快速引导滤波python-opencv

ゞ浴缸里的玫瑰 2022-11-20 02:59 426阅读 0赞

# -*- coding: utf-8 -*-
    import cv2
    import numpy as np
    import sys
    
    
    def integral(image):
        rows,cols = image.shape
        #行积分运算
        inteImageC = np.zeros((rows,cols),np.float32)
        for r in range(rows):
            for c in range(cols):
                if c == 0:
                    inteImageC[r][c] = image[r][c]
                else:
                    inteImageC[r][c] = inteImageC[r][c-1] + image[r][c]
        #列积分运算
        inteImage = np.zeros(image.shape,np.float32)
        for c in range(cols):
            for r in range(rows):
                if r == 0:
                    inteImage[r][c] = inteImageC[r][c]
                else:
                    inteImage[r][c] = inteImage[r-1][c] + inteImageC[r][c]
        #为了在快速均值平滑使用中省去判断边界的问题
        #上边和左边进行补零
        inteImage_0 = np.zeros((rows+1,cols+1),np.float32)
        inteImage_0[1:rows+1,1:cols+1] = inteImage
        return inteImage_0
    #快速均值平滑：返回数组的数据类型是浮点型，winSize = ( 高，宽 )
    def fastMeanBlur(image,winSize,borderType = cv2.BORDER_DEFAULT):
        halfH = (winSize[0]-1)//8
        halfW = (winSize[1]-1)//8
        ratio = 1.0/(winSize[0]*winSize[1])
        #边缘扩充
        paddImage = cv2.copyMakeBorder(image,halfH,halfH,halfW,halfW,borderType)
        #图像积分
        paddIntegral = integral(paddImage)
        #图像的宽高
        rows,cols = image.shape
        #均值滤波后的结果
        meanBlurImage = np.zeros(image.shape,np.float32)
        r,c = 0,0
        for h in range(halfH,halfH+rows,1):
            for w in range(halfW,halfW+cols,1):
                meanBlurImage[r][c] = (paddIntegral[h+halfH+1][w+halfW+1] + paddIntegral[h-halfH][w-halfW]
                                        -paddIntegral[h+halfH+1][w-halfW]-paddIntegral[h-halfH][w+halfW+1])*ratio
                c+=1
            r+=1
            c=0
        return meanBlurImage
    #导向滤波
    def guidedFilter(I,p,winSize,eps):
        #输入图像的宽高
        rows,cols = I.shape
        # I 的均值平滑
        mean_I = fastMeanBlur(I,winSize,cv2.BORDER_DEFAULT)
        # p 的均值平滑
        mean_p = fastMeanBlur(p,winSize,cv2.BORDER_DEFAULT)
        # I.*p 的均值平滑
        Ip = I*p
        mean_Ip = fastMeanBlur(Ip,winSize,cv2.BORDER_DEFAULT)
        #协方差
        cov_Ip = mean_Ip - mean_I*mean_p
        mean_II = fastMeanBlur(I*I,winSize,cv2.BORDER_DEFAULT)
        #方差
        var_I = mean_II - mean_I*mean_I
        a = cov_Ip/(var_I+eps)
        b = mean_p - a*mean_I
        # 对 a 和 b进行均值平滑
        mean_a = fastMeanBlur(a,winSize,cv2.BORDER_DEFAULT)
        mean_b = fastMeanBlur(b,winSize,cv2.BORDER_DEFAULT)
        q = mean_a*I + mean_b
        return q
    #主函数
    if __name__ =="__main__":
        # if len(sys.argv) > 1:
        #     I = cv2.imread(sys.argv[1],cv2.CV_LOAD_IMAGE_COLOR)
        #     p = cv2.imread(sys.argv[2],cv2.CV_LOAD_IMAGE_GRAYSCALE)
        # else:
        #     print "Usge:python guidedFilter.py imageFile"
        I = cv2.imread("fog1.jpg")
        p = I
        #将图像归一化
        image_0_1 = I/255.0
        p = p/255.0
        #显示原图
        cv2.imshow("image_0_1",image_0_1)
        #导向滤波
        result = np.zeros(I.shape)
        result[:,:,0] = guidedFilter(image_0_1[:,:,0],image_0_1[:,:,0],(17,17),pow(0.2,2.0))
        result[:,:,1] = guidedFilter(image_0_1[:,:,1],image_0_1[:,:,1],(17,17),pow(0.2,2.0))
        result[:,:,2] = guidedFilter(image_0_1[:,:,2],image_0_1[:,:,2],(17,17),pow(0.2,2.0))
        cv2.imshow("guidedFilter",result)
        #保存导向滤波的结果
        result = result*255
        result[result>255] = 255
        result = np.round(result)
        result = result.astype(np.uint8)
        cv2.imwrite("guidedFilter.jpg",result)
        cv2.waitKey(0)
        cv2.destroyAllWindows()