Saved in:
Bibliographic Details
Main Authors: Malidarreh, Parisa Boodaghi, Saurav, Jillur Rahman, Pham, Thuong Le Hoai, Hajighasemi, Amir, Samadi, Anahita, Maydeo, Saurabh Shrinivas, Nasr, Mohammad Sadegh, Luber, Jacob M.
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2503.10832
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915197358178304
author Malidarreh, Parisa Boodaghi
Saurav, Jillur Rahman
Pham, Thuong Le Hoai
Hajighasemi, Amir
Samadi, Anahita
Maydeo, Saurabh Shrinivas
Nasr, Mohammad Sadegh
Luber, Jacob M.
author_facet Malidarreh, Parisa Boodaghi
Saurav, Jillur Rahman
Pham, Thuong Le Hoai
Hajighasemi, Amir
Samadi, Anahita
Maydeo, Saurabh Shrinivas
Nasr, Mohammad Sadegh
Luber, Jacob M.
contents Vector Quantization (VQ) techniques face significant challenges in codebook utilization, limiting reconstruction fidelity in image modeling. We introduce a Dual Codebook mechanism that effectively addresses this limitation by partitioning the representation into complementary global and local components. The global codebook employs a lightweight transformer for concurrent updates of all code vectors, while the local codebook maintains precise feature representation through deterministic selection. This complementary approach is trained from scratch without requiring pre-trained knowledge. Experimental evaluation across multiple standard benchmark datasets demonstrates state-of-the-art reconstruction quality while using a compact codebook of size 512 - half the size of previous methods that require pre-training. Our approach achieves significant FID improvements across diverse image domains, particularly excelling in scene and face reconstruction tasks. These results establish Dual Codebook VQ as an efficient paradigm for high-fidelity image reconstruction with significantly reduced computational requirements.
format Preprint
id arxiv_https___arxiv_org_abs_2503_10832
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Dual Codebook VQ: Enhanced Image Reconstruction with Reduced Codebook Size
Malidarreh, Parisa Boodaghi
Saurav, Jillur Rahman
Pham, Thuong Le Hoai
Hajighasemi, Amir
Samadi, Anahita
Maydeo, Saurabh Shrinivas
Nasr, Mohammad Sadegh
Luber, Jacob M.
Computer Vision and Pattern Recognition
Vector Quantization (VQ) techniques face significant challenges in codebook utilization, limiting reconstruction fidelity in image modeling. We introduce a Dual Codebook mechanism that effectively addresses this limitation by partitioning the representation into complementary global and local components. The global codebook employs a lightweight transformer for concurrent updates of all code vectors, while the local codebook maintains precise feature representation through deterministic selection. This complementary approach is trained from scratch without requiring pre-trained knowledge. Experimental evaluation across multiple standard benchmark datasets demonstrates state-of-the-art reconstruction quality while using a compact codebook of size 512 - half the size of previous methods that require pre-training. Our approach achieves significant FID improvements across diverse image domains, particularly excelling in scene and face reconstruction tasks. These results establish Dual Codebook VQ as an efficient paradigm for high-fidelity image reconstruction with significantly reduced computational requirements.
title Dual Codebook VQ: Enhanced Image Reconstruction with Reduced Codebook Size
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2503.10832