Tag: Computer Vision and Pattern Recognition
BIR-Adapter reduces training needs for blind image restoration
—
What the study found BIR-Adapter is a parameter-efficient diffusion adapter for blind image restoration, a task where the degradation is not known in advance. The abstract says it achieves competitive performance, and in several settings superior performance, while requiring up to 36× fewer trained parameters. Why the authors say this matters The authors conclude that…
Graph matching improved IVUS-OCT sequence registration
—
Graph matching framework for cross-modality intravascular ultrasound and optical coherence tomography sequence registration with simultaneous temporal and rotational alignment.
L-DCS improved visibility in dense rock art scenes
—
Localised decorrelation stretch enhances visibility of rock art in dense scenes through automated windowed colour stretching, improving upon global DStretch methods for archaeological documentation.
