            a.extend(i)      
        return a
    h1=[]
    for j in range (0,9):
        h=listexpend(face_feature1.iloc[:,j])
        h1+=h
    points=[]
    def all_tuplelist(X):
        a=[]
        for i in list(X):
            a.append(list(i))
        return a
    points=np.array(all_tuplelist(h1))
    return points
# 普氏分析法映射变换
def transformation_from_points(points1, points2):
    '''0 - 先确定是float数据类型 '''
    points1 = points1.astype(numpy.float64)
    points2 = points2.astype(numpy.float64)

    '''1 - 消除平移的影响 '''
    c1 = numpy.mean(points1, axis=0)
    c2 = numpy.mean(points2, axis=0)
    points1 -= c1
    points2 -= c2

    '''2 - 消除缩放的影响 '''
    s1 = numpy.std(points1)
    s2 = numpy.std(points2)
    points1 /= s1
    points2 /= s2

    '''3 - 计算矩阵M=BA^T；对矩阵M进行SVD分解；计算得到R '''
    # ||RA-B||; M=BA^T
    A = points1.T # 2xN
    B = points2.T # 2xN
    M = np.dot(B, A.T)
    U, S, Vt = numpy.linalg.svd(M)
    R = np.dot(U, Vt)

    '''4 - 构建仿射变换矩阵 '''
    s = s2/s1
    sR = s*R
    c1 = c1.reshape(2,1)
    c2 = c2.reshape(2,1)
    T = c2 - np.dot(sR,c1) # 模板人脸的中心位置减去 需要对齐的中心位置（经过旋转和缩放之后）

    trans_mat = numpy.hstack([sR,T])   # 2x3

    return trans_mat


def get_affine_image(image1, image2, face_landmarks1, face_landmarks2,M):
    """
    获取图片1仿射变换后的图片
    :param image1: 图片1, 要进行仿射变换的图片
    :param image2: 图片2, 只要用来获取图片大小，生成与之大小相同的仿射变换图片
    :param face_landmarks1: 图片1的人脸特征点
    :param face_landmarks2: 图片2的人脸特征点
    :return: 仿射变换后的图片
    """
    three_points_index = [18, 8, 25]
    M = M
 #   cv2.getAffineTransform(face_landmarks1[three_points_index].astype(np.float32),face_landmarks2[three_points_index].astype(np.float32))
    dsize = (image2.shape[1], image2.shape[0])
    affine_image = cv2.warpAffine(image1, M, dsize)
    return affine_image.astype(np.uint8)

def get_face_mask(img,face_landmarks):
    """
    获取人脸掩模
    :param image_size: 图片大小
    :param face_landmarks: 68个特征点
    :return: image_mask, 掩模图片
    """
    img_gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
    image_size = (img.shape[0], img.shape[1])
    mask = np.zeros_like(img_gray)
#    convexhull = cv2.convexHull(points)
#    cv2.polylines(img,[convexhull],True,(255,0,0), 3)
#    cv2.fillConvexPoly(mask,convexhull,255)
    mask = np.zeros(image_size, dtype=np.uint8)
    points = np.concatenate([face_landmarks[0:16], face_landmarks[26:17:-1]])
#    points = np.concatenate([face_landmarks])

    cv2.fillPoly(img=mask ,pts=[points],color=(255,255,255))

    # mask = np.zeros(image_size, dtype=np.uint8)
    # points = cv2.convexHull(face_landmarks)  # 凸包
    # cv2.fillConvexPoly(mask, points, color=255)
    return mask
def get_mask_union(mask1, mask2):
    """
    获取两个掩模掩盖部分的并集
    :param mask1: mask_image, 掩模1
    :param mask2: mask_image, 掩模2
    :return: 两个掩模掩盖部分的并集
    """
    mask = np.min([mask1, mask2], axis=0)  # 掩盖部分并集
    mask = ((cv2.blur(mask, (3, 3)) == 255) * 255).astype(np.uint8)  # 缩小掩模大小
    mask = cv2.blur(mask, (5, 5))#.astype(np.uint8)  # 模糊掩模
    return mask
def get_mask_center_point(image_mask):
    """
    获取掩模的中心点坐标
    :param image_mask: 掩模图片
    :return: 掩模中心
    """
    image_mask_index = np.argwhere(image_mask > 0)
    miny, minx = np.min(image_mask_index, axis=0)
    maxy, maxx = np.max(image_mask_index, axis=0)
    center_point = ((maxx + minx) // 2, (maxy + miny) // 2)
    return center_point

import numpy
import cv2
import numpy as np
import pandas as pd
import os
from matplotlib import pyplot as plt
import face_recognition
img1=cv2.imread("lif.jpg")#原图像
img2=cv2.imread("222.jpg")#目标图像
points1=ladmask(img1)
points2=ladmask(img2)
img1_mask=get_face_mask(img1,points1)
img2_mask=get_face_mask(img2,points2)
landmarks1=ladmask(img1)
landmarks2=ladmask(img2)
M = transformation_from_points(landmarks1,landmarks2)
affine_im1 = get_affine_image(img1, img2, landmarks1, landmarks2,M)
affine_im1_mask = get_affine_image(img1_mask, img2, landmarks1, landmarks2,M)

union_mask = get_mask_union(img2_mask, affine_im1_mask)  # 掩模合并
point = get_mask_center_point(union_mask)  # im1（脸图）仿射变换后的图片的人脸掩模的中心点 cv2.MIXED_CLONE cv2.MONOCHROME_TRANSFER
seamless_im = cv2.seamlessClone(np.uint8(affine_im1), img2, mask=union_mask, p=point, flags=cv2.NORMAL_CLONE)  # 进行泊松融合
plt.imshow(cv2.cvtColor(seamless_im, cv2.COLOR_BGR2RGB))
plt.xticks([]),plt.yticks([])  #to hide tick values on X and Y axis
plt.show()