Transmission Electron Microscopy Images with Motion Blur for Deblurring Models

2048 x 2048 pixel high-quality images of kidney tissue and a calibration grid, paired with 1024x1024 motion-blurred counterparts, were collected using a MiniTEM microscope. The dataset, created by Håkan Wieslander of Uppsala University, includes raw files and a partitioned set for training, validation, and testing. Five registered low-quality images are cropped and aligned to each high-quality image.

Use Cases

• Train motion deblurring models based on paired low-quality and high-quality image sequences.
• Benchmark image registration algorithms based on the alignment of five low-quality frames to each high-quality frame.
• Develop super-resolution techniques based on the availability of high-resolution (GT_hr) and downsampled ground truth images.

Strengths

• Provides precisely paired high-quality (2048x2048) and motion-blurred (1024x1024) image sets, enabling supervised learning.
• Includes a predefined train/validation/test split derived from four imaging sequences per sample.
• Images are stored as 16-bit TIFF files, preserving high dynamic range for detailed analysis.

Limitations

• Row count and total dataset size are unknown, which may limit suitability assessment for large-scale training.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

Uppsala University

Collection Method

Images were captured using a MiniTEM microscope (Vironova AB), with motion blur introduced by physically moving the sample.