图像增强总结和代码
Posted buyizhiyou
tags:
篇首语：本文由小常识网(cha138.com)小编为大家整理，主要介绍了图像增强总结和代码相关的知识，希望对你有一定的参考价值。
调整对比度
- 直方图均衡化/对比度限制自适应直方图均衡化(CLAHE)/gamma变换/对数变化
- 代码实现
import cv2
import matplotlib.pyplot as plt
import numpy as np
def equalizeHist(im):
# 对彩色图像分r,g,b三个通道分别做，效果不好
# (b, g, r) = cv2.split(im)
# bH = cv2.equalizeHist(b)
# gH = cv2.equalizeHist(g)
# rH = cv2.equalizeHist(r)
# dst = cv2.merge((bH, gH, rH))
#在hls空间操作,l为亮度相关通道
hls = cv2.cvtColor(im, cv2.COLOR_BGR2HLS)
l = hls[:, :, 1]
l_ = cv2.equalizeHist(l)
hls[:, :, 1] = l_
dst = cv2.cvtColor(hls, cv2.COLOR_HLS2BGR)
#在Lab空间操作,L为亮度相关的通道
# lab = cv2.cvtColor(im, cv2.COLOR_BGR2Lab)
# l = lab[:, :, 0]
# l_ = cv2.equalizeHist(l)
# lab[:, :, 0] = l_
# dst = cv2.cvtColor(lab, cv2.COLOR_Lab2BGR)
return dst
def clahe(im):
‘‘‘
contrast limited adaptive histogram equalization for color image
‘‘‘
hls = cv2.cvtColor(im, cv2.COLOR_BGR2HLS)
l = hls[:, :, 1]
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(16, 16))
l_ = clahe.apply(l)
hls[:, :, 1] = l_
dst = cv2.cvtColor(hls, cv2.COLOR_HLS2BGR)
return dst
def logTransform(im):
img_norm = im/255.0
img_log = np.log10(1+9*img_norm)*255
img_log = img_log.astype(np.uint8)
return img_log
def gammaTransform(im):
# hls = cv2.cvtColor(im, cv2.COLOR_BGR2HLS)
# l = hls[:, :, 1]
# l_ = l/255.0 # 注意255.0得采用浮点数
# l_gamma = np.power(l_, 0.4)*255.0
# l_gamma = l_gamma.astype(np.uint8)
# hls[:, :, 1] = l_gamma
# dst = cv2.cvtColor(hls, cv2.COLOR_HLS2BGR)
img_norm = im/255.0 # 注意255.0得采用浮点数
dst = np.power(img_norm, 0.4)*255.0
dst = dst.astype(np.uint8)
return dst
# im = cv2.imread("test4.jpg")
# dst = gammaTransform(im)
# cv2.imshow("dst", dst)
# cv2.imshow("ori", im)
# cv2.waitKey()
去雾
- 暗通道先验去雾
- 代码实现
import cv2
import numpy as np
from adjustContrast import logTransform, gammaTransform
def zmMinFilterGray(src, r=7):
‘‘‘最小值滤波，r是滤波器半径‘‘‘
‘‘‘if r <= 0:
return src
h, w = src.shape[:2]
I = src
res = np.minimum(I , I[[0]+range(h-1) , :])
res = np.minimum(res, I[range(1,h)+[h-1], :])
I = res
res = np.minimum(I , I[:, [0]+range(w-1)])
res = np.minimum(res, I[:, range(1,w)+[w-1]])
return zmMinFilterGray(res, r-1)‘‘‘
return cv2.erode(src, np.ones((2*r+1, 2*r+1))) # 使用opencv的erode函数更高效
def guidedfilter(I, p, r, eps):
‘‘‘引导滤波，直接参考网上的matlab代码‘‘‘
height, width = I.shape
m_I = cv2.boxFilter(I, -1, (r, r))
m_p = cv2.boxFilter(p, -1, (r, r))
m_Ip = cv2.boxFilter(I*p, -1, (r, r))
cov_Ip = m_Ip-m_I*m_p
m_II = cv2.boxFilter(I*I, -1, (r, r))
var_I = m_II-m_I*m_I
a = cov_Ip/(var_I+eps)
b = m_p-a*m_I
m_a = cv2.boxFilter(a, -1, (r, r))
m_b = cv2.boxFilter(b, -1, (r, r))
return m_a*I+m_b
def getV1(m, r, eps, w, maxV1): # 输入rgb图像，值范围[0,1]
‘‘‘计算大气遮罩图像V1和光照值A, V1 = 1-t/A‘‘‘
V1 = np.min(m, 2) # 得到暗通道图像
V1 = guidedfilter(V1, zmMinFilterGray(V1, 7), r, eps) # 使用引导滤波优化
bins = 2000
ht = np.histogram(V1, bins) # 计算大气光照A
d = np.cumsum(ht[0])/float(V1.size)
for lmax in range(bins-1, 0, -1):
if d[lmax] <= 0.999:
break
A = np.mean(m, 2)[V1 >= ht[1][lmax]].max()
V1 = np.minimum(V1*w, maxV1) # 对值范围进行限制
return V1, A
def deHaze(m, r=81, eps=0.001, w=0.95, maxV1=0.80, bGamma=False):
