【5】python-opencv3教程：图像平滑2（联合双边滤波平滑）

快来打我* 2022-03-16 05:20 306阅读 0赞

# 第五节：图像平滑2 #

### **一：联合双边滤波** ###

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3NoYXdyb2FkODg_size_16_color_FFFFFF_t_70][]

代码实现：

import numpy as np
    import cv2
    import math
    
    
    def getClosenessWeight(sigma_g, H, W):
        # 计算空间距离权重模板
        r, c = np.mgrid[0:H:1, 0:W:1]  # 构造三维表
        r -= int((H-1) / 2)
        c -= int((W-1) / 2)
        closeWeight = np.exp(-0.5*(np.power(r, 2)+np.power(c, 2))/math.pow(sigma_g, 2))
        return closeWeight
    
    def jointBLF(I, H, W, sigma_g, sigma_d, borderType=cv2.BORDER_DEFAULT):
    
        # 构建空间距离权重模板
        closenessWeight = getClosenessWeight(sigma_g, H, W)
    
        # 对I进行高斯平滑
        Ig = cv2.GaussianBlur(I, (W, H), sigma_g)
    
        # 模板的中心点位置
        cH = int((H - 1) / 2)
        cW = int((W - 1) / 2)
    
        # 对原图和高斯平滑的结果扩充边界
        Ip = cv2.copyMakeBorder(I, cH, cH, cW, cW, borderType)
        Igp = cv2.copyMakeBorder(Ig, cH, cH, cW, cW, borderType)
    
        # 图像矩阵的行数和列数
        rows, cols = I.shape
        i, j = 0, 0
    
        # 联合双边滤波的结果
        jblf = np.zeros(I.shape, np.float64)
        for r in range(cH, cH+rows, 1):
            for c in range(cW, cW+cols, 1):
                # 当前位置的值
                pixel = Igp[r][c]
    
                # 当前位置的邻域
                rTop, rBottom = r-cH, r+cH
                cLeft, cRight = c-cW, c+cW
    
                # 从 Igp 中截取该邻域，用于构建相似性权重模板
                region = Igp[rTop: rBottom+1, cLeft: cRight+1]
    
                # 通过上述邻域，构建该位置的相似性权重模板
                similarityWeight = np.exp(-0.5*np.power(region-pixel, 2.0)) / math.pow(sigma_d, 2.0)
    
                # 相似性权重模板和空间距离权重模板相乘
                weight = closenessWeight * similarityWeight
    
                # 将权重归一化
                weight = weight / np.sum(weight)
    
                # 权重模板和邻域对应位置相乘并求和
                jblf[i][j] = np.sum(Ip[rTop:rBottom+1, cLeft:cRight+1]*weight)
    
                j+=1
            j = 0
            i += 1
        return jblf
    
    if __name__ == '__main__':
        I = cv2.imread('p2.jpg', cv2.IMREAD_GRAYSCALE)
        # 将8位图转换为浮点型
        fI = I.astype(np.float64)
    
        # 联合双边滤波，返回值的数据类型为浮点型
        jblf = jointBLF(fI, 33, 33, 7, 2)
        jblf = np.round(jblf)
        jblf = jblf.astype(np.uint8)
    
        cv2.imshow('origin', I)
        cv2.imshow('jblf', jblf)
        cv2.waitKey(0)
        cv2.destroyAllWindows()

输出结果：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3NoYXdyb2FkODg_size_16_color_FFFFFF_t_70 1][]

###    ###

### 图像平滑就这些算法  。。随后学到新方法再继续更新    ###

###                                                                               下一节我们学习阈值分割。。。。。 ###

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3NoYXdyb2FkODg_size_16_color_FFFFFF_t_70]: /images/20220316/c4d535af551341ea98ca8b60c5a3e359.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3NoYXdyb2FkODg_size_16_color_FFFFFF_t_70 1]: /images/20220316/ccfa0ea5b957461094ad00b7f961dacb.png