对图像做倾斜/仿射处理，数据增强

对图像做倾斜/仿射处理，数据增强

import numpy as np
import cv2,struct
def affine(img, dx=30, dy=30):
    kk=0
    img_h,img_w,_=img.shape
    if  dx<0 or dy<0:
        kk=1
        dx=abs(dx)
        dy=abs(dy)
        img=img[:,::-1,:]
    H, W, C = img.shape
    a = 1.

    b = dx / H

    c = dy / W

    d = 1.
    tx = 0.

    ty = 0.
    _img = np.zeros((H+2, W+2, C), dtype=np.float32)
    _img[1:H+1, 1:W+1] = img
    H_new = np.ceil(dy + H).astype(np.int)

    W_new = np.ceil(dx + W).astype(np.int)
    out = np.zeros((H_new, W_new, C), dtype=np.float32)
    x_new = np.tile(np.arange(W_new), (H_new, 1))

    y_new = np.arange(H_new).repeat(W_new).reshape(H_new, -1)
    adbc = a * d - b * c

    x = np.round((d * x_new-b*y_new)/adbc).astype(np.int)-tx+1

    y = np.round((-c * x_new + a * y_new) / adbc).astype(np.int) - ty + 1

    x = np.minimum(np.maximum(x, 0), W+1).astype(np.int)

    y = np.minimum(np.maximum(y, 0), H+1).astype(np.int)
    out[y_new, x_new] = _img[y, x]

    out = out.astype(np.uint8)
    if kk==1:
        out=out[:,::-1,:]
    out=cv2.resize(out,(img_w,img_h))
    return out

支持正负输入
eg：

img=cv2.imread('xxx.jpg')
out = affine(img, dx=30, dy=30)
out1 = affine(img, dx=-30, dy=30)
cv2.imshow('img',img)
cv2.imshow('out',out)
cv2.imshow('out1',out1)

import matplotlib.pyplot as plt
from scipy import signal
import numpy as np
import copy as cp
import random
import math
import cv2
import collections

#按比例缩小
def DOWNRESIZE(gray , ratio=0.8) :
    height , width = gray.shape[0] , gray.shape[1]
    resize_width , resize_height = int(width * ratio) , int(height * ratio)
    resize_gray = np.zeros([resize_height,resize_width],np.uint8)   #创建一个和原图像 *缩小比例相同像素的空图
    stepsize = 1.0/ratio    #原图像的采集步长
    for i in range (0,resize_height) :
        for j in range (0,resize_width) :
            if i  == 0 and j == 0 :     #缩小图(0,0)地方的像素用原图(1,1)的地方代替
                resize_gray[i][j] = gray[1][1]
            else :      #缩小图其他地方的像素用此时原图对应的坐标-1的地方代替
                resize_gray[i][j] = gray[int(i*stepsize)-1][int(j*stepsize)-1]
    cv2.imshow("old",gray)
    cv2.imshow("Downresize",resize_gray)
    cv2.waitKey(0)


#按比例放大
def UPSIZE(gray , ratio=1.2) :
    height , width = gray.shape[0] , gray.shape[1]
    resize_width , resize_height = int(width * ratio) , int(height * ratio)
    resize_gray = np.zeros([resize_height,resize_width],np.uint8)
    """
    #最邻近插值法
    for i in range (0,resize_height) :
        for j in range (0,resize_width) :
            if i  == 0 and j == 0 :
                resize_gray[i][j] = gray[int(i/ratio)+1][int(j/ratio)+1]
            else :
                resize_gray[i][j] = gray[int(i/ratio)-1][int(j/ratio)-1]
    """
    #双线性插值
    for i in range (0,resize_height) :
        for j in range (0,resize_width) :
            x = i/ratio
            y = j/ratio
            p=(i+0.0)/ratio-x
            q=(j+0.0)/ratio-y
            x=int(x)-1
            y=int(y)-1
            if x+1<i and y+1<j:
                resize_gray[i,j]=int(gray[x,y]*(1-p)*(1-q)+gray[x,y+1]*q*(1-p)+gray[x+1,y]*(1-q)*p+gray[x+1,y+1]*p*q)
    cv2.imshow("old",gray)
    cv2.imshow("Upresize",resize_gray)
    cv2.waitKey(0)

