ENHANCING CERVICAL CANCER IMAGES QUALITY: HYBRID SMO-PMD FILTER FOR NOISE REDUCTION

Abstract

This study presents an image denoising method for cervical cancer images using the Perona–Malik Diffusion (PMD) filter optimized with the Spider Monkey Optimization (SMO) algorithm. The BRISQUE is proposed as the new objective function. The method was simulated on three datasets: SIPaKMeD, Herlev, and Mendeley Liquid-Based Cytology (LBC). Enhanced image quality was evaluated using MSE, SSIM, PSNR, and Entropy. On the SIPaKMeD dataset, the SMO-PMD filter achieved an average MSE of 0.0454, SSIM of 0.9984, PSNR of 62.27 dB, and Entropy of 5.425. The Mendeley dataset recorded an MSE of 0.3991, SSIM of 0.9994, PSNR of 53.08 dB, and Entropy of 5.489. The Herlev dataset achieved an MSE of 8.1191, SSIM of 0.9688, PSNR of 55.77 dB, and Entropy of 5.203. The SMO algorithm was compared with Particle Swarm Optimization (PSO) and Genetic Algorithm (GA). SMO showed better results across all metrics. The proposed method produces images with lower noise, higher structural similarity, and improved visual quality. The stable entropy values across the datasets indicate that essential diagnostic information was preserved. These findings provide a new perspective for enhancing cervical cancer images using a hybrid SMO-PMD filter. A limitation of this study is that experiments were limited to three datasets, and SMO’s reliance on extreme κ values might reduce stability in other contexts