ANALISIS KINERJA NOISE FILTERING UNTUK MENGHILANGKAN DERAU PADA CITRA SEDIAAN DARAH TIPIS MALARIA

Abstract

In automated malaria diagnosis systems, the images resulting from the acquisition process may
contain noise, which can obscure the visual appearance of the Plasmodium. Therefore, noise filtering
techniques are needed to remove noise from the images. This study evaluates the performance of noise
filtering techniques under various noise conditions to determine the most effective method for removing
noise in blood smear images.The research consists of several stages: acquisition of thin blood smear
images, noise addition, noise filtering, and performance comparison of the noise filtering methods. The
image acquisition was carried out using the publicly available MP-IDB malaria dataset. Noise addition was
performed using salt and pepper, Gaussian, and mixed noise. The tested noise filtering techniques include
median filtering, Gaussian filtering, and average filtering. The performance of the noise filtering methods
was assessed using Mean Squared Error (MSE) and Peak Signal-to-Noise Ratio (PSNR)
measurements.The experimental results show that for blood smear images with salt and pepper noise, the
median filtering method provided the best results with an MSE of 6.51 and a PSNR of 43.13. Visually, the
noise-removed image using median filtering closely resembled the original image. For images with
Gaussian noise, the average filtering method achieved the best performance with an MSE of 31.81 and a
PSNR of 33.19. For images with mixed noise, the median filtering method again provided the best
performance with an MSE of 46.95 and a PSNR of 31.46.