How to set the best value for gamma correction

Question

I am trying to use the gamma correction to the image. But I only change values of gamma correction manually. Is there any way, to calculate the best value for gamma correction automaticly? E.g. with brightness histogram.

Code:

# import the necessary packages
from __future__ import print_function
import numpy as np
import argparse
import cv2
def adjust_gamma(image, gamma=1.0):
    # build a lookup table mapping the pixel values [0, 255] to
    # their adjusted gamma values
    invGamma = 1.0 / gamma
    table = np.array([((i / 255.0) ** invGamma) * 255
        for i in np.arange(0, 256)]).astype("uint8")
    # apply gamma correction using the lookup table
    return cv2.LUT(image, table)


# load the original image
original = cv2.imread('image.jpg')

# loop over various values of gamma
for gamma in np.arange(0.0, 3.5, 0.5):
    # ignore when gamma is 1 (there will be no change to the image)
    if gamma == 1:
        continue
    # apply gamma correction and show the images
    gamma = gamma if gamma > 0 else 0.1
    adjusted = adjust_gamma(original, gamma=gamma)
    cv2.putText(adjusted, "g={}".format(gamma), (10, 30),
        cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 0, 255), 3)
    cv2.imshow("Images", np.hstack([original, adjusted]))
    cv2.waitKey(0)

user_7355741 · Answer

Here are two ways to do that in Python/OpenCV. Both are based upon the ratio of the log(mid-gray)/log(mean). Results are often reasonable, especially for dark image, but do not work in all cases. For bright image, invert the gray or value image, process as for dark images, then invert again and recombine if using the value image.

Read the input
Convert to gray or HSV value
Compute the ratio log(mid-gray)/log(mean) on the gray or value channel
Raise the input or value to the power of the ratio
If using the value channel, combine the new value channel with the hue and saturation channels and convert back to RGB

Input:

import cv2
import numpy as np
import math

# read image
img = cv2.imread('lioncuddle1.jpg')

# METHOD 1: RGB

# convert img to gray
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# compute gamma = log(mid*255)/log(mean)
mid = 0.5
mean = np.mean(gray)
gamma = math.log(mid*255)/math.log(mean)
print(gamma)

# do gamma correction
img_gamma1 = np.power(img, gamma).clip(0,255).astype(np.uint8)



# METHOD 2: HSV (or other color spaces)

# convert img to HSV
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
hue, sat, val = cv2.split(hsv)

# compute gamma = log(mid*255)/log(mean)
mid = 0.5
mean = np.mean(val)
gamma = math.log(mid*255)/math.log(mean)
print(gamma)

# do gamma correction on value channel
val_gamma = np.power(val, gamma).clip(0,255).astype(np.uint8)

# combine new value channel with original hue and sat channels
hsv_gamma = cv2.merge([hue, sat, val_gamma])
img_gamma2 = cv2.cvtColor(hsv_gamma, cv2.COLOR_HSV2BGR)

# show results
cv2.imshow('input', img)
cv2.imshow('result1', img_gamma1)
cv2.imshow('result2', img_gamma2)
cv2.waitKey(0)
cv2.destroyAllWindows()

# save results
cv2.imwrite('lioncuddle1_gamma1.jpg', img_gamma1)
cv2.imwrite('lioncuddle1_gamma2.jpg', img_gamma2)

Result from method 1:

Result from method 2:

How to set the best value for gamma correction

1 Answers