Towards 1000-fold Electron Microscopy Image Compression for Connectomics via VQ-VAE with Transformer Prior

TL;DR

This paper introduces a VQ-VAE-based compression framework for electron microscopy images that achieves up to 1024x compression, enabling efficient storage and analysis of petascale datasets with optional texture restoration using a Transformer prior.

Contribution

It presents a novel VQ-VAE compression method with a Transformer prior for EM images, supporting extreme compression and selective high-resolution reconstruction.

Findings

Achieves 16x to 1024x compression ratios.

Enables pay-as-you-decode with optional texture restoration.

Supports ROI-driven high-resolution reconstruction.

Abstract

Petascale electron microscopy (EM) datasets push storage, transfer, and downstream analysis toward their current limits. We present a vector-quantized variational autoencoder-based (VQ-VAE) compression framework for EM that spans 16x to 1024x and enables pay-as-you-decode usage: top-only decoding for extreme compression, with an optional Transformer prior that predicts bottom tokens (without changing the compression ratio) to restore texture via feature-wise linear modulation (FiLM) and concatenation; we further introduce an ROI-driven workflow that performs selective high-resolution reconstruction from 1024x-compressed latents only where needed.