Efficient Noise Removal in Palmprint Images Using Various Filters in a Machine-Learning Approach

Abstract

A biological identification technique, palm print identification, takes advantage of the distinctive patterns on a person's palm for authentication. It falls under the broader category of biometrics, which deals with evaluating and statistically assessing each individual's distinctive personality characteristics. The efficiency of three well-known noise-removal methods the non-local mean (NLM) filter, Wiener filter, and median filter when utilized on palmprint images are examined in the present research. Peak signal-to-noise ratio (PSNR), mean squared error (MSE), and structural similarity index measure (SSIM) were used to evaluate the performance. The objective is to identify the best technique for reducing noise in palmprint photos without compromising important details. NLM filter beat the Wiener and Median filters by producing an MSE of 0.000143, PSNR of 41.79, and SSIM of 0.998, respectively and also the tool used for executing Jupyter Notebook and the language used is Python. Regarding the various types of noises frequently present in palmprint photos, the NLM filter demonstrated superior noise reduction abilities. The NLM filter successfully improved image quality while maintaining the images' structure.