A saliency prediction model on 360 degree images using color dictionary based sparse representation

Ling, Jing; Zhang, Kao; Zhang, Yingxue; Yang, Daiqin; Chen, Zhenzhong

doi:10.1016/j.image.2018.03.007

Cited by 65 publications

(29 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Table 5. Results for the head-eye movement prediction with dataset [22] Method KLD CC NSS AUC GVBS360 [8] 0.698 0.527 0.851 0.714 [9] 0.481 0.532 0.918 0.734 [10] 0.431 0.659 0.971 0.746 [11] 0.42 0.61 0.81 0.72 [13] 0.477 0.550 0.936 0.736 Proposed 0.469 0.570 1.027 0.731…”

Section: Resultsmentioning

confidence: 99%

“…ERP images centered on two different longitude along the equator and cube map faces generated by rotating the cube center to several angles are used in [11] to generate the saliency map. In [13], the ERP image is split into patches and the sparse feature is extracted and an integrated saliency map is produced after taking the visual acuity and latitude bias into consideration.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Saliency Prediction for Omnidirectional Images Considering Optimization on Sphere Domain

Dedhia

Chiang

Char

2019

ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

There are several formats to describe the omnidirectional images. Among them, equirectangular projection (ERP), represented as 2D image, is the most widely used format. There exist many outstanding methods capable of well predicting the saliency maps for the conventional 2D images. But these works cannot be directly extended to predict the saliency map of the ERP image, since the content on ERP is not for direct display. Instead, the viewport image on demand is generated after converting the ERP image to the sphere domain, followed by rectilinear projection. In this paper, we propose a model to predict the saliency maps of the ERP images using existing saliency predictors for the 2D image. Some pre-processing and post-processing are used to manage the problem mentioned above. In particular, a smoothing based optimization is realized on the sphere domain. A public dataset of omnidirectional images is used to perform all the experiments and competitive results are achieved.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Saliency Prediction for Omnidirectional Images Considering Optimization on Sphere Domain

Dedhia

Chiang

Char

2019

ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

show abstract

“…Hu et al [31] introduced a model that predicts relevant areas of a 360-degree video and decides in which direction a human observer should look for each frame. Some authors have also focused on omni-directional images [30,32,33].…”

Section: Visual Saliency Predictionmentioning

confidence: 99%

PathGAN: Visual Scanpath Prediction with Generative Adversarial Networks

Assens

Giró-i-Nieto

McGuinness

et al. 2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

We introduce PathGAN, a deep neural network for visual scanpath prediction trained on adversarial examples. A visual scanpath is defined as the sequence of fixation points over an image defined by a human observer with its gaze. PathGAN is composed of two parts, the generator and the discriminator. Both parts extract features from images using off-the-shelf networks, and train recurrent layers to generate or discriminate scanpaths accordingly. In scanpath prediction, the stochastic nature of the data makes it very difficult to generate realistic predictions using supervised learning strategies, but we adopt adversarial training as a suitable alternative. Our experiments prove how PathGAN improves the state of the art of visual scanpath prediction on the iSUN and Salient360! datasets.

show abstract

“…The recorded visual information is sent to a closed circuit television (CCTV) control center that supervises the VSS device concerned, and to administrator(s) there with proper access rights. The information is then made available for preventing possible incidents/accidents in target area(s), and for understanding the root cause of and investigating any incidents/accidents once they have occurred, and presenting evidence for resolving situations [1][2][3][4].…”

Section: Introductionmentioning

confidence: 99%

A Mechanism of Masking Identification Information regarding Moving Objects Recorded on Visual Surveillance Systems by Differentially Implementing Access Permission

Park

Kim

2019

Electronics

View full text Add to dashboard Cite

Video surveillance systems (VSS), used as a measure of security strengthening as well as investigation, are provided principally in heavily crowded public places. They record images of moving objects and transmit them to the control center. Typically, the recorded images are stored after being encrypted, or masked using visual obfuscations on a concerned image(s) in the identification-enabling data contained in the visual information. The stored footage is recovered to its original state by authorized users. However, the recovery entails the restoration of all information in the visual data, possibly infiltrating the privacy of the object(s) other than the one(s) whose images are requested. In particular, Artificial Intelligence Healthcare that checks the health status of an object through images has the same problem and must protect the patient's identification information. This study proposes a masking mechanism wherein the infiltration of visual data privacy on videos is minimized by limiting the objects whose images are recovered with differential use of access permission granted to the requesting users.

show abstract

A saliency prediction model on 360 degree images using color dictionary based sparse representation

Cited by 65 publications

References 28 publications

Saliency Prediction for Omnidirectional Images Considering Optimization on Sphere Domain

Saliency Prediction for Omnidirectional Images Considering Optimization on Sphere Domain

PathGAN: Visual Scanpath Prediction with Generative Adversarial Networks

A Mechanism of Masking Identification Information regarding Moving Objects Recorded on Visual Surveillance Systems by Differentially Implementing Access Permission

Contact Info

Product

Resources

About