Practical Video Denoising Dataset (PVDD)
收藏资源简介:
PVDD是一个专为视频去噪研究设计的数据集,包含200对噪声和清洁的动态视频,支持sRGB和RAW格式。与现有数据集相比,PVDD涵盖了更多样化和自然的运动场景。数据集通过在RAW域中使用物理意义明确的传感器噪声模型合成真实噪声,并通过ISP处理生成sRGB视频对。PVDD不仅用于评估视频去噪网络,还推动了视频去噪领域的研究进展,特别是在处理复杂自然运动和真实噪声方面。
PVDD is a dataset specifically tailored for video denoising research. It comprises 200 pairs of noisy and clean dynamic videos supporting both sRGB and RAW formats. Compared with existing datasets, PVDD covers a more diverse and natural set of motion scenarios. Realistic noise is synthesized using a physically meaningful sensor noise model in the RAW domain, and sRGB video pairs are generated via ISP processing. PVDD not only serves as a benchmark for evaluating video denoising networks, but also advances research progress in the field of video denoising, especially in handling complex natural motions and realistic noise.




