python实现滚动截屏功能

本文对应GitHub项目地址

之前写了像微信一样的区域截屏功能,中提到,我的小工具集里面有滚动截屏的功能，这篇文章将讨论一下具体的实现和公布源码。。。

实现思路类似某乎上一个关于滚动截屏实现的回答。

不同于移动端的滚动截屏，pc端浏览器之外是不能获得准确的鼠标滚动距离的(或者有我不知道的方法)，要实现滚动截屏就不是那么容易了；

现在网上也有一些可以滚动截屏的软件(像FSCapture、qq截屏)，实现方法应该也是先截屏然后根据图片元素拼接的。本文的方法虽有拼接出错的情况，但是能用哇。。h，源码全长四百多行。。。由于是直接从我的小工具集里面整合出来的，有些地方会显得不太合理，当作学习就好，具体思路都差不多

源码

import math
import operator
import os
import sys
import time
from functools import reducefrom PIL import Image
from PyQt5.QtCore import QThread, QTimer
from PyQt5.QtWidgets import QApplication
from pynput import mouse
from pynput.mouse import Controller as MouseControllerclass Splicing_shots(object):def __init__(self):self.init_splicing_shots_thread = Commen_Thread(self.init_splicing_shots)self.init_splicing_shots_thread.start()self.clear_timer = QTimer()self.clear_timer.timeout.connect(self.setup)def init_splicing_shots(self):"""后台初始化"""self.img = []self.img_list = []self.images_data_line_list = []self.img_width = 0self.img_height = 0self.compare_row = 50self.cut_width = 0self.rate = 0.85self.roll_speed = 5self.min_head = 0self.left_border = 0self.right_border = 0self.head_pos = {}self.maybe_errorlist = []self.in_rolling = Falseself.arrange = 0self.max_arrange = 999if not os.path.exists("j_temp"):os.mkdir("j_temp")def setup(self):"""清理&初始化"""if self.clear_timer.isActive():self.clear_timer.stop()print('clear')self.img = []self.img_list = []self.images_data_line_list = []self.img_width = 0self.img_height = 0self.compare_row = 50self.cut_width = 0self.rate = 0.85self.roll_speed = 3self.min_head = 0self.left_border = 0self.right_border = 0self.head_pos = {}self.maybe_errorlist = []self.in_rolling = Falseself.arrange = 0self.max_arrange = 999self.clear_timer = QTimer()self.clear_timer.timeout.connect(self.setup)def find_left_side(self):"""寻找相同的左边界"""images_data_list = []for img in self.img_list[:2]:rotate_img = img.rotate(270, expand=1)rotate_img_data = list(rotate_img.convert('L').getdata())an_imgdata = []for line in range(rotate_img.height - 1):an_imgdata.append(rotate_img_data[line * rotate_img.width:(line + 1) * rotate_img.width])images_data_list.append(an_imgdata)rotate_height = len(images_data_list[0])min_head = rotate_heightfor i in range(1):for j in range(1, rotate_height):img1 = images_data_list[i][:j]img2 = images_data_list[i + 1][:j]if img2 != img1:if j == 1:print('没有重复左边界！')returnelif j < (min_head + 1):min_head = j - 1breakself.left_border = min_headprint('minleft', min_head)def find_right_size(self):"""寻找相同的右边界"""images_data_list = []for img in self.img_list[:2]:rotate_img = img.rotate(90, expand=1)rotate_img_data = list(rotate_img.convert('L').getdata())an_imgdata = []for line in range(rotate_img.height - 1):an_imgdata.append(rotate_img_data[line * rotate_img.width:(line + 1) * rotate_img.width])images_data_list.append(an_imgdata)rotate_height = len(images_data_list[0])min_head = rotate_heightfor i in range(1):for j in range(1, rotate_height):img1 = images_data_list[i][:j]img2 = images_data_list[i + 1][:j]# print(img2)if img2 != img1:if j == 1:print('没有重复右边界！')self.right_border = self.img_width# print(self.majority_color(self.images_data_line_list[0]))returnelif j < (min_head + 1):min_head = j - 1breakself.right_border = self.img_width - min_headprint('minright', min_head)def find_the_same_head_to_remove(self):"""寻找相同的头部(上边界)"""# if self.images_datamin_head = self.img_heightfor i in range(len(self.img_list) - 1):for j in range(1, self.img_height):img1 = self.images_data_line_list[i][:j]img2 = self.images_data_line_list[i + 1][:j]# print(img2)if img2 != img1:if j == 1:print('没有重复头！')