Overfitted codecs have demonstrated promising performance for image and video compression. In particular, for image compression, the Cool-chic family of models has shown competitive performance against scene-agnostic models, with orders of magnitude lower decoding complexity at the cost of a longer overfitting process. However, these overfitted image codecs are not fully optimized toward the rate-distortion objective: their network weights remain in full precision during training, and the associated quantization parameters are selected in a separate post-training stage. Furthermore, their synthesis operates at a single scale, which overlooks cross-scale redundancy. In this paper, we propose MIRC, an overfitted image codec in which every coded component, including the latents, the synthesis network, and the entropy models, is quantized and entropy coded under a single rate-distortion objective, adopting the end-to-end compression pipeline of the neural video representation codec NVRC. We further introduce a multi-scale representation with cross-stage parameter sharing, which improves coding efficiency at a small transmitted overhead. On the CLIC2020 professional validation set, MIRC achieves a 10.5% BD-rate saving against VVC (VTM 22.0). Moreover, MIRC offers a family of configurations spanning 1.2 to 2.9 kMAC per pixel, so the decoding budget can be selected to match the deployment target.
Figures & tables
Fig. 1 : (a) Prior overfitted codecs [ 8 , 9 , 10 ] train their latents against a rate-distortion objective, but quantize and entropy-code the network weights in a separate stage. (b) MIRC (ours) optimizes every coded component, including latents, synthesis weights and entropy model (EM) weights, under a single rate-distortion objective with quantization-aware training.
Fig. 2 : Overview of the proposed MIRC framework, with insets detailing the shared decoder block (top) and the causal context of the ARM (right). All coded components ( z^ , θ^ , ϕ^ ) are optimized under the single rate-distortion objective.
Fig. 3 : Rate-distortion and complexity comparison. (a), (b): rate-distortion curves (PSNR versus bpp), with VTM 22.0 as the anchor and , on Kodak and the CLIC2020 professional validation dataset, respectively. (c): BD-rate against VTM 22.0 versus decoding complexity, in multiply-accumulate operations (MAC) per pixel.
Encoding
Decoding
Time [s/image]
Complexity [kMAC/px]
Time [ms/Mpx]
Cool-chic 4.0
3782
1.43
135.3
C3
2382
1.79
-
MoRIC (HOP)
16,231
2.93
-
MIRC ( High )
3615
2.92
17.8
TABLE I : Encoding and decoding time with complexity, measured and averaged on CLIC2020 pro. dataset using an NVIDIA RTX 4090 GPU and Intel Core i7-14700 CPU. Orange indicates GPU computation, blue indicates CPU computation.
Configuration
BD-rate [%] ↓
w/o Multi-scale decoder
+2.85
w/o Decoder block sharing between stages
+4.95
w/o End-to-end optimization
+13.01
MIRC (default)
0.0
TABLE II : Ablation of the main design choices. BD-rate is computed relative to the default MIRC configuration ( High ) on ten images of CLIC2020 pro. dataset.
Most learned image compression systems rely on a single latent representation combined with a hyperprior, which limits their ability to efficiently capture image structure across spatial scales. In this work, we propose a hierarchical latent representation to improve the efficiency of the entropy model. By using multiple latents at different scales, each with its own entropy model, we better capture the spatial structure of the latent representation. Our experiments show that this approach achieves a 17.9% BD-rate reduction over VVC on Kodak, demonstrating the effectiveness of multi-scale latent representations. Furthermore, the approach is orthogonal to other advances in learned image compression, making it a versatile addition to existing methods.
Jonas Brenig, Radu Timofte
Computer Vision Lab, CAIDAS & IFI, University of Würzburg, Germany
Recent progress in learning-based image compression has demonstrated that end-to-end optimization can substantially outperform traditional codecs by jointly learning compact latent representations and probabilistic entropy models. However, many existing approaches achieve high rate-distortion efficiency at the expense of increased computational cost and limited parallelism. This paper presents ARCHE - Autoregressive Residual Compression with Hyperprior and Excitation, an end-to-end learned image compression framework that balances modeling accuracy and computational efficiency. The proposed architecture unifies hierarchical, spatial, and channel-based priors within a single probabilistic framework, capturing both global and local dependencies in the latent representation of the image, while employing adaptive feature recalibration and residual refinement to enhance latent representation quality. Without relying on recurrent or transformer-based components, ARCHE attains state-of-the-art rate-distortion efficiency: it reduces the BD-Rate by approximately 48% relative to the commonly used benchmark model of Balle et al., 30% relative to the channel-wise autoregressive model of Minnen & Singh and 5% against the VVC Intra codec on the Kodak benchmark dataset. The framework maintains computational efficiency with 95M parameters and 222ms running time per image. Visual comparisons confirm sharper textures and improved color fidelity, particularly at lower bit rates, demonstrating that accurate entropy modeling can be achieved through efficient convolutional designs suitable for practical deployment.
Sofia Iliopoulou, Dimitris Ampeliotis, Athanassios Skodras
Dept. of Electrical and Computer Engineering, University of Patras, Patras, Greece · Dept. of Digital Media and Communication, Ionian University, Argostoli, Greece
Learned image compression (LIC) is bottlenecked by the need to store independent models for each rate-distortion operating point. Existing variable bit-rate (VBR) methods aim to reduce this overhead via dense parameter modulation, but forcing a shared backbone to approximate divergent mappings causes severe feature entanglement. Specifically, low-rate smoothing gradients inherently conflict with the preservation of high-frequency textural details, leading to sub-optimal performance. To resolve this, we propose MixCompress, a unified VBR framework based on sparse structural specialization. While sparsely gated Mixture-of-Experts (MoE) routing successfully mitigates gradient conflict, it operates on a fixed computational budget. To address the increased representational demands of higher bit-rates we introduce a Mixture-of-Depths (MoD) extension to dynamically scale model capacity. Combined with Conditional Auxiliary Transforms (CAT) for dynamic sub-band energy modulation, our hierarchical framework effectively dynamically scales capacity. Extensive evaluations demonstrate that MixCompress not only matches individually optimized single-rate baselines but can even surpass them, establishing a new Pareto frontier for computationally efficient image coding.