Quality assessment for microscopic parasite images
Muhammad Amirul Aiman Asri, Norrima Binti Mokhtar, Heshalini Rajagopal, Wan Amirul Wan Mohd Mahiyiddin, Yvonne Ai Lian Lim, Masahiro Iwahashi, Anees ul Husnain · Proceedings of International Conference on Artificial Life and Robotics · 2023
Microscopy lens distortion will cause errors during parasites image acquisition for water sample inspection.Since water sample inspection is crucial for treated water monitoring, the quality of microscopic parasite images such as Giardia and Cryptosporidium need to be monitored as well to avoid errors in treated water inspection.In this work, the subjective and objective evaluation of parasite images were performed.The parasite species studied were Cryptosporidium and Giardia (oo)cysts.Parasite image database consisting of 20 reference images and 360 distorted images were used in the evaluation.The distorted images were generated from the reference images by applying distortion to the reference images with Gaussian White Noise and Motion Blur, at 9 levels of distortions.Twenty subjects were selected to assess the distorted images for the subjective evaluation.The scores obtained from the subjects were transformed into Mean Opinion Score (MOS).In the objective evaluation, six Full Reference-IQA (FR-IQA) metrics, namely MSSIM, SSIM, FSIM, IWSSIM, GMSD and VIF were used to evaluate the distorted images.The subjective MOS scores were used as the benchmark to determine the most suitable objective IQA to assess parasite images.The relationship between the subjective MOS and objective IQAs are examined using performance metrics namely PLCC and RMSE.It was found that MSSIM is the most suitable IQA to assess parasite images distorted with Gaussian White Noise and Motion Blur.