Cryptosporidium spp. and Giardia spp. Parasite Image with Distortion (Dataset)

This dataset of Cryptosporidium spp. and Giardia spp. was obtained from the Department of Parasitology, Faculty of Medicine, Universiti Malaya with collaboration with PROF. DR. YVONNE LIM AI LIAN (https://umexpert.um.edu.my/limailian). A total of 23 high-resolution reference images were collected.

To construct the training dataset, each reference image was subjected to five types of distortions: Gaussian White Noise (GWN), Gaussian Blur (GB), Salt and Pepper Noise (SnP), Speckle Noise, and JPEG Compression. For each distortion type, 9 levels were applied using the following controlling parameters:

Gaussian White Noise (GWN) — Standard deviation (σ): 10, 20, 30, 40, 50, 60, 70, 80, 90

Gaussian Blur (GB) — Kernel size (σ): 10, 20, 30, 40, 50, 60, 70, 80, 90

Salt and Pepper Noise (SnP) — Noise density (d): 0.01, 0.02, 0.03, 0.04, 0.05, 0.06, 0.07, 0.08, 0.09

Speckle Noise — Variance (σ²): 0.01, 0.02, 0.03, 0.04, 0.05, 0.06, 0.07, 0.08, 0.09

JPEG Compression — Quality factor (Q): 10, 15, 20, 25, 30, 35, 40, 45, 50

These distortion levels were applied uniformly across all reference images. The distorted images were further augmented to improve the robustness of the dataset for deep learning model training. In total, the training dataset consists of 1,058 images.

A separate test dataset containing 125 images was also created by applying different distortion levels not used in the training set to prevent image redundancy and allow unbiased evaluation.