使用 Python 实现视频 Logo 消除处理

ridLogoManner_staticImg = 1ridLogoManner_frameImg = 2ridLogoManner_inpaint = 3ridLogoManner_multiSampleInpaint = 4

def OnMouseEvent( event, x, y, flags, param):    try:        mouseEvent = param        mouseEvent.processMouseEvent(event, x, y, flags)
    except Exception as e:        print("使用回调函数OnMouseEvent的方法错误，所有使用该回调函数处理鼠标事件的对象，必须满足如下条件:")        print("    1、必须将自身通过param传入")        print("    2、必须定义一个processMouseEvent(self)方法来处理鼠标事件")        print(e)

class CImgMouseEvent():    def __init__(self,parent,img=None,winName=None):        self.img = img        self.winName = winName        self.parent = parent        self.ignoreEvent = [cv2.EVENT_MBUTTONDOWN,cv2.EVENT_MBUTTONUP,cv2.EVENT_MBUTTONDBLCLK,cv2.EVENT_MOUSEWHEEL,cv2.EVENT_MOUSEHWHEEL] #需要忽略的鼠标事件        self.needRecordEvent = [cv2.EVENT_MOUSEMOVE,cv2.EVENT_LBUTTONDOWN,cv2.EVENT_LBUTTONUP] #需要记录当前信息的鼠标事件        self.windowCreated = False #窗口是否创建标记        if img is not None:self.showImg(img,winName)        self.open(winName)
    def open(self, winName=None):    #初始化窗口相关属性，一般情况下此时窗口还未创建，因此鼠标回调函数设置不会执行        if winName:            if self.winName != winName:                if self.winName:                    cv2.destroyWindow(self.winName)                    self.windowCreated = False                self.WinName = winName
        self.mouseIsPressed = self.playPaused = False        self.previousePos = self.pos = self.previousEvent = self.event = self.flags = self.previouseFlags = None        if self.winName and self.windowCreated : cv2.setMouseCallback(self.winName, OnMouseEvent, self)
    def showImg(self,img,winName=None):        """        在窗口winName中显示img图像，并设置鼠标回调函数为OnMouseEvent        """        if not winName:winName = self.winName        self.img = img        if winName != self.winName:            self.winName = winName            self.open()
        if not self.windowCreated:            self.windowCreated = True            cv2.namedWindow(winName)#cv2.WINDOW_NORMAL| cv2.WINDOW_KEEPRATIO | cv2.WINDOW_GUI_EXPANDED            cv2.setMouseCallback(winName, OnMouseEvent, self)        cv2.imshow(winName, img)
    def processMouseEvent(self,event, x, y, flags):        #鼠标回调函数调用该函数处理鼠标事件，包括记录当前事件信息、判断是否记录上次鼠标事件信息、是否暂停视频播放，调用parent.processMouseEvent() 执行响应操作        #mouseventDict = {cv2.EVENT_MOUSEMOVE:"鼠标移动中",cv2.EVENT_LBUTTONDOWN:"鼠标左键按下",cv2.EVENT_RBUTTONDOWN:"鼠标右键按下",cv2.EVENT_MBUTTONDOWN:"鼠标中键按下",cv2.EVENT_LBUTTONUP:"鼠标左键释放",cv2.EVENT_RBUTTONUP:"鼠标右键释放",cv2.EVENT_MBUTTONUP:"鼠标中键释放",cv2.EVENT_LBUTTONDBLCLK:"鼠标左键双击",cv2.EVENT_RBUTTONDBLCLK:"鼠标右键双击",cv2.EVENT_MBUTTONDBLCLK:"鼠标中键双击",cv2.EVENT_MOUSEWHEEL:"鼠标轮上下滚动",cv2.EVENT_MOUSEHWHEEL:"鼠标轮左右滚动"}        #print(f"processMouseEvent {mouseventDict[event]} ")
        if event in self.ignoreEvent:return        if self.event in [cv2.EVENT_LBUTTONDOWN,cv2.EVENT_LBUTTONUP]:#当上次鼠标事件左键按下或释放时，上次信息保存            self.previousEvent,self.previousePos,self.previouseFlags  = self.event,self.pos,self.flags