Y = np.zeros(m.shape)
V1, A = getV1(m, r, eps, w, maxV1) # 得到遮罩图像和大气光照
for k in range(3):
Y[:, :, k] = (m[:, :, k]-V1)/(1-V1/A) # 颜色校正
Y = np.clip(Y, 0, 1)
if bGamma:
Y = Y**(np.log(0.5)/np.log(Y.mean())) # gamma校正,默认不进行该操作
return Y
if __name__ == ‘__main__‘:
src = cv2.imread(‘test2.jpg‘)
dst = deHaze(src/255.0)*255
dst = dst.astype(np.uint8)
dst2 = logTransform(dst)
cv2.imshow("src", src)
cv2.imshow("dehaze", dst)
cv2.imshow("dehaze2", dst2)
cv2.waitKey(0)
同态滤波和retinex增强
import numpy as np
import cv2
config = {
"sigma_list": [15, 80, 250],
"G": 5.0,
"b": 25.0,
"alpha": 125.0,
"beta": 46.0,
"low_clip": 0.01,
"high_clip": 0.99
}
def homomorphic_filter(src, d0=10, r1=0.5, rh=2, c=4, h=2.0, l=0.5):
gray = src.copy()
if len(src.shape) > 2:
gray = cv2.cvtColor(src, cv2.COLOR_BGR2GRAY)
gray = np.float64(gray)
rows, cols = gray.shape
gray_fft = np.fft.fft2(gray)
gray_fftshift = np.fft.fftshift(gray_fft)
dst_fftshift = np.zeros_like(gray_fftshift)
M, N = np.meshgrid(np.arange(-cols // 2, cols // 2), np.arange(-rows//2, rows//2))
D = np.sqrt(M ** 2 + N ** 2)
Z = (rh - r1) * (1 - np.exp(-c * (D ** 2 / d0 ** 2))) + r1
dst_fftshift = Z * gray_fftshift
dst_fftshift = (h - l) * dst_fftshift + l
dst_ifftshift = np.fft.ifftshift(dst_fftshift)
dst_ifft = np.fft.ifft2(dst_ifftshift)
dst = np.real(dst_ifft)
dst = np.uint8(np.clip(dst, 0, 255))
return dst
def singleScaleRetinex(img, sigma):
retinex = np.log10(img) - np.log10(cv2.GaussianBlur(img, (0, 0), sigma))
return retinex
def multiScaleRetinex(img, sigma_list):
retinex = np.zeros_like(img)
for sigma in sigma_list:
retinex += singleScaleRetinex(img, sigma)
retinex = retinex / len(sigma_list)
return retinex
def colorRestoration(img, alpha, beta):
img_sum = np.sum(img, axis=2, keepdims=True) # (h,w,1)
color_restoration = beta * (np.log10(alpha * img) - np.log10(img_sum)) # (h,w,3)
return color_restoration
def simplestColorBalance(img, low_clip, high_clip):
total = img.shape[0] * img.shape[1]
for i in range(img.shape[2]):
unique, counts = np.unique(img[:, :, i], return_counts=True)
current = 0
for u, c in zip(unique, counts):
if float(current) / total < low_clip:
low_val = u
if float(current) / total < high_clip:
high_val = u
current += c
img[:, :, i] = np.maximum(np.minimum(img[:, :, i], high_val), low_val)
return img
def MSRCR(img, sigma_list, G, b, alpha, beta, low_clip, high_clip):
img = np.float64(img) + 1.0
img_retinex = multiScaleRetinex(img, sigma_list)
img_color = colorRestoration(img, alpha, beta)
img_msrcr = G * (img_retinex * img_color + b)
for i in range(img_msrcr.shape[2]):
img_msrcr[:, :, i] = (img_msrcr[:, :, i] - np.min(img_msrcr[:, :, i])) / (np.max(img_msrcr[:, :, i]) - np.min(img_msrcr[:, :, i])) * 255
img_msrcr = np.uint8(np.minimum(np.maximum(img_msrcr, 0), 255))
img_msrcr = simplestColorBalance(img_msrcr, low_clip, high_clip)
return img_msrcr
def automatedMSRCR(img, sigma_list):
img = np.float64(img) + 1.0
img_retinex = multiScaleRetinex(img, sigma_list)
for i in range(img_retinex.shape[2]):
