Learning-based image coding: early solutions reviewing and subjective quality evaluation

Ascenso, João; Akyazi, Pinar; Pereira, Fernando; Ebrahimi, Touradj

doi:10.1117/12.2555368

Cited by 24 publications

(28 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We notice that Ascenso et al [4] and Valenzise et al [24] have done some similar works on evaluating learning-based image coding methods. [24] compared a CNN method [16] and a RNN method [23] with JPEG2000 and BPG.…”

Section: Introductionmentioning

confidence: 82%

Subjective evaluation of traditional and learning-based image coding methods

Fang,

Zhang,

et al. 2021

Preprint

View full text Add to dashboard Cite

We conduct a subjective experiment to compare the performance of traditional image coding methods and learning-based image coding methods. HEVC and VVC, the state-of-the-art traditional coding methods, are used as the representative traditional methods. The learning-based methods used contain not only CNN-based methods, but also a GAN-based method, all of which are advanced or typical. Single Stimuli (SS), which is also called Absolute Category Rating (ACR), is adopted as the methodology of the experiment to obtain perceptual quality of images. Additionally, we utilize some typical and frequently used objective quality metrics to evaluate the coding methods in the experiment as comparison. The experiment shows that CNN-based and GAN-based methods can perform better than traditional methods in low bit-rates. In high bit-rates, however, it is hard to verify whether CNNbased methods are superior to traditional methods. Because the GAN method doesn't provide models with high target bit-rates, we can't exactly tell the performance of the GAN method in high bit-rates. Furthermore, some popular objective quality metrics have not shown the ability well to measure quality of images generated by learning-based coding methods, especially the GAN-based one.

show abstract

Section: Introductionmentioning

confidence: 82%

Subjective evaluation of traditional and learning-based image coding methods

Fang,

Zhang,

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…It is nowadays proven that end-to-end neural networks are able to beat the most performing image compression standards such as JPEG [ 15 ], JPEG 2000 [ 40 ], BPG [ 41 ], AV1 [ 42 ] or VTM [ 43 ]. The field has been very active, and more and more performing architectures, on top of extensive surveys, are regularly proposed [ 5 , 6 , 7 , 8 ]. Most of the existing architectures share the same goal and principles that are depicted in Figure 3 .…”

Section: Learning-based Transmissionmentioning

confidence: 99%

“…This survey focuses not on providing a comprehensive overview of the literature for each part of the multimedia encoding and delivery ecosystem but on introducing the main challenges, emphasizing the latest advances in each area, and giving our perspective for building more efficient intelligent multimedia coding and delivery systems. For surveys focusing on particular parts of this ecosystem, interested readers are referred to [ 5 , 6 , 7 , 8 ] for compression/decompression, ref. [ 9 ] for interactivity, refs.…”

Section: Introductionmentioning

confidence: 99%

Machine Learning for Multimedia Communications

Thomos

Maugey

Toni

2022

Sensors

View full text Add to dashboard Cite

Machine learning is revolutionizing the way multimedia information is processed and transmitted to users. After intensive and powerful training, some impressive efficiency/accuracy improvements have been made all over the transmission pipeline. For example, the high model capacity of the learning-based architectures enables us to accurately model the image and video behavior such that tremendous compression gains can be achieved. Similarly, error concealment, streaming strategy or even user perception modeling have widely benefited from the recent learning-oriented developments. However, learning-based algorithms often imply drastic changes to the way data are represented or consumed, meaning that the overall pipeline can be affected even though a subpart of it is optimized. In this paper, we review the recent major advances that have been proposed all across the transmission chain, and we discuss their potential impact and the research challenges that they raise.

show abstract

“…The diagonal terms are maximal when most popular items belongs to the subset Y. The popularity of an item with a feature b is given by ||Ub|| 2 2 . The off-diagonal terms depict the correlations between the features in the subset Y.…”

Section: Pi-based Compression Algorithmmentioning

confidence: 99%

“…Storage growth is exceeding even the highest estimates with no sign of it slowing down anytime soon: 2.5 quintillion bytes of data are created each day at our current pace [1], and it will only accelerate with the advent of IoTs, volumetric videos, and new sensors. The storage burden has been partially alleviated by state-ofthe-art compression algorithms, which can substantially reduce the amount of bits needed to store one or multiple sources, e.g., end-to-end learning-based image compression algorithms to minimize the compression rate [2], MPEG standards to ensure exploitation of spatial and temporal correlation [3], joint source compression [4], [5], [6], [7]. All these coding strategies have led to impressive compression ratio, which however will be scaling always with the number of sources.…”

Section: Introductionmentioning

confidence: 99%

Large Database Compression Based on Perceived Information

Maugey

Toni

2020

IEEE Signal Process. Lett.

View full text Add to dashboard Cite

Lossy compression algorithms trade bits for quality, aiming at reducing as much as possible the bitrate needed to represent the original source (or set of sources), while preserving the source quality. In this letter, we propose a novel paradigm of compression algorithms, aimed at compressing as much as possible a set of sources, minimizing the information loss perceived by the final user instead of the actual source quality loss, under compression rate constraints. As main contributions, we first introduce the concept of perceived information (PI), which reflects the information perceived by a given user experiencing a data collection, and which is evaluated as the volume spanned by the sources features in a personalized latent space. We then formalize the rate-PI optimization problem and propose an algorithm to solve this compression problem. Finally, we validate our algorithm against benchmark solutions with simulation results, showing the gain in taking into account users' preferences while also maximizing the perceived information in the feature domain.

show abstract

Learning-based image coding: early solutions reviewing and subjective quality evaluation

Cited by 24 publications

References 0 publications

Subjective evaluation of traditional and learning-based image coding methods

Subjective evaluation of traditional and learning-based image coding methods

Machine Learning for Multimedia Communications

Large Database Compression Based on Perceived Information

Contact Info

Product

Resources

About