视频人脸识别马赛克处理

2023-12-27 15:11:07


前言

前面几篇文章我们尝试了使用opencv完成图像人脸识别以及识别后贴图或者打马赛克的方法。
偶尔我们也会有需求在视频中将人脸马赛克化,opencv也提供了相应的方法来实现这个功能。


一、实现思路?

视频究其本质是图像按照一定的帧率去播放。如果需要将视频中的人脸马赛克化,那么我们可以逐帧输出图像后进行识别人脸再对其马赛克化,最终将所有的图像再按一定的帧率组合播放。

二、Coding

#识别视频人脸并增加马赛克
#实现原理:cv2读取视频后逐帧识别人脸并增加马赛克/贴图,处理完毕后保存视频

import cv2
# laod opencv schema
classifier = cv2.CascadeClassifier("haarcascade_frontalface_alt2.xml")#实践下来貌似这个schema匹配度最高
blockimg = "block.jpg"#贴图路径

#马赛克化
def do_mosaic(frame, x, y, w, h, neighbor=20):
    fh, fw = frame.shape[0], frame.shape[1]
    if (y + h > fh) or (x + w > fw):
        return
    for i in range(0, h - neighbor, neighbor):  # 关键点0 减去neightbour 防止溢出
        for j in range(0, w - neighbor, neighbor):
            rect = [j + x, i + y, neighbor, neighbor]
            color = frame[i + y][j + x].tolist()  # 关键点1 tolist
            left_up = (rect[0], rect[1])
            right_down = (rect[0] + neighbor - 1, rect[1] + neighbor - 1)  # 关键点2 减去一个像素
            cv2.rectangle(frame, left_up, right_down, color, -1)

#贴图处理
def do_blockpic(frame, x, y, w, h):
    resizeimg = cv2.imread(blockimg)
    resizeimg_f = cv2.resize(resizeimg,(w,h))
    frame[y:y+h, x:x+w] = resizeimg_f

#识别人脸
def do_identifyFace(frame):
    color = (0, 255, 0)
    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)  # convert to grey
    # begin to identify face
    faceRects = classifier.detectMultiScale(gray, scaleFactor=1.3, minNeighbors=3, minSize=(32, 32))
    if len(faceRects):  # get faces if above zero
        for faceRect in faceRects:  # loop each face
            x, y, w, h = faceRect
            #do_blockpic(frame, x, y, w, h)
            do_mosaic(frame, x, y, w, h)

#main starts
srcVideo = "srcVideo.mp4"#源视频
savedVideo = "savedVideo.mp4"#处理后的视频
cap = cv2.VideoCapture(srcVideo)

if not cap.isOpened():
    print("error to open source video")
    exit()

print("got source video")
w = cap.get(cv2.CAP_PROP_FRAME_WIDTH)
h = cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
fps = cap.get(cv2.CAP_PROP_FPS)
fcount = cap.get(cv2.CAP_PROP_FRAME_COUNT)
print("total frames %s"%fcount)#获取所有帧数
writer = cv2.VideoWriter(savedVideo, cv2.VideoWriter_fourcc('m', 'p', '4', 'v'), int(fps), (int(w), int(h)))

print("start handle source video")
i = 0
while cap.isOpened():
    success, frame = cap.read()
    while success:
        do_identifyFace(frame)
        print("finish frame %s"%i)
        writer.write(frame)
        #cv2.imwrite("frame%s.jpg"%i, frame)
        i += 1

        success, frame = cap.read()
        if (cv2.waitKey(20) & 0xff) == ord('q'):
            break
    cap.release()

print("finish handle source video")
writer.release()
cv2.destroyAllWindows()

三、实现效果

处理后的视频效果

文章来源:https://blog.csdn.net/gxiangzi/article/details/135243944
本文来自互联网用户投稿,该文观点仅代表作者本人,不代表本站立场。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。