阅读 19853

几十行代码,修复马赛克表情包|Python 主题月

本文正在参加「Python主题月」,详情查看活动链接

是不是每次偷表情包时,都感觉模糊感, 想找一个清楚点

先观察

  1. 模糊的地方就是有些小黑色点
  2. 只需去掉小黑点,就清楚了

先看下效果

原图修复后
src=http___5b0988e595225.cdn.sohucs.com_images_20190705_099f2e9f3718465aa3ec54bd94410558.jpeg&refer=http___5b0988e595225.cdn.sohucs.jpegrepair-929ef834-c0ac-483f-81b4-9db25598d85a.png
WX20210513-112540@2x.pngrepair-9cd7d9ff-c567-4f66-be6a-4102b540a862.png

是不是看着还不错

想思路

  1. 如何去掉小黑点,拿到每个像素点
  2. 将小黑点变成白色的就OK了

去操作

  1. 观察每个像素值,找到小黑点的范围

    • 范围大概在 200 - 230

    image.png

  2. 找到眼睛、嘴 的颜色,防止吧眼也变白了

    • 大概在 20-140

    image.png

  3. 遍历每个像素点

    • 200 - 230 颜色加几十,变白一些,不能超过255
    • 20 - 140 颜色减几十,变黑一些,不能设置0,会太黑了
  4. 代码思路

    • 读取图片
    • 遍历像素点
    • 更具修复程度算出,变白或变黑

上代码

代码能力有限,写的不咋地,有兴趣可以研究一下

# !/usr/bin/python3
import threading
import uuid
from tkinter import *
from tkinter import filedialog, messagebox

import cv2
from PIL import ImageTk, Image


class MY_GUI():

    def __init__(self, init_window_name):
        self.init_window_name = init_window_name
        self.out_frame = None
        self.in_frame = None

    # 绘制窗口
    def set_init_window(self):
        self.init_window_name.title("图片处理")  # 窗口名
        self.init_window_name.geometry('950x681+10+10')

        self.input_label = Label(self.init_window_name, text="路径:")
        self.input_label.place(x=0, y=0, width=80, height=30)
        self.input_path = Entry(self.init_window_name)
        self.input_path.place(x=80, y=0, width=850, height=30)

        self.init_data_label = Label(self.init_window_name, text="修复前")
        self.init_data_label.place(x=0, y=60, width=80, height=30)
        self.result_data_label = Label(self.init_window_name, text="修复后")
        self.result_data_label.place(x=550, y=60, width=80, height=30)

        self.in_img_button = Button(self.init_window_name, text="选择图片", bg="lightblue", width=10,
                                    command=self.select_in_frame)
        self.in_img_button.place(x=400, y=100, width=150, height=30)

        self.degree_label = Label(self.init_window_name, text="修复程度(1-10)")
        self.degree_label.place(x=400, y=150, width=150, height=30)
        self.degree_input = Spinbox(self.init_window_name, from_=1, to=10)  # 调用内部方法  加()为直接调用
        self.degree_input.place(x=400, y=180, width=150, height=30)
        
        self.repair_button = Button(self.init_window_name, text="修复", bg="lightblue", width=10,
                                    command=self.access_pixels)
        self.repair_button.place(x=400, y=230, width=150, height=30)
        self.d_img_button = Button(self.init_window_name, text="保存", bg="lightblue", width=10,
                                   command=self.d_img)
        self.d_img_button.place(x=400, y=280, width=150, height=30)

    def select_in_frame(self):
        path = filedialog.askopenfilename()
        if path is None or path == '':
            messagebox.showinfo("提示", "请选择图片")
            return
        self.input_path.delete(0, END)
        self.input_path.insert(0, path)
        frame = cv2.imread(path)
        self.in_frame = frame

    # 这里就修复图片的代码了
    def access_pixels(self):
        if self.in_frame is None:
            messagebox.showinfo("提示", "请选择图片")
            return
        degreeStr = self.degree_input.get()
        if degreeStr is None or degreeStr == '':
            degreeStr = '0'

        degree = int(degreeStr)
        if degree < 1:
            degree = 1
        elif degree > 10:
            degree = 10

        frame = self.in_frame.copy()

        print(frame.shape)  # shape内包含三个元素:按顺序为高、宽、通道数
        height = frame.shape[0]
        weight = frame.shape[1]
        channels = frame.shape[2]

        showImgHeight = int(height / (weight / 400))
        # 这里是转换rgb 不然 保存下来会变色    
        b1, g1, r1 = cv2.split(self.in_frame.copy())
        rgb_img1 = cv2.merge([r1, g1, b1])
        im = Image.fromarray(cv2.resize(rgb_img1, (400, showImgHeight)))
        imgtk = ImageTk.PhotoImage(image=im)

        self.before2_img = Label(self.init_window_name, image=imgtk).place(x=0, y=90, width=400,
                                                                           height=showImgHeight)
        print("weight : %s, height : %s, channel : %s" % (weight, height, channels))
        for row in range(height):  # 遍历高
            for col in range(weight):  # 遍历宽
                for c in range(channels):  # 便利通道
                    pv = frame[row, col, c]
                    # 这里是判断是不是小黑点
                    if pv > 23 * (11 - degree):
                        # 这里就是计算增加白色 原颜色+修复程度*10,防止太白 颜色都一样
                        if (pv + degree * 10) > 255:
                            frame[row, col, c] = 255
                        else:
                            frame[row, col, c] = (pv + degree * 10)
                    # 这里是判断黑色,是不是需要更黑一些
                    if pv < degree * 15:
                        frame[row, col, c] = 0

        self.out_frame = frame
        b, g, r = cv2.split(self.out_frame.copy())
        rgb_img = cv2.merge([r, g, b])
        im2 = Image.fromarray(cv2.resize(rgb_img, (400, showImgHeight)))
        imgtk2 = ImageTk.PhotoImage(image=im2)
        self.img2 = Label(self.init_window_name, image=imgtk2).place(x=550, y=90, width=400, height=showImgHeight)
        self.img2.place()

    def d_img(self):
        if self.out_frame is None:
            messagebox.showinfo("提示", "请先修复图片")
            return
        path = filedialog.askdirectory()
        if path is None or path == '':
            messagebox.showinfo("提示", "请选择输出路径")
            return
        savepath = path + "/repair-" + str(uuid.uuid4()) + ".png"
        cv2.imwrite(savepath, self.out_frame)


if __name__ == '__main__':
    init_window = Tk()
    ZMJ_PORTAL = MY_GUI(init_window)
    ZMJ_PORTAL.set_init_window()
    init_window.mainloop()

复制代码

代码差不多了,看看效果吧,有不懂的可以评论问

j1.png

j2.png

j3.png

效果还行,不适用于颜色复杂的表情包, 简单的颜色还是可以的

文章分类
后端
文章标签