Towards improved lossy image compression: Human image reconstruction with public-domain images

Bhown, Ashutosh; Mukherjee, Soham; Yang, Sean; Chandak, Shubham; Fischer-Hwang, Irena; Tatwawadi, Kedar; Fan, Judith E.; Weissman, Tsachy

doi:10.48550/arxiv.1810.11137

Cited by 2 publications

(3 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Human Compression. (Bhown et al, 2018; demonstrates a hand-crafted compression scheme in which humans write down text descriptions of the image to compress; the decoder consists of another human who has access to a database of images and image editing software. Humanrated scores for this scheme were higher than WebP at similar rates, despite the fact that the reconstructions may not necessarily be faithful at the pixel-level.…”

Section: Related Workmentioning

confidence: 99%

“…Recent works from the lossy compression literature have demonstrated that when human satisfaction or semantic visual information is prioritized, compression schemes that manually encode images using human-written text descriptions as the compressed representation (Bhown et al, 2018; yield significant improvements compared to traditional compressors. These works show that when operating at such low bit-rates, high levels of human satisfaction can still be achieved despite low pixel-wise fidelity.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Federated Neural Compression Under Heterogeneous Data

Lei,

Hassani,

Bidokhti

2023

2023 IEEE International Symposium on Information Theory (ISIT)

View full text Add to dashboard Cite

Recent advances in text-to-image generative models provide the ability to generate high-quality images from short text descriptions. These foundation models, when pre-trained on billion-scale datasets, are effective for various downstream tasks with little or no further training. A natural question to ask is how such models may be adapted for image compression. We investigate several techniques in which the pre-trained models can be directly used to implement compression schemes targeting novel low rate regimes. We show how text descriptions can be used in conjunction with side information to generate high-fidelity reconstructions that preserve both semantics and spatial structure of the original. We demonstrate that at very low bit-rates, our method can significantly improve upon learned compressors in terms of perceptual and semantic fidelity, despite no end-to-end training.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Federated Neural Compression Under Heterogeneous Data

Lei,

Hassani,

Bidokhti

2023

2023 IEEE International Symposium on Information Theory (ISIT)

View full text Add to dashboard Cite

show abstract

“…Going further and completely discarding distortion for extremely low bitrates, several methods rely on semantic descriptions for image representation. The semantic description can be done using text as first showcased in [4] † These authors contributed equally to this work. The names are arranged in alphabetical order.…”

Section: Introductionmentioning

confidence: 99%