# print(self.majority_color(self.images_data_line_list[0]))returnelif j < (min_head + 1):min_head = j - 1breakself.min_head = min_headprint('minhead', min_head)def majority_color(self, classList):'''返回颜色列表中最多的颜色'''count_dict = {}for label in classList:if label not in count_dict.keys():count_dict[label] = 0count_dict[label] += 1# print(max(zip(count_dict.values(), count_dict.keys())))return max(zip(count_dict.values(), count_dict.keys()))def isthesameline(self, line1, line2):"""判断是否两行是否相同"""same = 0rate = self.rateline1_majority_color = self.majority_color(line1)line2_majority_color = self.majority_color(line2)if line2_majority_color[1] != line1_majority_color[1]:# print(self.majority_color(line2),self.majority_color(line1))return 0elif abs(line1_majority_color[0] - line2_majority_color[0]) > self.img_width * (1 - rate) * 0.5:return 0else:majority_color_count, majority_color = line2_majority_color# print(majority_color_count,majority_color)if majority_color_count > int(self.cut_width * rate):return 1for i in range(self.cut_width):if line1[i] == majority_color or line2[i] == majority_color:# print('maj')continueelse:if abs(line1[i] - line2[i]) < 10:same += 1if same >= (self.cut_width - majority_color_count) * rate:return 1else:return 0def efind_the_pos(self):"""在滚动的同时后台寻找拼接点"""while self.in_rolling or self.arrange < self.max_arrange:  # 如果正在截屏或截屏没有处理完print(self.arrange, '  max:', self.max_arrange)min_head = self.min_headleft = self.left_borderright = self.right_borderself.cut_width = right - leftimages_data_line_list = self.images_data_line_listcompare_row = self.compare_rowi = self.arrangetry:img1 = images_data_line_list[i]  # 前一张图片img2 = images_data_line_list[i + 1]  # 后一张图片except IndexError:time.sleep(0.1)  # 图片索引超出则等待下一张截屏continuemax_line = [0, 0]for k in range(min_head, self.img_height - compare_row):  # 前一张图片从相同头部开始遍历到最后倒数compare_row行if self.in_rolling:  # 如果正在截屏则sleep一下避免过多占用主线程,也没什么用...time.sleep(0.001)sameline = 0chance_count = 0chance = 0for j in range(min_head, min_head + compare_row):  # 后一张图片从相同头部开始逐行遍历compare_row行lin1 = img1[k + sameline][left:right]lin2 = img2[min_head + sameline][left:right]if self.isthesameline(lin1, lin2):  # 如果是行相同,则sameline+1sameline += 1chance_count += 1if chance_count >= 7:  # 每7行增加一个chance,避免误判chance_count = 0chance += 1if sameline > max_line[1]:max_line[0] = kmax_line[1] = sameline  # 记录最大行数备用else:  # 否则chance-1直到退出if chance <= 0:breakelse:chance -= 1sameline += 1if sameline >= compare_row - compare_row // 20:self.head_pos[i] = kprint(i, k)print(self.head_pos)breakif i not in self.head_pos.keys():#如果没有找到符合的拼接点,则取最大的配合点,并标记为可能出错的地方if max_line[1] >= 1:self.head_pos[i] = max_line[0]print(self.head_pos)max_line.append(i)self.maybe_errorlist.append(max_line)print('max_line', i, max_line)  # 测试self.arrange += 1def find_the_pos(self):#和上面的efind_the_pos类似"""寻找拼接点,当图片数少时可以直接截完屏调用"""min_head = self.min_headleft = self.left_borderright = self.right_borderself.cut_width = right - leftimages_data_line_list = self.images_data_line_listcompare_row = self.compare_row# print(min_head, self.img_height - compare_row)for i in range(len(self.img_list) - 1):# print(i)img1 = images_data_line_list[i]img2 = images_data_line_list[i + 1]max_line = [0, 0]  # 测试for k in range(min_head, self.img_height - compare_row):sameline = 0chance_count = 0chance = 0for j in range(min_head, min_head + compare_row):lin1 = img1[k + sameline][left:right]lin2 = img2[min_head + sameline][left:right]# print(len(lin2),len(lin1))res = self.isthesameline(lin1, lin2)if res:sameline += 1chance_count += 1if chance_count >= 5:chance_count = 0chance += 1if sameline > max_line[1]:max_line[0] = kmax_line[1] = sameline  # 测试# print(i, j, k)else:# print(chance)if chance <= 0:breakelse:chance -= 1sameline += 1if sameline >= compare_row - compare_row // 20:self.head_pos[i] = kprint(i, k)print(self.head_pos)breakif i not in self.head_pos.keys():if max_line[1] >= 1:self.head_pos[i] = max_line[0]print(self.head_pos)max_line.append(i)self.maybe_errorlist.append(max_line)print('max_line', i, max_line)  # 测试def merge_all(self):"""根据拼接点拼接所有图片"""majority_pos = self.majority_color(self.head_pos.values())for i in range(len(self.img_list) - 1):if i not in self.head_pos.keys():self.head_pos[i] = majority_pos[1]print(i, '丢失,补', majority_pos)  # 丢失则补为图片拼接点的众数,虽然没有什么用...img_width = self.img_widthimg_height = 0# head_pos = []# for i in len(self.head_pos)for i in self.head_pos.keys():img_height += self.head_pos[i] - self.min_head# print(img_height)img_height += self.img_height  # 加最后一张newpic = Image.new(self.img_list[0].mode, (img_width, img_height))height = 0if self.min_head:height += self.min_headnewpic.paste(self.img_list[0].crop((0, 0, img_width, self.min_head)), (0, 0))for i in range(len(self.img_list) - 1):if self.min_head:newpic.paste(self.img_list[i].crop((0, self.min_head, self.img_width, self.head_pos[i])),(0, height))height += self.head_pos[i] - self.min_headelse:newpic.paste(self.img_list[i].crop((0, self.min_head, self.img_width, self.head_pos[i])),(0, height))height += self.head_pos[i]if self.min_head:newpic.paste(self.img_list[-1].crop((0, self.min_head, img_width, img_height)), (0, height))else:newpic.paste(self.img_list[-1], (0, height))# name = str(time.strftime("%Y-%m-%d_%H.%M.%S", time.localtime()))newpic.save('j_temp/jam_outputfile.png')print('saved in j_temp/jam_outputfile.png')def is_same(self, img1, img2):"""简单判断两幅图片是否相同,用于停止滚动截屏,速度非常快!"""h1 = img1.histogram()h2 = img2.histogram()result = math.sqrt(reduce(operator.add, list(map(lambda a, b: (a - b) ** 2, h1, h2))) / len(h1))if result <= 5:return Trueelse:return Falsedef auto_roll(self, area):"""自动滚动截屏,总函数"""x, y, w, h = areaself.img_width = wself.img_height = hspeed = round(1 / self.roll_speed, 2)screen = QApplication.primaryScreen()controler = MouseController()find_left = Commen_Thread(self.find_left_side)find_right = Commen_Thread(self.find_right_size)find_head = Commen_Thread(self.find_the_same_head_to_remove)find_pos = Commen_Thread(self.efind_the_pos)threads = [find_left, find_pos, find_right, find_head]self.in_rolling = Truei = 0img_height = 0controler.position = (area[0] + int(area[2] / 2), area[1] + int(area[3] / 2))while self.in_rolling:pix = screen.grabWindow(QApplication.desktop().winId(), x, y, w, h)  # 区域截屏获取图像pixmapimg = Image.fromqpixmap(pix)  # 将qt的pixmap转为pil模块的img对象self.img_list.append(img)  # 储存图片的列表img_data = list(img.convert('L').getdata())  # 图片灰度化,并把灰度值转为列表an_img_line_data = []for line in range(h):an_img_line_data.append(img_data[line * w:(line + 1) * w])  # 列表分行储存self.images_data_line_list.append(an_img_line_data)if i >= 1:if self.is_same(self.img_list[i - 1], self.img_list[i]):  # 每帧检查是否停止(图片是否相同)self.in_rolling = Falsei += 1breakif img_height == 0:  # 图片有两张以上后,启动线程寻找图片边界点img_height = 1find_head.start()find_left.start()find_right.start()if i == 5:  # 图片大于5张才开始寻找拼接点find_pos.start()controler.scroll(dx=0, dy=-3)  # 滚动屏幕time.sleep(speed)  # 速度控制# img.save('j_temp/{0}.png'.format(i))i += 1print('图片数', i)self.max_arrange = i - 1  # 获取图片序列用于控制寻找边界点的结束for thread in threads:  # 遍历并等待各线程结束thread.wait()# print(thread)if i <= 2:print('过短！一张图还不如直接截呐')self.clear_timer.start(0)returnelif i <= 5:self.find_the_pos()  # 图片小于5张则截完屏在拼接else:find_pos.wait()  # 等待拼接点寻找完成# self.find_the_pos()print('found_pos_done')# try:self.merge_all()  # 调用图片拼接函数# except:#     print('拼接出错错误!请重新截屏！')#     self.clear_timer.start(10000)#     returnprint('可能错误的地方:', self.maybe_errorlist)self.clear_timer.start(10000)  # 10s后初始化内存def listen():"""鼠标监听,截屏中当按下鼠标停止截屏"""global listenerprint("listen")def on_click(x, y, button, pressed):if button == mouse.Button.left:if roller.in_rolling:roller.in_rolling = Falselistener = mouse.Listener(on_click=on_click)listener.start()class Commen_Thread(QThread):"""造的轮子...可用于多线程中不同参数的输入"""def __init__(self, action, *args):super(QThread, self).__init__()self.action = actionself.args = args# print(self.args)def run(self):print('start_thread')if self.args:if len(self.args) == 1:self.action(self.args[0])print(self.args[0])elif len(self.args) == 2:self.action(self.args[0], self.args[1])else:self.action()if __name__ == '__main__':app = QApplication(sys.argv)roller = Splicing_shots()listen()roller.auto_roll((350, 50, 800, 700))#注意选好区域,边界区域过大可能会拼接失败;不能滚动的时候会立即停止!sys.exit(app.exec_())

