A model of perceived dynamic range for HDR images

Hulusić, Vedad; Debattista, Kurt; Valenzise, Giuseppe; Dufaux, Frédéric

doi:10.1016/j.image.2016.11.005

Cited by 11 publications

(14 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally, Screenshot is a screen capture, cropped to match the resolution of test stimuli. These images were selected out of 19 candidate pictures, on the basis of their spatial information, key and colorfulness [19], as well as on their semantics (outdoor, people/faces, man-made objects). Screenshot was selected to include an example of synthetic image, which might be representative of a screen-content compression scenario.…”

Section: Subjective Evaluationmentioning

confidence: 99%

Quality Assessment of Deep-Learning-Based Image Compression

Valenzise¹,

Purica²,

Hulusić³

et al. 2018

2018 IEEE 20th International Workshop on Multimedia Signal Processing (MMSP)

Self Cite

View full text Add to dashboard Cite

Image compression standards rely on predictive coding, transform coding, quantization and entropy coding, in order to achieve high compression performance. Very recently, deep generative models have been used to optimize or replace some of these operations, with very promising results. However, so far no systematic and independent study of the coding performance of these algorithms has been carried out. In this paper, for the first time, we conduct a subjective evaluation of two recent deeplearning-based image compression algorithms, comparing them to JPEG 2000 and to the recent BPG image codec based on HEVC Intra. We found that compression approaches based on deep auto-encoders can achieve coding performance higher than JPEG 2000, and sometimes as good as BPG. We also show experimentally that the PSNR metric is to be avoided when evaluating the visual quality of deep-learning-based methods, as their artifacts have different characteristics from those of DCT or wavelet-based codecs. In particular, images compressed at low bitrate appear more natural than JPEG 2000 coded pictures, according to a no-reference naturalness measure. Our study indicates that deep generative models are likely to bring huge innovation into the video coding arena in the coming years.

show abstract

Section: Subjective Evaluationmentioning

confidence: 99%

Quality Assessment of Deep-Learning-Based Image Compression

Valenzise¹,

Purica²,

Hulusić³

et al. 2018

2018 IEEE 20th International Workshop on Multimedia Signal Processing (MMSP)

Self Cite

View full text Add to dashboard Cite

show abstract

“…Although the formulated model [13] can predict well overall subjective DR scores, we also found significant exceptions and prediction failures. In fact, any DR predictor learned on such a small-size dataset might incur the risk of overfitting.…”

Section: Introductionmentioning

confidence: 57%

“…The scores were obtained by asking observers to rate images based on the magnitude of the overall difference between the brightest and darkest region(s) of the picture. Later, we have leveraged this data to evaluate robust dynamic range (DR) measures [6], and to derive a DR predictor which takes into account also the area of highlight regions [13].…”

Section: Introductionmentioning

confidence: 99%

Perceived Dynamic Range of HDR Images with no Semantic Information

Hulusić¹,

Valenzise²,

Dufaux³

2018

Self Cite

View full text Add to dashboard Cite

Computing dynamic range of high dynamic range (HDR) content is an important procedure when selecting the test material, designing and validating algorithms, or analyzing aesthetic attributes of HDR content. It can be computed on a pixelbased level, measured through subjective tests or predicted using a mathematical model. However, all these methods have certain limitations. This paper investigates whether dynamic range of modeled images with no semantic information, but with the same first order statistics as the original, natural content, is perceived the same as for the corresponding natural images. If so, it would be possible to improve the perceived dynamic range (PDR) predictor model by using additional objective metrics, more suitable for such synthetic content. Within the subjective study, three experiments were conducted with 43 participants. The results show significant correlation between the mean opinion scores for the two image groups. Nevertheless, natural images still seem to provide better cues for evaluation of PDR.

show abstract

“…The data used during the GP run were six images downscaled to half their original resolution. These images were obtained from the subset of 36 images taken from the Fairchild database used in perceptual studies of dynamic range [HDVD16]. This database of images was chosen because of their availability and because they had been selected as they broadly cover the HDR range in the afore-mentioned study.…”

Section: Gp Methodsmentioning

confidence: 99%

“…To ensure running the method is practical, results comparing the proposed method with state‐of‐the‐art methods are based on the full set of 36 HDR images at HD resolution used in the perceptual study [HDVD16]. They are further augmented for this test by the 19 left eye images from the set of images used for the feature mapping application discussed below and the Stanford Memorial Chapel image for a total of 56 images.…”

Section: Applicationsmentioning

confidence: 99%

Application‐Specific Tone Mapping Via Genetic Programming

Debattista

2017

Computer Graphics Forum

View full text Add to dashboard Cite

High dynamic range (HDR) imagery permits the manipulation of real‐world data distinct from the limitations of the traditional, low dynamic range (LDR), content. The process of retargeting HDR content to traditional LDR imagery via tone mapping operators (TMOs) is useful for visualizing HDR content on traditional displays, supporting backwards‐compatible HDR compression and, more recently, is being frequently used for input into a wide variety of computer vision applications. This work presents the automatic generation of TMOs for specific applications via the evolutionary computing method of genetic programming (GP). A straightforward, generic GP method that generates TMOs for a given fitness function and HDR content is presented. Its efficacy is demonstrated in the context of three applications: Visualization of HDR content on LDR displays, feature mapping and compression. For these applications, results show good performance for the generated TMOs when compared to traditional methods. Furthermore, they demonstrate that the method is generalizable and could be used across various applications that require TMOs but for which dedicated successful TMOs have not yet been discovered.

show abstract

A model of perceived dynamic range for HDR images

Cited by 11 publications

References 49 publications

Quality Assessment of Deep-Learning-Based Image Compression

Quality Assessment of Deep-Learning-Based Image Compression

Perceived Dynamic Range of HDR Images with no Semantic Information

Application‐Specific Tone Mapping Via Genetic Programming

Contact Info

Product

Resources

About