Datasets
Standard Dataset
NMNIST (random noise and impulse noise)
- Citation Author(s):
- Submitted by:
- Qixuan Li
- Last updated:
- Thu, 04/10/2025 - 17:52
- DOI:
- 10.21227/0cbz-n616
- Data Format:
- License:
23 Views- Categories:
- Keywords:
Abstract
This dataset consists of images with two types of artificially added noise, intended for evaluating the robustness of machine learning models against noise perturbations. The first type of noise introduces randomly generated pixel values ranging from 0 to 255 at random positions in the image. The second type of noise adds binary noise by setting pixels at random locations to either 0 or 255. The dataset includes images with varying amounts of noisy pixels, allowing for detailed analysis under different noise intensities. All images were generated using MATLAB, and the dataset is provided in a format that can be directly used within MATLAB environments for further simulation and testing.
The dataset includes two .mat files that can be directly loaded and used in MATLAB: RandomNoiseImage.mat and ImpulseNoiseImage.mat. Each file is organized as a cell array containing multiple image sets with varying noise levels.
-
RandomNoiseImage.mat contains 6 layers (cells), where each layer represents a collection of images with a specific number of noisy pixels. The layers correspond to images with 50, 100, 150, 200, 250, and 300 randomly placed noisy pixels, respectively. The noise values are random integers between 0 and 255, added at random positions in the images.
-
ImpulseNoiseImage.mat contains 9 layers (cells), each corresponding to a different impulse noise level applied to the images. The layers represent noise levels of 0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, and 0.5, where the noise is applied by randomly setting pixel values to either 0 or 255 (simulating salt-and-pepper noise).
This dataset is intended for evaluating the noise robustness of image processing or machine learning models, and can be readily used in MATLAB environments for training, testing, and analysis.
