low_light.py

import imutils
import cv2
import numpy as np
import cv2
import scipy.linalg as la
import scipy.sparse.linalg as sla
from scipy.signal import convolve2d
import skimage.color as sc


# import matplotlib.pyplot as plt
# %matplotlib inline


def conv2(x, y, mode='same'):
    return np.rot90(convolve2d(np.rot90(x, 2), np.rot90(y, 2), mode=mode), 2)


def gauss2d(shape=(15, 15), sigma=0.5):
    m, n = [(ss - 1.) / 2. for ss in shape]
    y, x = np.ogrid[-m:m + 1, -n:n + 1]
    h = np.exp(-(x * x + y * y) / (2. * sigma * sigma))
    h[h < np.finfo(h.dtype).eps * h.max()] = 0
    sumh = h.sum()
    if sumh != 0:
        h /= sumh
    return h


def low_image_enhancement(im):
    imhsv = sc.rgb2hsv(im)
    row, col, ch = np.shape(im)
    h1 = imhsv[:, :, 0]
    s1 = imhsv[:, :, 1]
    v1 = imhsv[:, :, 2]

    H1 = gauss2d(sigma=15)
    H2 = gauss2d(sigma=80)
    H3 = gauss2d(sigma=250)

    img1 = conv2(v1, H1)
    img2 = conv2(v1, H2)
    img3 = conv2(v1, H3)

    imgguas = (1 / 3 * img1) + (1 / 3 * img2) + (1 / 3 * img3)

    alpha5 = .1
    k5 = alpha5 * np.sum(s1.flatten()) / (row * col)
    vnew5 = (v1 * (1 + k5)) / (np.maximum(v1, imgguas) + k5)

    X1 = v1.flatten()
    X2 = vnew5.flatten()
    C = np.cov(X1, X2)
    D, V = np.linalg.eig(C)
    diagD = D
    V1 = np.zeros(np.size(V))

    if (diagD[0] > diagD[1]):
        V1 = V[:, 0]
    else:
        V1 = V[:, 1]

    w1 = V1[0] / (V1[0] + V1[1])
    w2 = V1[1] / (V1[0] + V1[1])

    F = (w1 * v1) + (w2 * vnew5)
    out1 = np.zeros((row, col, ch))
    out1[:, :, 0] = h1
    out1[:, :, 1] = s1
    out1[:, :, 2] = F

    out1 = sc.hsv2rgb(out1)

    out1 = np.uint8(255 * out1 / np.max(np.ravel(out1)))

    return out1

# path=r'E:\MAIN PROJECT\FIGURES\lowlight.jpg'
# ite = cv2.imread(path)
# out=low_image_enhancement(ite)
# cv2.imshow('image',out)
# cv2.imwrite(r'E:\MAIN PROJECT\OUTPUT\out1.jpg',out)
# cv2.waitKey(0)