A Study on Denoising of Poisson Noise in Pap Smear Microscopic Image

Abstract

Objectives: To study and implement various spatial domain denoising filters on various Pap Smear microscopic images. Methods/Statistical Analysis: In this paper, we calculated the performance metrics to estimate best fit filtering technique among the filters considered for this study. Study of around 300 microscopic Pap Smear images with low and high color densities are considered which includes images from two different standard slide preparation techniques i.e. Conventional Based Cytology (CBC) and Liquid Based Cytology (LBC). Performance of different filters is analyses based on performance metrics like PSNR, MSE, NAE and SNR. Findings: Poisson noise is more likely in Pap Smear images as it is captured using CCD camera at microscopic eye piece. From this study it was found that adaptive wiener filter can be used to reduce Poisson noise in both LBC and CBC Pap Smear Microscopic image. Application/Improvements: This helps to improve analysis result for pathologists using computer aided diagnosis system. Keywords: Denoising, Microscopic Image, Pap Smear, Performance Metrics, Poisson Noise