Building Custom Filters in Scipy for Noise Reduction in Signal and Image Processing

by

Filtering is a common technique used in signal and image processing to reduce noise and improve data quality. SciPy provides tools to create custom filters tailored to specific noise characteristics and processing needs. This article explains how to build and apply custom filters using SciPy for noise reduction tasks.

Understanding Noise and Filtering

Noise can originate from various sources, such as sensor imperfections or environmental interference. Filtering aims to suppress this unwanted information while preserving the essential features of the data. Different types of filters, like low-pass, high-pass, and band-pass, target specific frequency components.

Creating Custom Filters in SciPy

To build a custom filter, you typically define a filter kernel or transfer function that matches your noise profile. SciPy’s signal processing module offers functions such as scipy.signal.convolve for applying filters and scipy.signal.firwin for designing finite impulse response (FIR) filters.

Applying Filters to Data

Once the filter is designed, it can be applied to signals or images. For signals, convolution is used to filter the data. For images, 2D convolution applies the filter kernel across the image. SciPy functions facilitate these operations efficiently.

Example: Designing a Low-Pass Filter

Below is an example of creating a simple low-pass filter to reduce high-frequency noise in a signal.

Code Example:

“`python import numpy as np from scipy.signal import firwin, lfilter import matplotlib.pyplot as plt # Generate a noisy signal fs = 500 # Sampling frequency t = np.linspace(0, 1, fs, endpoint=False) signal_clean = np.sin(2 * np.pi * 5 * t) noise = 0.5 * np.random.randn(fs) signal_noisy = signal_clean + noise # Design a low-pass FIR filter numtaps = 51 cutoff = 10 # Cutoff frequency in Hz fir_coeff = firwin(numtaps, cutoff, fs=fs) # Apply the filter filtered_signal = lfilter(fir_coeff, 1.0, signal_noisy) # Plot the results plt.figure(figsize=(10, 6)) plt.plot(t, signal_noisy, label=’Noisy Signal’) plt.plot(t, filtered_signal, label=’Filtered Signal’, linewidth=2) plt.legend() plt.xlabel(‘Time [s]’) plt.ylabel(‘Amplitude’) plt.title(‘Noise Reduction Using Custom Low-Pass Filter’) plt.show() “`