        if  event==cv2.EVENT_LBUTTONUP:            self.mouseIsPressed = False        elif event == cv2.EVENT_LBUTTONDOWN:            self.mouseIsPressed = True            self.playPaused = True        elif  event in [cv2.EVENT_LBUTTONDBLCLK,cv2.EVENT_RBUTTONDBLCLK,cv2.EVENT_RBUTTONDOWN,cv2.EVENT_RBUTTONUP]:#鼠标右键动作、鼠标双击动作恢复视频播放            self.playPaused = False         if event in self.needRecordEvent:            self.event,self.flags,self.pos = event,flags,(x,y)
        self.parent.processMouseEvent()  #调用者对象的鼠标处理方法执行
    def getMouseSelectRange(self):        """        获取鼠标左键按下位置到当前鼠标移动位置或左键释放位置的对应的矩形以及矩形最后位置的鼠标事件类型        :return: 由鼠标左键按下开始到鼠标左键释放或鼠标当前移动位置的矩形，为None表示当前没有这样的操作        """        if self.previousEvent is None or self.event is None:            return None        if (self.event!=cv2.EVENT_LBUTTONUP)  and (self.event!=cv2.EVENT_MOUSEMOVE): #最近的事件不是鼠标左键释放或鼠标移动            return None        if self.pos == self.previousePos:#与上次比位置没有变化            return None        if (self.previousEvent== cv2.EVENT_LBUTTONDOWN ) and (self.event==cv2.EVENT_LBUTTONUP): #鼠标左键按下位置到鼠标左键释放位置            return [self.previousePos,self.pos,cv2.EVENT_LBUTTONUP]        elif (self.previousEvent== cv2.EVENT_LBUTTONDOWN ) and (self.event==cv2.EVENT_MOUSEMOVE):#鼠标左键按下位置到鼠标当前移动位置            return [self.previousePos, self.pos, cv2.EVENT_MOUSEMOVE]        return None
    def drawRect(self,color,specRect=None,filled=False):        """        :param color: 矩形颜色        :param specRect: 不为None画specRect指定矩形，否则根据鼠标操作来判断        :param filled: 是画实心还是空心矩形，缺省为空心矩形        :return: 画下的矩形，specRect不为None时是specRect指定矩形，否则根据鼠标操作来判断        """        if specRect:            rect = specRect        else:            rect = self.getMouseSelectRange()
        if rect:            img = self.img            img = self.img.copy()            if not filled:                cv2.rectangle(img, rect[0], rect[1], color,1)            else:                cv2.rectangle(img, rect[0], rect[1], color,-1)            cv2.imshow(self.winName, img)            return rect        else:            return None
    def drawEllipse(self, color,specRect=None, filled=False):        """        :param color: 椭圆颜色        :param specRect: 不为None画specRect指定椭圆，否则根据鼠标操作来判断        :param filled: 是画实心还是空心椭圆，缺省为空心椭圆        :return: 画下的椭圆对应的外接矩形，specRect不为None时是specRect指定矩形，否则根据鼠标操作来判断        """        if specRect:            rect = specRect        else:            rect = self.getMouseSelectRange()        if rect:            x0, y0 = rect[0]            x1, y1 = rect[1]            x = int((x0+x1)/2)            y = int((y0+y1)/2)            axes = (int(abs(x1-x0)/2),int(abs(y1-y0)/2))            img = self.img.copy()            if not filled:                cv2.ellipse(img, (x, y),axes, 0,0,360,  color,1)            else:                cv2.ellipse(img, (x, y),axes, 0,0,360,  color,-1)            cv2.imshow(self.winName, img)            return rect        else:            return None
    def close(self):        cv2.destroyWindow(self.winName)        self.windowCreated = False
    def __del__(self):        self.close()