unique, count = np.unique(
np.int32(img_retinex[:, :, i] * 100), return_counts=True)
for u, c in zip(unique, count):
if u == 0:
zero_count = c
break
low_val = unique[0] / 100.0
high_val = unique[-1] / 100.0
for u, c in zip(unique, count):
if u < 0 and c < zero_count * 0.1:
low_val = u / 100.0
if u > 0 and c < zero_count * 0.1:
high_val = u / 100.0
break
img_retinex[:, :, i] = np.maximum(np.minimum(
img_retinex[:, :, i], high_val), low_val)
img_retinex[:, :, i] = (img_retinex[:, :, i] - np.min(img_retinex[:, :, i])) / (np.max(img_retinex[:, :, i]) - np.min(img_retinex[:, :, i])) * 255
img_retinex = np.uint8(img_retinex)
return img_retinex
def MSRCP(img, sigma_list, low_clip, high_clip):
img = np.float64(img) + 1.0
intensity = np.sum(img, axis=2) / img.shape[2]
retinex = multiScaleRetinex(intensity, sigma_list)
intensity = np.expand_dims(intensity, 2)
retinex = np.expand_dims(retinex, 2)
intensity1 = simplestColorBalance(retinex, low_clip, high_clip)
intensity1 = (intensity1 - np.min(intensity1)) / (np.max(intensity1) - np.min(intensity1)) * 255.0 + 1.0
img_msrcp = np.zeros_like(img)
for y in range(img_msrcp.shape[0]):
for x in range(img_msrcp.shape[1]):
B = np.max(img[y, x])
A = np.minimum(256.0 / B, intensity1[y, x, 0] / intensity[y, x, 0])
img_msrcp[y, x, 0] = A * img[y, x, 0]
img_msrcp[y, x, 1] = A * img[y, x, 1]
img_msrcp[y, x, 2] = A * img[y, x, 2]
img_msrcp = np.uint8(img_msrcp - 1.0)
return img_msrcp
im = cv2.imread("test3.jpg")
msrcr = MSRCR(im,
config[‘sigma_list‘],
config[‘G‘],
config[‘b‘],
config[‘alpha‘],
config[‘beta‘],
config[‘low_clip‘],
config[‘high_clip‘])
amsrcr = automatedMSRCR(
im,
config[‘sigma_list‘]
)
msrcp = MSRCP(im,
config[‘sigma_list‘],
config[‘low_clip‘],
config[‘high_clip‘])
cv2.imshow("msrcr", msrcr)
cv2.imshow("amsrcr", amsrcr)
cv2.imshow("msrcp", msrcp)
cv2.waitKey()
自动色阶
- 自动色阶去雾
- 代码实现
import numpy as np
import cv2
‘‘‘
autoLevel自动色阶去雾
‘‘‘
def nothing(*arg):
pass
MAX_VALUE = 30
cv2.namedWindow("autoLevel", cv2.WINDOW_NORMAL)
# 滑动块
cv2.createTrackbar("min", "autoLevel", 0, MAX_VALUE, nothing)
cv2.createTrackbar("max", "autoLevel", 0, MAX_VALUE, nothing)
def ComputeHist(img):
h, w = img.shape
hist, bin_edge = np.histogram(img.reshape(1, w*h), bins=list(range(257)))
return hist
def ComputeMinLevel(hist, rate, pnum):
sum = 0
for i in range(256):
sum += hist[i]
if (sum >= (pnum * rate * 0.01)):
return i
def ComputeMaxLevel(hist, rate, pnum):
sum = 0
for i in range(256):
sum += hist[255-i]
if (sum >= (pnum * rate * 0.01)):
return 255-i
def LinearMap(minlevel, maxlevel):
if (minlevel >= maxlevel):
return []
else:
newmap = np.zeros(256)
for i in range(256):
if (i < minlevel):
newmap[i] = 0
elif (i > maxlevel):
newmap[i] = 255
else:
newmap[i] = (i-minlevel)/(maxlevel-minlevel) * 255
return newmap
def autoLevel(img, minrate, maxrate):
h, w, d = img.shape
newimg = np.zeros([h, w, d])
for i in range(d):
imgmin = np.min(img[:, :, i])
imgmax = np.max(img[:, :, i])
imghist = ComputeHist(img[:, :, i])
minlevel = ComputeMinLevel(imghist, minrate, h*w)
maxlevel = ComputeMaxLevel(imghist, maxrate, h*w)
