Horse Iris Super-Resolution Dataset from Zhaosu Horse Farm, Xinjiang

This super-resolution dataset is constructed based on high-resolution horse iris images and is intended to support iris image super-resolution reconstruction and related research tasks. The dataset contains paired high-resolution and low-resolution iris image samples. The high-resolution images are obtained from video frames captured under near-infrared conditions, followed by manual annotation and standardized cropping, and are uniformly resized to 256×256 pixels.The low-resolution samples are generated from the corresponding high-resolution iris images through bicubic downsampling combined with a Gaussian blur degradation model, with a degradation factor set to ×4. This process is designed to simulate image quality degradation caused by factors such as shooting distance, optical system performance, and device resolution limitations in real-world acquisition scenarios. Each high- and low-resolution image pair maintains a strict one-to-one correspondence, which facilitates supervised super-resolution reconstruction experiments.The dataset covers iris images from both eyes of multiple horses, with samples collected in a natural pasture environment using non-contact near-infrared imaging. As a result, the data reflect variations in image clarity, texture visibility, and imaging conditions commonly encountered in practical applications. The dataset can be used for training and evaluation of iris image super-resolution methods and also serves as a data foundation for related computer vision research.