具体思路

前面已经说了，实现滚动截屏的步骤：滚动–>截屏–>寻找拼接点–>拼接

滚动

滚动部分主要用了pynput模块的滚动功能,该模块还可以实现全局快捷键,具体方法本文不深究。

截屏

由于我的小工具集是采用pyqt作为界面的，就直接用qt的截屏方法了，可以自行改为pil或win32的截屏方法
截屏中止(滚动中止)采用了双重判定，当前后两张图片相同(到了尽头)时可以自动停止，当按下鼠标左键时也会停止

寻找拼接点

寻找拼接点就是比较相邻的图片，寻找下一张图片在前一张图片的相同部分的位置，并记录下来。
但考虑到有些截屏区域是包含不滚动部分的，即所有图片都有相同的头部或边框，所以截取的图片就不能直接用来寻找拼接点，需要比较多张图片并去除相同部分的影响；
所以这一部分又可以分几步：把图片灰度化(不然就要比较rgb3个通道,处理时间大大增加)–>把图片的灰度值储存于数组中–>比较前n张图片并找出所有相同的边界点(上、左、右，下由于拼接的时候会覆盖掉就不用识别了)–>排除相同的部分开始寻找拼接的界限(逐行比较)

更具体的实现已经在代码中标明了,仔细看注释应该可以看懂,可以根据需要改动代码,例如可以完全去除qt库(这个库是真的大!)

听说还可以用opencv实现图片拼接功能(而且效率upup的),等我研究一下再来改蛤

忘了回来更新了…
opencv实现的滚动截屏方法,比暴力拼接不知道快到哪里去了…自行查看
https://github.com/fandesfyf/roll_screenshot

本文对应GitHub项目地址开源万岁!
作者Fandes,转载请标明出处,其实不标明我也不知道…