#水平内凹
def CONCAVE(gray) :
    RANGE = 400
    height, width = gray.shape[0], gray.shape[1]
    resize_gray = np.zeros([height, width], np.uint8)
    for i in range(height):
        # 得到正弦波的波形，即j对应的起点
        temp = float((width- RANGE) * math.sin(((2 * math.pi * i) / height) / 2))
        for j in range (int(temp+0.5),int(width-temp)) :
            #每行非黑色区域的长度
            distance = int(width-temp) - int(temp+0.5)
            #缩小的倍率
            ratio = distance / width
            #取点的步长
            stepsize = 1.0/ratio
            #将同意行缩小相同倍率
            resize_gray[i][j] = gray[i][int((j-temp)*stepsize)]
    cv2.imshow("old", gray)
    cv2.imshow("Concave", resize_gray)
    cv2.waitKey(0)


#水平外凸
def CONVEX(gray) :
    RANGE = 400
    height, width = gray.shape[0], gray.shape[1]
    resize_gray = np.zeros([height, width], np.uint8)
    temp1 = []
    for i in range(height):
        # 得到正弦波的波形，即j对应的起点
        temp = float((width- RANGE) * math.sin(((2 * math.pi * i) / height) / 2))
        temp = 200 - temp
        for j in range (int(temp+0.5),int(width-temp)) :
            #每行非黑色区域的长度
            distance = int(width-temp) - int(temp+0.5)
            #缩小的倍率
            ratio = distance / width
            #取点的步长
            stepsize = 1.0/ratio
            #将同意行缩小相同倍率
            resize_gray[i][j] = gray[i][int((j-temp)*stepsize)]
    cv2.imshow("old", gray)
    cv2.imshow("Convex", resize_gray)
    cv2.waitKey(0)

#梯形形变
def TRAPEZOID(gray,k=0.3) :
    height, width = gray.shape[0], gray.shape[1]
    value = k*height
    resize_gray = np.zeros([height, width], np.uint8)
    for i in range (height) :
        temp = int( value - k * i ) 
        for j in range (temp,width-temp) :
            #每行非黑色区域的长度
            distance = int(width-temp) - int(temp+0.5)
            #缩小的倍率
            ratio = distance / width
            #取点的步长
            stepsize = 1.0/ratio
            #将同意行缩小相同倍率
            resize_gray[i][j] = gray[i][int((j-temp)*stepsize)]
    cv2.imshow("old", gray)
    cv2.imshow("Trapezoid", resize_gray)
    cv2.waitKey(0)

#三角形形变
def TRIANGLE(gray,k=0.5) :
    height, width = gray.shape[0], gray.shape[1]
    resize_gray = np.zeros([height, width], np.uint8)
    for i in range (height) :
        temp = int( k * i ) 
        for j in range (temp) :
            #每行非黑色区域的长度
            distance = temp
            #缩小的倍率
            ratio = distance / width
            #取点的步长
            stepsize = 1.0/ratio
            #将同意行缩小相同倍率
            resize_gray[i][j] = gray[i][int((j-temp)*stepsize)]
    cv2.imshow("old", gray)
    cv2.imshow("Triangle", resize_gray)
    cv2.waitKey(0)

#S形变
def SSHAPE(gray,RANGE=450) :
    height, width = gray.shape[0], gray.shape[1]
    resize_gray = np.zeros([height, width], np.uint8)
    for i in range(height):
        # 得到正弦波的波形，即j对应的起点
        temp = float((width - RANGE) / 2 + (width- RANGE) * math.sin((2 * math.pi * i) / height + math.pi) / 2)
        for j in range (int(temp+0.5),int(RANGE+temp)) :
            #映射关系
            m = int(((j-temp) * width / RANGE))
            if m >= width :
                m=width-1
            if m < 0 :
                m=0
            resize_gray[i,j]=gray[i,m]

    cv2.imshow("old", gray)
    cv2.imshow("Sshape", resize_gray)
    cv2.waitKey(0)

#图片旋转
def ROTATE(gray, ANGLE = 45,center=None, scale=1.0) :
    #将角度转化为弧度
    """ #特别神奇的图案
    ANGLE =  math.radians(ANGLE)
    height, width = gray.shape[0], gray.shape[1]
    resize_gray = np.zeros([800,800], np.uint8)
    for i in range (height) :
        for j in range (width) :
            angle = math.asin( i / width )
            ANGLE = angle + ANGLE
            x,y = int(width * math.cos(ANGLE)) , int(width * math.sin(ANGLE))
            resize_gray[x][y] = gray[i][j]
    cv2.imshow("old", gray)
    cv2.imshow("Rotate", resize_gray)
    cv2.waitKey(0)
    """
    (height, width) = gray.shape[:2]
    (cX, cY) = (width // 2, height // 2)
 
