圖像統(tǒng)計特征

熵

用于度量圖像所具有的信息量，它反映了圖像中紋理的紊亂度或復(fù)雜程度。熵值越大，說明紋理越復(fù)雜；熵值越小，說明紋理越平滑。

均值

灰度均值是對區(qū)域內(nèi)亮度的一個度量，可以用來反應(yīng)圖像的明暗程度。

方差

方差就是數(shù)據(jù)的分散程度（偏離均值）。

對比度

反映了圖像的清晰度和紋理溝紋深淺的程度。紋理溝紋越深，其對比度越大，視覺效果越清晰；反之，對比度小，則溝紋淺，效果模糊。

公式

測試圖片-從均值、方差做對比

對于第一組六張圖片，圖片越輛均值越大，但方差不會改變（同一張圖片只改變亮度）
對于第二組圖片，對比度不做調(diào)節(jié)，只調(diào)節(jié)亮度的情況下，偏離均值的離散度會變大

測試圖片-從對比度出發(fā)

對比度越大 視覺效果越清晰；紋理越深對比度越大

測試圖片-從熵出發(fā)

熵值越大，說明紋理越復(fù)雜；熵值越小，說明紋理越平滑。

文章來源地址http://www.zghlxwxcb.cn/news/detail-593822.html

img_1 = cv2.imread('office_1.jpg')
img_2 = cv2.imread('office_2.jpg')
img_3 = cv2.imread('office_3.jpg')
img_4 = cv2.imread('office_4.jpg')
img_5 = cv2.imread('office_5.jpg')
img_6 = cv2.imread('office_6.jpg')
img_7 = cv2.imread('dalishi.jpg')
img_8 = cv2.imread('wall.jpg')
img_9 = cv2.imread('muwen.jpg')
img_10 = cv2.imread('shuiniwen.jpg')
img_11 = cv2.imread('400x400_zhi_wenli.jpg')
img_12 = cv2.imread('400x400_zhi_zhezhouwenli.jpg')
imgs = [img_1,img_2,img_3,img_4,img_5,img_6]
titles = ['office_1.jpg','office_2.jpg','office_3.jpg','office_4.jpg','office_5.jpg','office_6.jpg']
for m in range(6):
    plt.subplot(2,3,m + 1)
    plt.imshow(imgs[m])
    plt.title(titles[m])
def rgb2gray(img):
    h = img.shape[0]
    w = img.shape[1]
    gray = np.uint8(np.zeros((h,w)))
    for i in range(h):
        for j in range(w):
            gray[i,j] = 0.144 * img[i,j,0] + 0.587 * img[i,j,1] + 0.299 * img[i,j,2]  # BGR
    return gray
def average(img):
    img1 = rgb2gray(img)
    height,width = img1.shape
    size = img1.size
    ave = 0
    for i in range(height):
        for j in range(width):
            ave += img1[i][j] / size
    return ave
def contrast(img):
    img1 = rgb2gray(img)
    m,n = img1.shape
    # 圖片矩陣向外擴展一個像素
    img1_ext=cv2.copyMakeBorder(img1,1,1,1,1,cv2.BORDER_REPLICATE)  # 用邊界顏色填充
    height,width = img1_ext.shape
    b = 0.0
    for i in range(1,height - 1):
        for j in range(1,width - 1):
            b += (int((img1_ext[i,j]) - int(img1_ext[i,j + 1])) ** 2 + (
                    int(img1_ext[i,j]) - int(img1_ext[i,j - 1])) ** 2 + (
                          int(img1_ext[i,j]) - int(img1_ext[i + 1,j])) ** 2 + (
                          int(img1_ext[i,j]) - int(img1_ext[i - 1,j])) ** 2)
    cg = b / (4 * (m - 2) * (n - 2) + 3 * (2 * (m - 2) + 2 * (n - 2)) + 2 * 4)  #
    return cg
def variance(img):
    img1 = rgb2gray(img)
    height,width = img1.shape
    var = 0
    size = img1.size
    average = 0
    for i in range(height):
        for j in range(width):
            average += img1[i][j] / size
    for i in range(height):
        for j in range(width):
            var += img1[i,j] * (i - average) ** 2
    return var
def Contrast_and_Brightness(alpha,bete,img):  
    blank = np.zeros(img.shape,img.dtype)
    dst = cv2.addWeighted(img,alpha,blank,1 - alpha,bete)
    return dst
def dec2bin(p):
    floatbinstr = ""
    if p == 0:
        return floatbinstr
    for kk in range(len(str(p)) - 2):
        p *= 2
        if p > 1:
            floatbinstr += "1"
            p = p - int(p)
        else:
            floatbinstr += "0"
        if p == 0:
            break
        return str(floatbinstr)
def total_entropy(img):
    n = []
    P = []
    lenavg = []
    avg_sum = 0
    grey_lvl = 0
    k = 0
    res = 0
    # test = [[5,4,3,2,1]]
    weight = img.shape[0]
    height = img.shape[1]
    total = weight * height
    for i in range(256):
        n.append(0)
    for i in range(weight):
        for j in range(height):
            grey_lvl = img[i][j]
            n[grey_lvl] = float(n[grey_lvl] + 1)
            k = float(k + 1)
    for i in range(256):
        P.append(0)
    P = n
    for i in range(len(n)):
        P[i] = (n[i] / k)
    for i in range(256):
        lenavg.append(0)
    lenavg = P
    for i in range(len(n)):
        if P[i] == 0.0:
            continue
        lenavg[i] = lenavg[i] * len(dec2bin(lenavg[i]))
        avg_sum = lenavg[i] + avg_sum
    for i in range(len(n)):
        if (P[i] == 0):
            res = res
        else:
            res = float(res - P[i] * (math.log(P[i]) / math.log(2.0)))
    return res
if __name__ == '__main__':
    if input(keyboard.wait('A')):
        for i in range(6):
            ave_1 = average(imgs[i])
            ave_1 = Decimal(ave_1).quantize(Decimal("0.000"))
            print("average_office" + "_" + "123456"[i],ave_1)
            with open('E:\\untitled12\\image_practice\\entropy.txt','a',encoding='utf-8') as f:
                f.write('{:^30}\n'.format(str(ave_1)))
                f.close()
        plt.show()
    if input(keyboard.wait('V')):
        for i in range(6):
            var_1 = variance(img_1)
            var_1 = Decimal(var_1).quantize(Decimal("0.000"))
            print("variance_office" + "_" + "123456"[i],var_1)
        img_contrast = Contrast_and_Brightness(2.0,0,img_1)
        var_orign = variance(img_1)
        var_con = variance(img_contrast)
        print("原圖：       ",var_orign)
        print("原圖調(diào)節(jié)對比：",var_con)
        plt.subplot(1,2,1),plt.imshow(img_1),plt.title('img_1')   plt.subplot(1,2,2),plt.imshow(img_contrast),plt.title('img_contrast')
    plt.show()
    if input(keyboard.wait('C')):
        con_1 = contrast(img_1)
        plt.subplot(2,2,1),plt.imshow(img_1),plt.title('bangong')