class CSubVideoImg():    def __init__(self,videoFName):        super().__init__()        self.imgMouseEvent = CImgMouseEvent(self)     #创建鼠标事件处理对象        self.videoFName = videoFName        self.exitKeys = [ord('q') ,ord('q'), 27] #视频图像播放时退出键定义，包括q、Q以及ESC        self.initStatus()
    def initStatus(self):#初始化相关变量，self.rect为记录最后一个鼠标选择框        self.rect = self.logoObjList = self.replaceObject = self.frameMask = None

    def processMouseEvent(self):#鼠标事件响应函数，将当前选择框显示出来        self.drawSelectRange() 
    def drawSelectRange(self,specRect=None):        if specRect:            rect = self.imgMouseEvent.drawRect((255, 0, 0),specRect)        else:            rect = self.imgMouseEvent.drawRect((255, 0, 0))            if rect: self.rect = rect
    def displayImg(self,winname,img,seconds):        cv2.imshow(winname, img)        ch = cv2.waitKey(seconds*1000)        cv2.destroyWindow(winname)
    def getROI(self, operInfo, fps=24):        """       获取视频中的多个ROI区域(即鼠标选择区域)，选定一个ROI区域后，按N、n、S、s保存当前选择区域       按指定帧率播放视频（仅图像），并提供在视频图像中选中某个矩形范围，并在接下来播放中一直显示该矩形，按EsC或q或Q退出       退出后会显示当前选择的ROI图像       :param operInfo: 播放窗口提示信息，也即窗口名，必须是英文       :param fps: 播放的帧率       :return: 返回选择的ROI及对应帧的二元组，类似：（[(rect1,img1),...,(rectn,imgn)],frame)矩形和最后选中操作所在帧的选择图像        """        frame = None        cap = cv2.VideoCapture(self.videoFName)        self.imgMouseEvent.open(operInfo)
        ROIList = []        saveKeys = [ord('n'), ord('N'), ord('s'), ord('S')]+self.exitKeys #保存和退出键都保存最后一个选择矩阵范围
        self.rect = None        if not cap.isOpened():            print("Cannot open video")            return None
        while True:            if not self.imgMouseEvent.playPaused: #正在播放                ret, frame = cap.read()                if not ret:                    if frame is None:                        print("The video has end.")                    else:                        print("Read video error!")                    break                self.imgMouseEvent.showImg(frame, operInfo)                self.drawSelectRange(self.rect)
            ch = cv2.waitKey(int(1000 / fps))            if ch in saveKeys:                if self.rect is not None:                    x0, y0 = self.rect[0]                    x1, y1 = self.rect[1]                    ROI = frame[y0:y1, x0:x1]                    ROIList.append((ROI, self.rect))                    self.rect = None            if ch in self.exitKeys: break        # 完成所有操作后，释放捕获器        if len(ROIList) == 0:            self.imgMouseEvent.close()            cap.release()            return None,None
        self.imgMouseEvent.close()        cap.release()
        return ROIList, frame
    def replaceImgRegionBySpecImg(self,srcImg,regionTopLeftPos,specImg):        """        将srcImg的regionTopLeftPos开始位置的一个矩形图像替换为specImg        :return: True 成功，False失败        """        srcW, srcH = srcImg.shape[1::-1]        refW, refH = specImg.shape[1::-1]        x,y =  regionTopLeftPos        if (refW>srcW) or (refH>srcH):            #raise ValueError("specImg's size must less than srcImg")            print(f"specImg's size {specImg.shape[1::-1]} must less than srcImg's size {srcImg.shape[1::-1]}")            return False        else:            srcImg[y:y+refH,x:x+refW] = specImg            return True
    def replaceImgRegionBySpecRange(self,srcImg,regionTopLeftPos,specRect):        """        将srcImg的regionTopLeftPos开始位置的一个矩形图像替换为srcImg内specRect指定的一个矩形范围图像        :return: True 成功，False失败        """        srcW, srcH = srcImg.shape[1::-1]        refW, refH = specRect[1][0]-specRect[0][0],specRect[1][1]-specRect[0][1]        x,y =  regionTopLeftPos        if (refW>srcW) or (refH>srcH):            print(f"specImg's size {(refW, refH)} must less than srcImg's size {srcImg.shape[1::-1]}")            return False        else:            srcImg[y:y+refH,x:x+refW] = srcImg[specRect[0][1]:specRect[1][1],specRect[0][0]:specRect[1][0]]            return True
    def adjuestImgAccordingRefImg(self,img,refimg,color=None):        """        按照refimg大小调整img大小，如果是扩充，则采用img边缘像素的镜像复制或指定颜色创建扩充像素        :param img:        :param refimg:        :param color:        :return:        """        srcW,srcH = img.shape[1::-1]        refW,refH = refimg.shape[1::-1]        if srcW>refW:            diff = int((srcW-refW)/2)            img = img[:,diff:refW+diff]        if srcH>refH:            diff = int((srcH - refH) / 2)            img =img[diff:refH+diff,:]        srcW, srcH = img.shape[1::-1]        w = max(srcW,refW)        h = max(srcH,refH)        diffW = int((w-srcW)/2)        diffH = int((h-srcH)/2)        if color is None:            dest = cv2.copyMakeBorder(img,diffH,h-srcH-diffH,diffW,w-srcW-diffW,cv2.BORDER_REFLECT_101) #上下左右扩展当前图像        else:            dest = cv2.copyMakeBorder(img, diffH,h-srcH-diffH,diffW,w-srcW-diffW, cv2.BORDER_CONSTANT,color)#上下左右扩展当前图像，扩展部分颜色为color        rectSize = (h,w)        return dest
    def createImgMask(self, img):        # 创建img的掩码        img2gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)        ret, mask = cv2.threshold(img2gray, 35, 255, cv2.THRESH_BINARY) #转为像素值为0和255的二值图，阈值为35
        #对掩码进行膨胀处理        element = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))        mask = cv2.dilate(mask, element)
        return mask
    def genLogoFrameMask(self,frame,logoObject):    #将Logo掩码填充到一与视频帧大小相同的全0图像中        logoImg,logoRect = logoObject        if logoImg is  None:            return None        else:            logMask = self.createImgMask(logoImg)            frameMask = np.zeros(frame.shape[0]*frame.shape[1],dtype=np.uint8)            frameMask = frameMask.reshape(frame.shape[0:2])            x0,y0 = logoRect[0]            x1,y1 = logoRect[1]            frameMask[y0:y1,x0:x1] = logMask            return frameMask
    def genMultiLogoFrameMask(self, logoObjectList,frame ):    #将多次采样的Logo掩码填充到一与视频帧大小相同的全0图像中        composeFrameMask = None        for logoObject in logoObjectList:            frameMask = self.genLogoFrameMask(frame, logoObject)            if composeFrameMask is None:                composeFrameMask = frameMask            else:                composeFrameMask = cv2.add(composeFrameMask, frameMask)
        return composeFrameMask
    def convertVideo(self,outPutFName,ridLogoManner,logoObjects,replaceObject=None,frameMask=None):        #生成视频        global videoImgConvertParams
        if ridLogoManner in [ridLogoManner_staticImg,ridLogoManner_frameImg]:            if replaceObject is None:                return False,"替换图像尚未提供，请先选择替换图像"        else:            if frameMask is None:                return False,"替换frameMask尚未提供或未生成,请确保进行了Logo图像的截取操作，请先提供"
        self.frameMask = frameMask        self.replaceObject = replaceObject        self.logoObjList  = logoObjects        self.ridLogoManner = ridLogoManner
        try:            videoImgConvertParams = self, ridLogoManner            clipVideo = VideoFileClip(self.videoFName)            newclip = clipVideo.fl_image(processImg)            newclip.write_videofile(outPutFName, threads=8)            clipVideo.close()            newclip.close()        except Exception as e:            return False,f"生成视频时出现异常:\n{e}"        else:            return True,f"视频处理完成，生成的视频保存着在文件:{outPutFName}"
    def previewVideoByReplaceLogo(self,fps,logoObjects,replaceObject,ridLogoManner):        """        使用替换区域或替换图像替换logo区域后的视频效果预览        fps:fps        用于使用静态图像或同帧图像替换后预览视频使用        :param logoObjects:         二元组：(logoObjectList,FRAME)，实际形如（[(logoImg1,logoRect1),...,(logoImgn,logoRectn)],FRAME)         logoObjectList:列表，1...n个元素（只有当采用多次采样修复算法时才会n大于1），每个元素是个二元组，每个二元组表示一个logo图像信息，包括图像的数组以及图像的位置及大小等信息，         形如：[(logoImg1,logoRect1),...,(logoImgn,logoRectn)]         Frame：截取Logon图像的帧对应数组，当预览一个帧时可以使用
        :param replaceObject:四元组(replaceImg, replaceRect,targetReplaceImg frame)
        :param ridLogoManner:消除logo的方式
        :return:        """        global  videoImgConvertParams        videoImgConvertParams = self,ridLogoManner
        self.frameMask = None        self.replaceObject = replaceObject        self.logoObjList = logoObjects        self.ridLogoManner = ridLogoManner
        cap = cv2.VideoCapture(self.videoFName)
        if not cap.isOpened():            print("Cannot open video")            return        winName = f"video previewing fps={fps}"        while True:            ret, frame = cap.read()            if not ret:                if frame is None:                    print("The video has end.")                else:                    print("Read video error!")                break
            frame = processImg(frame)
            cv2.imshow(winName, frame)
            ch = cv2.waitKey(int(1000 / fps))            if ch in self.exitKeys:  break        # 完成所有操作后，释放捕获器
        cap.release()        cv2.destroyWindow(winName)
    def previewVideoByInpaintLogo(self,fps, logoObjects,frameMask, ridLogoManner):        """        使用图像修复术对logo区域处理后的视频效果预览        fps:fps        :param logoObjects:列表，1...n个元素（当多次采样Logo时n大于1），每个元素是个二元组，每个二元组表示一个logo图像信息，包括图像的数组以及图像的位置及大小等信息，        形如：[(logoImg1,logoRect1),...,(logoImgn,logoRectn)]        Frame：截取Logon图像的帧对应数组，当预览一个帧时可以使用
        :param ridLogoManner:消除logo的方式              """        global      videoImgConvertParams
        if ridLogoManner not in [ridLogoManner_inpaint, ridLogoManner_multiSampleInpaint]:            print("ridLogoManner is not fit previewVideoByInpaintLogo ")            return False

