# 0416midterm

### Data
讀取期中考提供的圖片，4張386*382大小，分別為過暗、過亮、灰階、普通。

import cv2
import matplotlib.pyplot as plt
import numpy as np

#cv2讀取圖片
dark = cv2.imread('dark.PNG')
bright = cv2.imread('bright.PNG')
gray = cv2.imread('gray.PNG')
normal = cv2.imread('normal.PNG')

#用pyplot呈現圖片
plt.figure(figsize=(20,5)) #设置窗口大小
plt.subplot(1,4,1), plt.title('dark') #把圖片排列成1行顯示4個
plt.imshow(dark), plt.axis('off') #把x軸的標籤去除
plt.subplot(1,4,2), plt.title('bright')
plt.imshow(bright), plt.axis('off')
plt.subplot(1,4,3), plt.title('gray')
plt.imshow(gray), plt.axis('off')
plt.subplot(1,4,4), plt.title('normal')
plt.imshow(normal), plt.axis('off')
plt.show() #印出圖片


---

### power-low
power-low公式s=cr^γ，將4張圖片分別做公式運算，先將圖片除以255以便把數值壓在0~1之間，做γ次方的運算後，在乘上255恢復色彩顯示範圍。

#設定兩個gamma值
gamma_h = 2.0 
gamma_l = 0.5

#power-low運算
dark_t_h = np.array(255*(dark/255)**gamma_h, dtype = 'uint8')
bright_t_h = np.array(255*(bright/255)**gamma_h, dtype = 'uint8')
gray_t_h = np.array(255*(gray/255)**gamma_h, dtype = 'uint8')
normal_t_h = np.array(255*(normal/255)**gamma_h, dtype = 'uint8')

dark_t_l = np.array(255*(dark/255)**gamma_l, dtype = 'uint8')
bright_t_l = np.array(255*(bright/255)**gamma_l, dtype = 'uint8')
gray_t_l = np.array(255*(gray/255)**gamma_l, dtype = 'uint8')
normal_t_l = np.array(255*(normal/255)**gamma_l, dtype = 'uint8')

plt.figure(figsize=(20,5))
plt.subplot(1,4,1), plt.title('dark_transform_h')
plt.imshow(dark_t_h), plt.axis('off')
plt.subplot(1,4,2), plt.title('bright_transform_h')
plt.imshow(bright_t_h), plt.axis('off')
plt.subplot(1,4,3), plt.title('gray_transform_h')
plt.imshow(gray_t_h), plt.axis('off')
plt.subplot(1,4,4), plt.title('normal_transform_h')
plt.imshow(normal_t_h), plt.axis('off')
plt.show()

plt.figure(figsize=(20,5))
plt.subplot(1,4,1), plt.title('dark_transform_l')
plt.imshow(dark_t_l), plt.axis('off')
plt.subplot(1,4,2), plt.title('bright_transform_l')
plt.imshow(bright_t_l), plt.axis('off')
plt.subplot(1,4,3), plt.title('gray_transform_l')
plt.imshow(gray_t_l), plt.axis('off')
plt.subplot(1,4,4), plt.title('normal_transform_l')
plt.imshow(normal_t_l), plt.axis('off')
plt.show()

cv2.imwrite("dark_transform_h.jpg", dark_t_h)
cv2.imwrite("bright_transform_h.jpg", bright_t_h)
cv2.imwrite("gray_transform_h.jpg", gray_t_h)
cv2.imwrite("normal_transform_h.jpg", normal_t_h)

cv2.imwrite("dark_transform_l.jpg", dark_t_l)
cv2.imwrite("bright_transform_l.jpg", bright_t_l)
cv2.imwrite("gray_transform_l.jpg", gray_t_l)
cv2.imwrite("normal_transform_l.jpg", normal_t_l)


---

### histogram_equalization
histogram-equalization方法能夠將cdf分布從原本的聚集，變為分散狀態以達到明暗對比分明的效果，做法式將灰度最大值推到256，把最小值推到0，其中他們彼此的區間會被等量的加大。

def img_cdf(img, title): #圖片的cdf
    hist, bins = np.histogram(img.flatten(),256,[0,256])
    plt.hist(img.flatten(),256,[0,256],color="g")
    plt.xlim([0,256])
    plt.title(title)
    plt.show()
    
def histogram_equalization(img_in):
    b, g, r = cv2.split(img_in) #切出rgb三個值
    hist_b, bin_b = np.histogram(b.flatten(),256,[0,256])
    hist_g, bin_g = np.histogram(g.flatten(),256,[0,256])
    hist_r, bin_r = np.histogram(r.flatten(),256,[0,256])
    
    cdf_b = np.cumsum(hist_b) #將數組累加
    cdf_g = np.cumsum(hist_g)
    cdf_r = np.cumsum(hist_r)
    
    cdf_m_b = np.ma.masked_equal(cdf_b,0) #直方圖均衡化
    cdf_m_b = (cdf_m_b - cdf_m_b.min())*255/(cdf_m_b.max()-cdf_m_b.min()) #公式運算
    cdf_final_b = np.ma.filled(cdf_m_b,0).astype('uint8') #將0補上平均值
    
    cdf_m_g = np.ma.masked_equal(cdf_g,0)
    cdf_m_g = (cdf_m_g - cdf_m_g.min())*255/(cdf_m_g.max()-cdf_m_g.min())
    cdf_final_g = np.ma.filled(cdf_m_g,0).astype('uint8')
    
    cdf_m_r = np.ma.masked_equal(cdf_r,0)
    cdf_m_r = (cdf_m_r - cdf_m_r.min())*255/(cdf_m_r.max()-cdf_m_r.min())
    cdf_final_r = np.ma.filled(cdf_m_r,0).astype('uint8')
    
    img_b = cdf_final_b[b]
    img_g = cdf_final_g[g]
    img_r = cdf_final_r[r]
    
    img_out = cv2.merge((img_b,img_g,img_r))
    
    return img_out
    
img_cdf(dark,"dark_cdf")
img_cdf(bright,"bright_cdf")
img_cdf(gray,"gray_cdf")
img_cdf(normal,"normal_cdf")

dark_transform = histogram_equalization(dark)
bright_transform = histogram_equalization(bright)
gray_transform = histogram_equalization(gray)

plt.figure(figsize=(20,5))
plt.subplot(1,3,1), plt.title('dark_transform')
plt.imshow(dark_transform), plt.axis('off')
plt.subplot(1,3,2), plt.title('bright_transform')
plt.imshow(bright_transform), plt.axis('off')
plt.subplot(1,3,3), plt.title('gray_transform')
plt.imshow(gray_transform), plt.axis('off')
plt.show()

cv2.imwrite("dark_transform.jpg", dark_transform)
cv2.imwrite("bright_transform.jpg", bright_transform)
cv2.imwrite("gray_transform.jpg", gray_transform)



---

### Conclusion
比較power-low以及histogram equalization，power-low能夠決定的是整張圖片的亮度，只能同時變亮或變暗，所以在過暗和過亮兩張圖片，可以有稍微好的效果，但作用在灰階圖片上完全不行。

至於histogram equalization，因為可以將亮的部分變得更亮，暗的部分變得更暗，明暗區分出來，可以讓物件更明顯。