python實現圖像增強算法

本文實例為大傢分享瞭python實現圖像增強算法的具體代碼,供大傢參考,具體內容如下

圖像增強算法,圖像銳化算法

1)基於直方圖均衡化

2)基於拉普拉斯算子

3)基於對數變換

4)基於伽馬變換

5)  CLAHE

6)retinex-SSR

7)retinex-MSR

其中,基於拉普拉斯算子的圖像增強為利用空域卷積運算實現濾波
基於同一圖像對比增強效果
直方圖均衡化:對比度較低的圖像適合使用直方圖均衡化方法來增強圖像細節
拉普拉斯算子可以增強局部的圖像對比度
log對數變換對於整體對比度偏低並且灰度值偏低的圖像增強效果較好
伽馬變換對於圖像對比度偏低,並且整體亮度值偏高(對於相機過曝)情況下的圖像增強效果明顯

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


# 直方圖均衡增強
def hist(image):
    r, g, b = cv2.split(image)
    r1 = cv2.equalizeHist(r)
    g1 = cv2.equalizeHist(g)
    b1 = cv2.equalizeHist(b)
    image_equal_clo = cv2.merge([r1, g1, b1])
    return image_equal_clo


# 拉普拉斯算子
def laplacian(image):
    kernel = np.array([[0, -1, 0], [-1, 5, -1], [0, -1, 0]])
    image_lap = cv2.filter2D(image, cv2.CV_8UC3, kernel)
   # cv2.imwrite('th1.jpg', image_lap)
    return image_lap


# 對數變換
def log(image):
    image_log = np.uint8(np.log(np.array(image) + 1))
    cv2.normalize(image_log, image_log, 0, 255, cv2.NORM_MINMAX)
    # 轉換成8bit圖像顯示
    cv2.convertScaleAbs(image_log, image_log)
    return image_log


# 伽馬變換
def gamma(image):
    fgamma = 2
    image_gamma = np.uint8(np.power((np.array(image) / 255.0), fgamma) * 255.0)
    cv2.normalize(image_gamma, image_gamma, 0, 255, cv2.NORM_MINMAX)
    cv2.convertScaleAbs(image_gamma, image_gamma)
    return image_gamma


# 限制對比度自適應直方圖均衡化CLAHE
def clahe(image):
    b, g, r = cv2.split(image)
    clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
    b = clahe.apply(b)
    g = clahe.apply(g)
    r = clahe.apply(r)
    image_clahe = cv2.merge([b, g, r])
    return image_clahe


def replaceZeroes(data):
    min_nonzero = min(data[np.nonzero(data)])
    data[data == 0] = min_nonzero
    return data


# retinex SSR
def SSR(src_img, size):
    L_blur = cv2.GaussianBlur(src_img, (size, size), 0)
    img = replaceZeroes(src_img)
    L_blur = replaceZeroes(L_blur)

    dst_Img = cv2.log(img/255.0)
    dst_Lblur = cv2.log(L_blur/255.0)
    dst_IxL = cv2.multiply(dst_Img, dst_Lblur)
    log_R = cv2.subtract(dst_Img, dst_IxL)

    dst_R = cv2.normalize(log_R,None, 0, 255, cv2.NORM_MINMAX)
    log_uint8 = cv2.convertScaleAbs(dst_R)
    return log_uint8


def SSR_image(image):
    size = 3
    b_gray, g_gray, r_gray = cv2.split(image)
    b_gray = SSR(b_gray, size)
    g_gray = SSR(g_gray, size)
    r_gray = SSR(r_gray, size)
    result = cv2.merge([b_gray, g_gray, r_gray])
    return result


# retinex MMR
def MSR(img, scales):
    weight = 1 / 3.0
    scales_size = len(scales)
    h, w = img.shape[:2]
    log_R = np.zeros((h, w), dtype=np.float32)

    for i in range(scales_size):
        img = replaceZeroes(img)
        L_blur = cv2.GaussianBlur(img, (scales[i], scales[i]), 0)
        L_blur = replaceZeroes(L_blur)
        dst_Img = cv2.log(img/255.0)
        dst_Lblur = cv2.log(L_blur/255.0)
        dst_Ixl = cv2.multiply(dst_Img, dst_Lblur)
        log_R += weight * cv2.subtract(dst_Img, dst_Ixl)

    dst_R = cv2.normalize(log_R,None, 0, 255, cv2.NORM_MINMAX)
    log_uint8 = cv2.convertScaleAbs(dst_R)
    return log_uint8


def MSR_image(image):
    scales = [15, 101, 301]  # [3,5,9]
    b_gray, g_gray, r_gray = cv2.split(image)
    b_gray = MSR(b_gray, scales)
    g_gray = MSR(g_gray, scales)
    r_gray = MSR(r_gray, scales)
    result = cv2.merge([b_gray, g_gray, r_gray])
    return result


if __name__ == "__main__":
    image = cv2.imread('img/FJ(93).png')
    image_gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

    plt.subplot(4, 2, 1)
    plt.imshow(image)
    plt.axis('off')
    plt.title('Offical')

    # 直方圖均衡增強
    image_equal_clo = hist(image)

    plt.subplot(4, 2, 2)
    plt.imshow(image_equal_clo)
    plt.axis('off')
    plt.title('equal_enhance')

    # 拉普拉斯算法增強
    image_lap = laplacian(image)
    plt.subplot(4, 2, 3)
    plt.imshow(image_lap)
    plt.axis('off')
    plt.title('laplacian_enhance')

    # LoG對象算法增強
    image_log = log(image)

    plt.subplot(4, 2, 4)
    plt.imshow(image_log)
    plt.axis('off')
    plt.title('log_enhance')

    # # 伽馬變換
    image_gamma = gamma(image)

    plt.subplot(4, 2, 5)
    plt.imshow(image_gamma)
    plt.axis('off')
    plt.title('gamma_enhance')

    # CLAHE
    image_clahe = clahe(image)

    plt.subplot(4, 2, 6)
    plt.imshow(image_clahe)
    plt.axis('off')
    plt.title('CLAHE')

    # retinex_ssr
    image_ssr = SSR_image(image)

    plt.subplot(4, 2, 7)
    plt.imshow(image_ssr)
    plt.axis('off')
    plt.title('SSR')

    # retinex_msr
    image_msr = MSR_image(image)

    plt.subplot(4, 2, 8)
    plt.imshow(image_msr)
    plt.axis('off')
    plt.title('MSR')

    plt.show()

以上就是本文的全部內容,希望對大傢的學習有所幫助,也希望大傢多多支持WalkonNet。

推薦閱讀: