(2021)matlab代码重改python代码，对应函数——全！

@前言

因工程需要将matlab代码重构成pyhton代码，以下是个人历程，希望对后来者有所帮助。有错误麻烦积极指正，我将不胜感激。

1. matlab 函数 对应python基本函数文档下载（全）：猛点免费下载
2. matlab 对应 numpy：https://numpy.org/doc/stable/user/numpy-for-matlab-users.html#table-of-rough-matlab-numpy-equivalents

• matla和pyhton 图像读取差别

  cv2读入的形式为B G R
  matlab读入的形式为R G B
  

• matlab 加载mat文件load('file.mat');

  import scipy.io as scio
  
  load_mat = scio.loadmat("file.mat")
  变量a = load_mat["变量a"]
  变量b = load_mat["变量b"]
  

• matlab 变换矩阵形状reshape(A,[row col])

  A.reshape((row, col), order='F')
  

• matlab 复制矩阵行列repmat(A ,[row col])

  np.tile(A ,[row col])
  

• matlab 拼接矩阵[a(:) b(:) c(:)....n(:)]

  np.c_[a.flatten(), b.flatten(), c.flatten(), d..flatten()]
  

• conv2(gray_temp1, Gx, 'same');

  scipy.ndimage.filters.convolve1d(gray_temp1, Gx, axis=1, mode='constant')
  

• circshift(MSCN,[1 0]);

  np.roll(MSCN, 1, axis=0)
  

• mrdivide(A,B);

  # A * B逆
  np.matmul(A, np.linalg.inv(B))
  

• mtimes(A, B);

  # A * B
  np.matmul(A, B)
  

• length(A);

  max(A.shape)
  

• rgb2hsv

  def rgb2hsv(img):
      """
      :param img: 彩色图片
      :return: H、S、V 三个通道
      """
      HSV = cv2.cvtColor(img, cv2.COLOR_RGB2HSV)
      H, S, V = cv2.split(HSV)
      return H, S, V
  

• // matlab 对图像进行高斯滤波
MSCN_window = fspecial('gaussian',7,7/6);
mu = imfilter(Ig,MSCN_window,'replicate');

 def gaussian_filter(img, Hsize=(3, 3), sigma=0.5, boedertype=cv2.BORDER_REPLICATE):
     """
         对图像进行高斯滤波
     :param img: 待滤波图像
     :param Hsize: 核的尺寸
     :param sigma: 标准差
     :param boedertype: 边界填充模式
     :return: 滤波后的图像
     """
     r, c = Hsize
     gaussian_kernel = np.multiply(cv2.getGaussianKernel(r, sigma), (cv2.getGaussianKernel(c, sigma)).T)
     
     filter_img = cv2.filter2D(img, -1, gaussian_kernel, borderType=boedertype)
     return filter_img

• im2col(MSCN, [ps ps], 'distinct');

def im2col(mtx, block_size):
    mtx_shape = mtx.shape
    if mtx_shape[0] % block_size[0] != 0:
        sx = mtx_shape[0] // block_size[0] + 1
    else:
        sx = mtx_shape[0] // block_size[0]

    if mtx_shape[1] % block_size[1] != 0:
        sy = mtx_shape[1] // block_size[1] + 1
    else:
        sy = mtx_shape[1] // block_size[1]

    result = np.empty((block_size[0] * block_size[1], sx * sy))
    
    for i in range(sy):
        for j in range(sx):

            Zeros = np.zeros((block_size[0], block_size[1]))
            block = mtx[j * block_size[0]: j * block_size[0] + block_size[0], i * block_size[1]: i * block_size[1] + block_size[1]]

            if block.shape[0] < Zeros.shape[0] and block.shape[1] == Zeros.shape[1]:
                concat = np.zeros((Zeros.shape[0] - block.shape[0], Zeros.shape[1]))
                block = np.vstack((block, concat))

            if block.shape[1] < Zeros.shape[1] and block.shape[0] == Zeros.shape[0]:
                concat = np.zeros((Zeros.shape[0], Zeros.shape[1] - block.shape[1]))
                block = np.hstack((block, concat))

            if block.shape[0] < Zeros.shape[0] and block.shape[1] < Zeros.shape[1]:
                concat = np.zeros((block_size[0], block_size[1]))
                concat[:block.shape[0], :block.shape[1]] = block
                block = concat

            for line in range(block.shape[0]):
                Zeros[line, :] = block[line, :]
            result[:, i * sx + j] = Zeros.ravel(order='F')
    return result

• entropy

  def entropy(img):
      rows, cols = img.shape
      entropy_list = []
      for col in range(cols):
  
          hist, _ = np.histogram(img[:, col], 256, [0, 256])
          noZero_list = []
          index = np.where(hist != 0)[0]
          for idx in index:
              noZero_list.append(hist[idx])
          noZero_array = np.array(noZero_list, dtype=np.float32)
          noZero_array = noZero_array / rows
          E = -np.sum(noZero_array * np.log2(noZero_array), axis=0)
          entropy_list.append(E)
  
      return np.array(entropy_list, dtype=np.float32)