    M = cv2.getRotationMatrix2D((cX, cY), -ANGLE, 1.0)
    cos = np.abs(M[0, 0])
    sin = np.abs(M[0, 1])
 
    nW = int((height * sin) + (width * cos))
    nH = int((height * cos) + (width * sin))
 
    M[0, 2] += (nW / 2) - cX
    M[1, 2] += (nH / 2) - cY
 
    shuchu=cv2.warpAffine(gray, M, (nW, nH))

    cv2.imshow("old", gray)
    cv2.imshow("Rotate", shuchu)
    cv2.waitKey(0)

#图片平移
def HORIZONTALMOVE(gray,move_x,move_y) :
    height, width = gray.shape[0], gray.shape[1]
    resize_gray = np.zeros([height, width], np.uint8)
    for i in range (height) :
        for j in range (width) :
            if i + move_x < height and j + move_y < width :
                resize_gray[i][j] = gray[i+move_x][j+move_y]
            else :
                resize_gray [i][j] = 0
    cv2.imshow("old", gray)
    cv2.imshow("Horizeontalmove", resize_gray)
    cv2.waitKey(0)

def MAIN() :
    image = cv2.imread(r"D:\Code\Python\2.jpg")
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    while True :
        print("0: 按比例缩小\n1: 按比例放大\n2: 水平外凹\n3: 水平外凸\n4: 梯形变形\n5: 三角形变形\n6: S形变形\n7: 图片旋转\n8: 图片平移\n")
        keyinput = input ("请输入数字0—8，选择你需要的变换(按'Q'或'q'退出)：\n")
        if keyinput == "0" :
            ratio = float(input("请输入缩小倍数：\n"))
            DOWNRESIZE(gray , ratio)
            continue
        elif keyinput == "1" :
            ratio = float(input("请输入放大倍数：\n"))
            UPSIZE(gray , ratio)
            continue
        elif keyinput == "2" :
            CONCAVE(gray)
            continue
        elif keyinput == "3" :
            CONVEX(gray)
            continue
        elif keyinput == "4" :
            k = float(input("请输入梯形的斜度：\n"))
            TRAPEZOID(gray,k)
            continue
        elif keyinput == "5" :
            k = float(input("请输入三角形的斜度：\n"))
            TRIANGLE(gray,k)
            continue
        elif keyinput == "6" :
            RANGE = float(input("请输入变换强度：\n"))
            SSHAPE(gray,RANGE)
            continue
        elif keyinput == "7" :
            ANGLE = float(input("请输入旋转角度：\n"))
            ROTATE(gray,ANGLE)
            continue
        elif keyinput == "8" :
            move_x = int(input("请输入x轴偏移值：\n"))
            move_y = int(input("请输入y轴偏移值：\n"))
            HORIZONTALMOVE(gray,move_x,move_y)
            continue
        elif keyinput == "Q" or "q" :
            break
        else :
            keyinput = input("您输入的字符有误，请输入 0-8 之间的数字：")

 
if __name__ == "__main__":
    MAIN()

rgb图片

import cv2
import numpy as np
from collections import Counter
import matplotlib.pyplot as plt
from scipy import signal
import numpy as np
import copy as cp
import random
import math
import cv2
import collections
def TRAPEZOID(img,k=0.1) :
    for mk in range(3):
        gray=img[...,mk]
        
        height, width = gray.shape[0], gray.shape[1]
        value = k*height
        resize_gray = np.zeros([height, width], np.uint8)
        for i in range (height) :
            temp = int( value - k * i ) 
            for j in range (temp,width-temp) :
                #每行非黑色区域的长度
                distance = int(width-temp) - int(temp+0.5)
                #缩小的倍率
                ratio = distance / width
                #取点的步长
                stepsize = 1.0/ratio
                #将同意行缩小相同倍率
                resize_gray[i][j] = gray[i][int((j-temp)*stepsize)]
        img[...,mk]=resize_gray
    return img
 #旋转梯形挤压，再回来
 img=cv2.rotate(img, cv2.ROTATE_90_CLOCKWISE)
 warped=TRAPEZOID(img)
 warped=cv2.rotate(warped, cv2.ROTATE_90_COUNTERCLOCKWISE)