### 计算图像或信号去噪处理前后的均方误差 (MSE) 对于计算图像或信号去噪处理前后的均方误差（Mean Squared Error, MSE），可以采用如下方法： #### 方法概述 均方误差是一种衡量两个数值集合差异程度的方法，在这里用于比较原始无噪声的数据与经过去噪处理之后的数据。其公式为: \[ \text{MSE} = \frac{1}{n}\sum_{i=1}^{n}(y_i-\hat{y}_i)^2 \] 其中 \( y_i \) 表示原始数据点，\( \hat{y}_i \) 是对应的去噪后数据点。 #### Python 实现代码 下面是使用Python编写的计算图像或信号去噪前后均方误差的代码实例： ```python import numpy as np def calculate_mse(original_signal, denoised_signal): """ Calculate Mean Squared Error between original and denoised signals/images. Parameters: original_signal : array_like Original signal/image data before denoising process. denoised_signal : array_like Signal/Image after applying the denoise algorithm. Returns: mse_value : float Computed mean squared error value. """ # Ensure both arrays have same shape if not original_signal.shape == denoised_signal.shape: raise ValueError('Both input shapes must match.') # Compute square difference of two vectors/matrices element-wise, # then take average over all elements to get final MSE score mse_value = ((original_signal - denoised_signal)**2).mean() return mse_value # Example usage with dummy data points representing an image or a signal if __name__ == "__main__": # Create some sample noisy 'image' data for demonstration purposes only img_size = (64, 64) clean_img = np.random.rand(*img_size) * 255 noise_level = 0.1 noise = np.random.normal(scale=noise_level*clean_img.max(), size=img_size) noisy_img = clean_img + noise # Simulate simple averaging filter as basic denoiser here just for example purpose kernel = np.ones((3, 3)) / 9 from scipy.signal import convolve2d denoised_img = convolve2d(noisy_img, kernel, mode='same', boundary='symm') # Now compute MSE comparing cleaned vs simulated-denoised images/signals mse_result = calculate_mse(clean_img, denoised_img) print(f"The computed Mean Squared Error is {mse_result:.4f}") ``` 此段代码展示了如何创建一个简单的函数`calculate_mse()`来接收两组相同尺寸的一维或多维数组作为输入，并返回它们之间均方误差的结果。此外还提供了一个具体的例子说明如何应用这个功能于二维图像上去评估一种简易平滑滤波器的效果[^1]。