Implicit neural representations (INRs) have achieved remarkable success in image representation and compression, but they require substantial training time and memory. Meanwhile, recent 2D Gaussian Splatting (GS) methods (\textit{e.g.}, GaussianImage) offer promising alternatives through efficient primitive-based rendering. However, these methods require excessive Gaussian primitives to maintain high visual fidelity. To exploit the potential of GS-based approaches, we present GaussianImage++, which utilizes limited Gaussian primitives to achieve impressive representation and compression performance. Firstly, we introduce a distortion-driven densification mechanism. It progressively allocates Gaussian primitives according to signal intensity. Secondly, we employ context-aware Gaussian filters for each primitive, which assist in the densification to optimize Gaussian primitives based on varying image content. Thirdly, we integrate attribute-separated learnable scalar quantizers and quantization-aware training, enabling efficient compression of primitive attributes. Experimental results demonstrate the effectiveness of our method. In particular, GaussianImage++ outperforms GaussianImage and INRs-based COIN in representation and compression performance while maintaining real-time decoding and low memory usage.
@article{arxiv.2512.19108,
title = {GaussianImage++: Boosted Image Representation and Compression with 2D Gaussian Splatting},
author = {Tiantian Li and Xinjie Zhang and Xingtong Ge and Tongda Xu and Dailan He and Jun Zhang and Yan Wang},
journal= {arXiv preprint arXiv:2512.19108},
year = {2026}
}
Comments
Accepted to AAAI 2026. Code URL:https://github.com/Sweethyh/GaussianImage_plus.git