LotteryCodec: Searching the Implicit Representation in a Random Network for Low-Complexity Image Compression

TL;DR

LotteryCodec leverages untrained subnetworks within random networks for efficient image compression, achieving state-of-the-art results and adaptable complexity without extensive training.

Contribution

Proposes the lottery codec hypothesis and a novel method to use untrained subnetworks for high-performance, low-complexity image compression.

Findings

Outperforms VTM in single-image compression.

Enables adaptive decoding complexity via mask ratios.

Achieves competitive rate-distortion performance.

Abstract

We introduce and validate the lottery codec hypothesis, which states that untrained subnetworks within randomly initialized networks can serve as synthesis networks for overfitted image compression, achieving rate-distortion (RD) performance comparable to trained networks. This hypothesis leads to a new paradigm for image compression by encoding image statistics into the network substructure. Building on this hypothesis, we propose LotteryCodec, which overfits a binary mask to an individual image, leveraging an over-parameterized and randomly initialized network shared by the encoder and the decoder. To address over-parameterization challenges and streamline subnetwork search, we develop a rewind modulation mechanism that improves the RD performance. LotteryCodec outperforms VTM and sets a new state-of-the-art in single-image compression. LotteryCodec also enables adaptive decoding complexity through adjustable mask ratios, offering flexible compression solutions for diverse device constraints and application requirements.