        videoImgConvertParams = self, ridLogoManner
        self.frameMask = None        self.replaceObject = None        self.logoObjList = logoObjects        self.ridLogoManner = ridLogoManner
        winName = f"video previewing,fps={fps}"
        self.frameMask = frameMask        self.multiFrameMask = frameMask        cap = cv2.VideoCapture(self.videoFName)
        if not cap.isOpened():            print("Cannot open video")            return
        while True:            ret, frame = cap.read()            if not ret:                if frame is None:                    print("The video has end.")                else:                    print("Read video error!")                break            frame = processImg(frame)            cv2.imshow(winName, frame)

            ch = cv2.waitKey(int(1000 / fps))
            if ch in self.exitKeys:  break        # 完成所有操作后，释放捕获器
        cap.release()        cv2.destroyWindow(winName)

def main(ridLogoManner):    videoOperation = CSubVideoImg(r"f:\video\mydream.mp4")    destFName = r"f:\video\mydream_new_"+str(ridLogoManner)+".mp4"    fps = 24    replaceObject = logoObjList = multiFrameMask = frameMask = None

    print("请在播放的视频中选择要去除Logo的区域：")    logobjs, frame = videoOperation.getROI("select multiLogo Imgs Range", fps)    if logobjs is not None and len(logobjs):        logoObjList = (logobjs, frame)        frameMask = videoOperation.genMultiLogoFrameMask([logobjs[-1]], frame)        multiFrameMask = videoOperation.genMultiLogoFrameMask(logobjs, frame)        frame = frame    else:        print("本次操作没有选择对应Logo图像，程序退出。")        return
    if ridLogoManner in ( ridLogoManner_staticImg, ridLogoManner_frameImg):  # ridLogoManner_inpaint , ridLogoManner_multiSampleInpaint        print("请在播放的视频中选择要去除Logo的区域：")        replaceObjList, frame = videoOperation.getROI("select Replace Img Range")        if replaceObjList is None:            replaceObject = None            print("本次操作没有选择对应替换区域或替换图像，如果要执行后续操作，请重新选择。")        else:            replaceImg, replaceRect = replaceObjList[-1]            if replaceRect is not None:                targetReplaceImg = videoOperation.adjuestImgAccordingRefImg(replaceImg, logoObjList[0][-1][0])                replaceObject = (replaceImg, replaceRect, targetReplaceImg, frame)
            else:                print("本次操作没有选择对应替换图像，程序退出。")                return    print("准备工作完成，开始进行视频转换：")    if ridLogoManner in [ridLogoManner_staticImg, ridLogoManner_frameImg]:        ret, inf = videoOperation.convertVideo(destFName, ridLogoManner, logoObjList, replaceObject)
    elif ridLogoManner == ridLogoManner_inpaint:        ret, inf = videoOperation.convertVideo(destFName, ridLogoManner, logoObjList, frameMask=frameMask)    else:        ret, inf = videoOperation.convertVideo(destFName, ridLogoManner, logoObjList, frameMask=multiFrameMask)    print(inf)
if __name__=='__main__':    main(ridLogoManner_multiSampleInpaint)