        img_GAUSS = cv2.GaussianBlur(img_1,(9,9),0)
        con_2 = contrast(img_GAUSS)
        plt.subplot(2,2,2),plt.imshow(img_GAUSS),plt.title('bangong_GUSS')
        con_3 = contrast(img_9)
        plt.subplot(2,2,3),plt.imshow(img_9),plt.title('mu_wen')
        con_4 = contrast(img_10)
        plt.subplot(2,2,4),plt.imshow(img_10),plt.title('shui_ni_wen')

        con_1 = Decimal(con_1).quantize(Decimal("0.000"))
        con_2 = Decimal(con_2).quantize(Decimal("0.000"))
        print("contrast_office" + "_oringel",con_1)
        print("contrast_office" + "_GAUSS  ",con_2)

        con_3 = Decimal(con_3).quantize(Decimal("0.000"))
        con_4 = Decimal(con_4).quantize(Decimal("0.000"))
        print("木紋  ",con_3)
        print("水泥紋",con_4)
    plt.show()

    if input(keyboard.wait('E')):
        img_grey = cv2.imread('dalishi.jpg',cv2.IMREAD_GRAYSCALE)
        img_grey1 = cv2.imread('wall.jpg',cv2.IMREAD_GRAYSCALE)
        img_grey2 = cv2.imread('muwen.jpg',cv2.IMREAD_GRAYSCALE)

        img_wenli = cv2.imread('400x400_zhi_wenli.jpg',cv2.IMREAD_GRAYSCALE)
        img_wenli1 = cv2.imread('400x400_zhi_zhezhouwenli.jpg',cv2.IMREAD_GRAYSCALE)

        ent_dalishi = total_entropy(img_grey)
        plt.subplot(1,3,1),plt.imshow(img_7)
        ent_wall = total_entropy(img_grey1)
        plt.subplot(1,3,2),plt.imshow(img_8)
        ent_muwen = total_entropy(img_grey2)
        plt.subplot(1,3,3),plt.imshow(img_9)

        ent_wenli = total_entropy(img_wenli)
        plt.subplot(1,2,1),plt.imshow(img_11),plt.title('junyun_weli')
        ent_wenli1 = total_entropy(img_wenli1)
        plt.subplot(1,2,2),plt.imshow(img_12),plt.title('not_junrun_wen')

        print("entropy_dalishi",ent_dalishi)
        print("entropy_wall   ",ent_wall)
        print("entrop_muwen   ",ent_muwen)

        print("E_junrun_wenli:        ",ent_wenli)
        print("E_NOT_junrun_wenli1:   ",ent_wenli1)
    plt.show()

到了這里，關(guān)于從圖像統(tǒng)計特征看圖像內(nèi)在-均值，方差，對比度，熵的文章就介紹完了。如果您還想了解更多內(nèi)容，請在右上角搜索TOY模板網(wǎng)以前的文章或繼續(xù)瀏覽下面的相關(guān)文章，希望大家以后多多支持TOY模板網(wǎng)！