GaussianToken: An Effective Image Tokenizer with 2D Gaussian Splatting

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Dong, Jiajun, Wang, Chengkun, Zheng, Wenzhao, Chen, Lei, Lu, Jiwen, Tang, Yansong
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910801476976640
author Dong, Jiajun
Wang, Chengkun
Zheng, Wenzhao
Chen, Lei
Lu, Jiwen
Tang, Yansong
author_facet Dong, Jiajun
Wang, Chengkun
Zheng, Wenzhao
Chen, Lei
Lu, Jiwen
Tang, Yansong
contents Effective image tokenization is crucial for both multi-modal understanding and generation tasks due to the necessity of the alignment with discrete text data. To this end, existing approaches utilize vector quantization (VQ) to project pixels onto a discrete codebook and reconstruct images from the discrete representation. However, compared with the continuous latent space, the limited discrete codebook space significantly restrict the representational ability of these image tokenizers. In this paper, we propose GaussianToken: An Effective Image Tokenizer with 2D Gaussian Splatting as a solution. We first represent the encoded samples as multiple flexible featured 2D Gaussians characterized by positions, rotation angles, scaling factors, and feature coefficients. We adopt the standard quantization for the Gaussian features and then concatenate the quantization results with the other intrinsic Gaussian parameters before the corresponding splatting operation and the subsequent decoding module. In general, GaussianToken integrates the local influence of 2D Gaussian distribution into the discrete space and thus enhances the representation capability of the image tokenizer. Competitive reconstruction performances on CIFAR, Mini-ImageNet, and ImageNet-1K demonstrate the effectiveness of our framework. Our code is available at: https://github.com/ChrisDong-THU/GaussianToken.
format Preprint
id arxiv_https___arxiv_org_abs_2501_15619
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle GaussianToken: An Effective Image Tokenizer with 2D Gaussian Splatting
Dong, Jiajun
Wang, Chengkun
Zheng, Wenzhao
Chen, Lei
Lu, Jiwen
Tang, Yansong
Computer Vision and Pattern Recognition
Artificial Intelligence
Effective image tokenization is crucial for both multi-modal understanding and generation tasks due to the necessity of the alignment with discrete text data. To this end, existing approaches utilize vector quantization (VQ) to project pixels onto a discrete codebook and reconstruct images from the discrete representation. However, compared with the continuous latent space, the limited discrete codebook space significantly restrict the representational ability of these image tokenizers. In this paper, we propose GaussianToken: An Effective Image Tokenizer with 2D Gaussian Splatting as a solution. We first represent the encoded samples as multiple flexible featured 2D Gaussians characterized by positions, rotation angles, scaling factors, and feature coefficients. We adopt the standard quantization for the Gaussian features and then concatenate the quantization results with the other intrinsic Gaussian parameters before the corresponding splatting operation and the subsequent decoding module. In general, GaussianToken integrates the local influence of 2D Gaussian distribution into the discrete space and thus enhances the representation capability of the image tokenizer. Competitive reconstruction performances on CIFAR, Mini-ImageNet, and ImageNet-1K demonstrate the effectiveness of our framework. Our code is available at: https://github.com/ChrisDong-THU/GaussianToken.
title GaussianToken: An Effective Image Tokenizer with 2D Gaussian Splatting
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2501.15619