newmap = LinearMap(minlevel, maxlevel)
if (newmap.size == 0):
continue
for j in range(h):
newimg[j, :, i] = newmap[img[j, :, i]]
return newimg
img = cv2.imread(‘test3.jpg‘)
while True:
minval = cv2.getTrackbarPos(‘min‘, ‘autoLevel‘)
maxval = cv2.getTrackbarPos(‘max‘, ‘autoLevel‘)
newimg = autoLevel(img, minval, maxval)
cv2.imshow(‘autoLevel‘, newimg/255)
ch = cv2.waitKey(5)
# 按 ESC 键退出
if ch == 27:
break
色彩增强
- 饱和度调整
- 代码实现
import numpy as np
import cv2
image = cv2.imread(‘test.jpg‘)
alpha = 150
MAX_VALUE = 200
# 调节饱和度窗口
cv2.namedWindow("ColorEnhance", cv2.WINDOW_NORMAL)
def nothing(*arg):
pass
def adjustSaturation(image, alpha):
# 图像归一化，且转换为浮点型
image2 = image.astype(np.float32)
image2 = image2 / 255.0
# 颜色空间转换 BGR转为HLS,在hls空间做调整比较好
hlsImg = cv2.cvtColor(image2, cv2.COLOR_BGR2HLS)
hlsImg[:, :, 2] = pow(hlsImg[:, :, 2], alpha/100) # 幂函数
hlsImg[:, :, 2][hlsImg[:, :, 2] > 1] = 1
dst = cv2.cvtColor(hlsImg, cv2.COLOR_HLS2BGR)
return dst
def coloeEnhance(image, alpha):
image2 = image.astype(np.float32)
image2 = image2 / 255.0
W = np.array([0.0721, 0.7154, 0.2125])
intensity = image2.dot(W)
intensity = np.expand_dims(intensity, 2)
dst = intensity*(1-alpha/100)+image2*(alpha/100)
dst = np.clip(dst, 0, 1)
dst = dst*255
dst = dst.astype(np.uint8)
‘‘‘
W=[0.2125,0.7154,0.0721]
等价于变换矩阵M=[[alpha+w1*(1-alpha),w2*(1-alpha),w3*(1-alpha)],
[w1*(1-alpha),alpha+w2*(1-alpha),w3*(1-alpha)],
[w1*(1-alpha),w2*(1-alpha),alpha+w3*(1-alpha)]]
[R‘,G‘,B‘]=M*[R,G,B]^T
‘‘‘
return dst
# 滑动块
cv2.createTrackbar("alpha", "ColorEnhance", alpha, MAX_VALUE, nothing)
dst = np.zeros(image.shape, np.float32)
while True:
alpha = cv2.getTrackbarPos(‘alpha‘, ‘ColorEnhance‘)
dst = coloeEnhance(image, alpha)
cv2.imshow("ColorEnhance", dst)
ch = cv2.waitKey(5)
# 按 ESC 键退出
if ch == 27:
break
elif ch == ord(‘s‘):
# 按 s 键保存并退出
dst = dst * 255
dst = dst.astype(np.uint8)
cv2.imwrite("dst.jpg", dst)
break
# 关闭所有的窗口
cv2.destroyAllWindows()
‘‘‘
选自python 标准库中的colorsys库,rbg2hls和hls2rgb
‘‘‘
# Some floating point constants
ONE_THIRD = 1.0/3.0
ONE_SIXTH = 1.0/6.0
TWO_THIRD = 2.0/3.0
def rgb_to_hls(r, g, b):
maxc = max(r, g, b)
minc = min(r, g, b)
# Can optimize (maxc+minc) and (maxc-minc)
l = (minc+maxc)/2.0
if minc == maxc:
return 0.0, l, 0.0
if l <= 0.5:
s = (maxc-minc) / (maxc+minc)
else:
s = (maxc-minc) / (2.0-maxc-minc)
rc = (maxc-r) / (maxc-minc)
gc = (maxc-g) / (maxc-minc)
bc = (maxc-b) / (maxc-minc)
if r == maxc:
h = bc-gc
elif g == maxc:
h = 2.0+rc-bc
else:
h = 4.0+gc-rc
h = (h/6.0) % 1.0
return h, l, s
def hls_to_rgb(h, l, s):
if s == 0.0:
return l, l, l
if l <= 0.5:
m2 = l * (1.0+s)
else:
m2 = l+s-(l*s)
m1 = 2.0*l - m2
return (_v(m1, m2, h+ONE_THIRD), _v(m1, m2, h), _v(m1, m2, h-ONE_THIRD))
def _v(m1, m2, hue):
hue = hue % 1.0
if hue < ONE_SIXTH:
return m1 + (m2-m1)*hue*6.0
if hue < 0.5:
return m2
if hue < TWO_THIRD:
return m1 + (m2-m1)*(TWO_THIRD-hue)*6.0
return m1
图像锐化
以上是关于图像增强总结和代码的主要内容，如果未能解决你的问题，请参考以下文章
AIchanllger农作物病害分类竞赛思路总结，也是pytorch进行图像分类的一般步骤，下一篇为源代码详解
图像锐化算法(Image sharpening):拉普拉斯增强和Unsharp Masking(附代码)