高斯影像：透過2D高斯點陣投影實現每秒1000幀的影像表示和壓縮。

GaussianImage: 1000 FPS Image Representation and Compression by 2D Gaussian Splatting

March 13, 2024

作者: Xinjie Zhang, Xingtong Ge, Tongda Xu, Dailan He, Yan Wang, Hongwei Qin, Guo Lu, Jing Geng, Jun Zhang

cs.AI

摘要

最近，隱式神經表示（INRs）在圖像表示和壓縮方面取得了巨大成功，提供高視覺品質和快速渲染速度，達到每秒10-1000幀，假設有足夠的GPU資源可用。然而，這種要求通常阻礙了它們在內存有限的低端設備上的使用。為此，我們提出了一種名為GaussianImage的基於2D高斯擴散的圖像表示和壓縮的開創性範式。我們首先引入2D高斯來表示圖像，其中每個高斯具有8個參數，包括位置、協方差和顏色。隨後，我們揭示了一種基於累積求和的新型渲染算法。值得注意的是，我們的方法在GPU內存使用量至少低3倍且擬合時間快5倍，不僅在表示性能上與INRs（例如WIRE、I-NGP）不相上下，而且無論參數大小如何，渲染速度都達到每秒1500-2000幀。此外，我們整合現有的向量量化技術來構建圖像編解碼器。實驗結果表明，我們的編解碼器在速率失真性能上與基於壓縮的INRs（如COIN和COIN++）相當，同時實現了約每秒1000幀的解碼速度。此外，初步概念證明，我們的編解碼器在使用部分bits-back編碼時優於COIN和COIN++的性能。

English

Implicit neural representations (INRs) recently achieved great success in image representation and compression, offering high visual quality and fast rendering speeds with 10-1000 FPS, assuming sufficient GPU resources are available. However, this requirement often hinders their use on low-end devices with limited memory. In response, we propose a groundbreaking paradigm of image representation and compression by 2D Gaussian Splatting, named GaussianImage. We first introduce 2D Gaussian to represent the image, where each Gaussian has 8 parameters including position, covariance and color. Subsequently, we unveil a novel rendering algorithm based on accumulated summation. Remarkably, our method with a minimum of 3times lower GPU memory usage and 5times faster fitting time not only rivals INRs (e.g., WIRE, I-NGP) in representation performance, but also delivers a faster rendering speed of 1500-2000 FPS regardless of parameter size. Furthermore, we integrate existing vector quantization technique to build an image codec. Experimental results demonstrate that our codec attains rate-distortion performance comparable to compression-based INRs such as COIN and COIN++, while facilitating decoding speeds of approximately 1000 FPS. Additionally, preliminary proof of concept shows that our codec surpasses COIN and COIN++ in performance when using partial bits-back coding.