模式识别导论（实验一）使用OpenCV进行图像读写与编码误差分析

在此记录一下自己课程实验的结果。

实验要求：

1. 通过OpenCV程序从硬盘读取一幅原始图像（格式不限，但最好是分辨率比较高的一幅图像）

2. 将读入内存的图像以不同的压缩率保存为多个JPEG文件，每一个文件对应一个压缩率。

3. 比较原始图像和每一个3.中生成的JPEG图像之间的颜色误差，且根据此颜色误差，画出一个率失真（rate-distortion）曲线图。Rate为编码率，可以使用平均每像素所用比特数来计算，亦即JPEG文件所占总比特数除以像素总数。两幅图像的误差比较可以通过逐像素比较颜色值来计算，具体公式需要自己设计。
4. 将以上3. 和 4. 的JPEG换成PNG，再做一遍同样的实验。

将读入内存的图像以不同的压缩率保存为多个JPEG文件，每一个文件对应一个压缩率。

下面代码段为读取JPG图像，以不同的图片质量保存为JPEG格式图像。其原本图片的默认值为95。

import cv2
PATH = 'C:/Users/lishu/Desktop/code/PatternRecognition/PatternRecognition/img/'
if __name__ == '__main__':
    img = cv2.imread(PATH+'test01.jpg')  #读取图像
    print(img)
    # 设值改变图像可缩放
    cv2.namedWindow('image', cv2.WINDOW_NORMAL)

    #将JPG文件保存为不同图像质量的图像
    cv2.imwrite(PATH+'QUALITY100.JPEG', img, [int(cv2.IMWRITE_JPEG_QUALITY),100])
    cv2.imwrite(PATH+'QUALITY90.JPEG', img, [int(cv2.IMWRITE_JPEG_QUALITY),90])
    cv2.imwrite(PATH+'QUALITY80.JPEG', img, [int(cv2.IMWRITE_JPEG_QUALITY),80])
    cv2.imwrite(PATH+'QUALITY70.JPEG', img, [int(cv2.IMWRITE_JPEG_QUALITY),70])
    cv2.imwrite(PATH+'QUALITY60.JPEG', img, [int(cv2.IMWRITE_JPEG_QUALITY),60])
    cv2.imwrite(PATH+'QUALITY50.JPEG', img, [int(cv2.IMWRITE_JPEG_QUALITY),50])
    cv2.imwrite(PATH+'QUALITY40.JPEG', img, [int(cv2.IMWRITE_JPEG_QUALITY),40])
    cv2.imwrite(PATH+'QUALITY30.JPEG', img, [int(cv2.IMWRITE_JPEG_QUALITY),30])
    cv2.imwrite(PATH+'QUALITY20.JPEG', img, [int(cv2.IMWRITE_JPEG_QUALITY),20])
    cv2.imwrite(PATH+'QUALITY10.JPEG', img, [int(cv2.IMWRITE_JPEG_QUALITY),10])


    # cv2.imshow('image', img)
    # cv2.waitKey(0)
    # cv2.destroyAllWindows()

下面代码段为读取JPG图像，以不同的图片压缩比保存为PNG格式图像。其原本图片的默认值为10。

import cv2
PATH = 'C:/Users/lishu/Desktop/code/PatternRecognition/PatternRecognition/img/'
if __name__ == '__main__':
    img = cv2.imread(PATH+'test01.jpg')  #读取图像
    print(img)
    # 设值改变图像可缩放
    cv2.namedWindow('image', cv2.WINDOW_NORMAL)

    #将JPG文件保存为不同图像质量的图像
    cv2.imwrite(PATH+'compression9.PNG', img, [int(cv2.IMWRITE_PNG_COMPRESSION),9])
    cv2.imwrite(PATH+'compression8.PNG', img, [int(cv2.IMWRITE_PNG_COMPRESSION),8])
    cv2.imwrite(PATH+'compression7.PNG', img, [int(cv2.IMWRITE_PNG_COMPRESSION),7])
    cv2.imwrite(PATH+'compression6.PNG', img, [int(cv2.IMWRITE_PNG_COMPRESSION),6])
    cv2.imwrite(PATH+'compression5.PNG', img, [int(cv2.IMWRITE_PNG_COMPRESSION),5])
    cv2.imwrite(PATH+'compression4.PNG', img, [int(cv2.IMWRITE_PNG_COMPRESSION),4])
    cv2.imwrite(PATH+'compression3.PNG', img, [int(cv2.IMWRITE_PNG_COMPRESSION),3])
    cv2.imwrite(PATH+'compression2.PNG', img, [int(cv2.IMWRITE_PNG_COMPRESSION),2])
    cv2.imwrite(PATH+'compression1.PNG', img, [int(cv2.IMWRITE_PNG_COMPRESSION),1])


    # cv2.imshow('image', img)
    # cv2.waitKey(0)
    # cv2.destroyAllWindows()

图片创建完成后，进行编码误差分析，并作出distortion-rate分析图像。

对于每个图像，其rate的如下：

    rate = 文件大小(Byte) / 像素总数

distortion计算方法如下：

对于生成图：
  计算其每个像素点RGB和原图像素点RGB的绝对值之和，记为total
  得到其像素点*3 计RGB总数 记为size
   distortion = total / size

代码如下：

JPEG：

import cv2
import os
import matplotlib.pyplot as plt
import numpy as np
DPATH = 'C:/Users/lishu/Desktop/code/PatternRecognition/experience1/img/'

if __name__ == '__main__':
    path95 = DPATH+'test01.jpg'

    pathList = []

    for i in range(10):
        pathList.append(DPATH+'QUALITY'+ str(i*10+10) +'.JPEG')
    length = len(pathList)
    img95 = cv2.imread(path95)  #读取图像

    imgList = []
    for i in range(length):
        imgList.append(cv2.imread(DPATH+'QUALITY'+ str(i*10+10) +'.JPEG'))
    rateList = []
    for i in range(length):
        rateList.append(os.path.getsize(pathList[i]) * 8 / (imgList[i].size / 3))

    distortionList = []

    a = img95.astype(np.float32)
    for k in range(length):
        b = imgList[k].astype(np.float32)
        c = np.maximum(a - b, b - a)
        total = np.sum(c)
        distortionList.append(total / img95.size)
    print(rateList)
    print(distortionList)

    plt.plot(rateList, distortionList)
    plt.xlabel('rate')  # 横坐标轴的标题
    plt.ylabel('distortion')  # 纵坐标轴的标题
    plt.show()

PNG:

import cv2
import os
import matplotlib.pyplot as plt
import numpy as np
DPATH = 'C:/Users/lishu/Desktop/code/PatternRecognition/experience1/img/'

if __name__ == '__main__':
    path95 = DPATH+'test01.jpg'
    pathList = []

    for i in range(9):
        pathList.append(DPATH+'compression'+ str(i+1) +'.PNG')
    length = len(pathList)
    img95 = cv2.imread(path95)  #读取图像

    imgList = []
    for i in range(length):
        imgList.append(cv2.imread(DPATH+'compression'+ str(i+1) +'.PNG'))
    rateList = []
    for i in range(length):
        rateList.append(os.path.getsize(pathList[i]) * 8 / (imgList[i].size / 3))

    distortionList = []

    a = img95.astype(np.float32)
    for k in range(length):
        b = imgList[k].astype(np.float32)
        total = np.sum(a-b)
        distortionList.append(total / img95.size)
    print(rateList)
    print(distortionList)
    plt.plot(rateList, distortionList)
    plt.xlabel('rate')  # 横坐标轴的标题
    plt.ylabel('distortion')  # 纵坐标轴的标题
    plt.show()