Filters








3,852 Hits in 7.2 sec

On the Use of a Convolutional Neural Network to Predict Perceptual Quality of Images without Reference for Different Viewing Distances

Aladine Chetouani, Marius Pedersen
2019 2019 IEEE International Conference on Image Processing (ICIP)  
In this study, we propose to consider this information by estimating the quality of a given image without a reference image for different viewing distances.  ...  For that, a Convolutional Neural Network (CNN) model was used in this study. Relevant patches are first selected from the image and they are then used as inputs to the CNN.  ...  In this work we use a Convolutional Neural Network (CNN) to predict perceived image quality at different viewing distances.  ... 
doi:10.1109/icip.2019.8803056 dblp:conf/icip/ChetouaniP19 fatcat:ugcviowsnfcljeaya6pcjym7pm

Learned Perceptual Image Enhancement [article]

Hossein Talebi, Peyman Milanfar
2017 arXiv   pre-print
This metric is implemented using a CNN (convolutional neural network) trained on a large-scale dataset labelled with aesthetic preferences of human raters.  ...  This loss allows us to conveniently perform back-propagation in our learning framework to simultaneously optimize for similarity to a given ground truth reference and perceptual quality.  ...  These algorithms rely on training a convolutional neural network (CNN) architecture, with respect to a loss function.  ... 
arXiv:1712.02864v1 fatcat:jjh4m6axcrhppgzzxop46xg7e4

Learning structure of stereoscopic image for no-reference quality assessment with convolutional neural network

Wei Zhang, Chenfei Qu, Lin Ma, Jingwei Guan, Rui Huang
2016 Pattern Recognition  
In this paper, we propose to learn the structures of stereoscopic image based on convolutional neural network (CNN) for no-reference quality assessment.  ...  Multilayer perceptron (MLP) is further employed to summarize the learned representation to a final value to indicate the perceptual quality of the stereo image patch pair.  ...  For onecolumn CNN, these parameters are used to initialize the network before training on the difference images of the stereoscopic images.  ... 
doi:10.1016/j.patcog.2016.01.034 fatcat:cend7opm4bempp2vbqpocy33qu

Embedding Novel Views in a Single JPEG Image [article]

Yue Wu and Guotao Meng and Qifeng Chen
2021 arXiv   pre-print
We propose a novel approach for embedding novel views in a single JPEG image while preserving the perceptual fidelity of the modified JPEG image and the restored novel views.  ...  We conducted experiments on public datasets with different novel view synthesis methods, and the results show that the proposed method can restore high-fidelity novel views from a slightly modified JPEG  ...  For each scene, some views are used as input, while others are used as target views. The objective of the network is to predict target views based on given views.  ... 
arXiv:2108.13003v1 fatcat:afx7sy3inbhcldtr5jpkv7vi54

Learning wavelet coefficients for face super-resolution

Liu Ying, Sun Dinghua, Wang Fuping, Lim Keng Pang, Chiew Tuan Kiang, Lai Yi
2020 The Visual Computer  
Firstly, this paper uses prior knowledge of face images to manually emphases relevant facial features with more attention. Then, a linear low-rank convolution in the network is used.  ...  To overcome this problem, we propose a novel deep neural network to predict the super-resolution wavelet coefficients in order to obtain clearer facial images.  ...  Reference [13] uses perceptual loss and adversarial loss to improve the realism of the predicted image. Zhang Y et al.  ... 
doi:10.1007/s00371-020-01925-2 fatcat:lqnjva6ttrgk7cxinwcbo5gupi

Image Inpainting using Block-wise Procedural Training with Annealed Adversarial Counterpart [article]

Chao Yang, Yuhang Song, Xiaofeng Liu, Qingming Tang, C.-C. Jay Kuo
2018 arXiv   pre-print
Recent advances in deep generative models have shown promising potential in image inpanting, which refers to the task of predicting missing pixel values of an incomplete image using the known context.  ...  We present a new approach to address the difficulty of training a very deep generative model to synthesize high-quality photo-realistic inpainting.  ...  All images have original size 256x256. Zoom in for best viewing quality. Figure 7 .Figure 8 . 78 Visual comparison of a result using different types of convolutional layers.  ... 
arXiv:1803.08943v2 fatcat:ckfhqrnndjc6vdav42evwym7e4

Learning to generate images with perceptual similarity metrics

Jake Snell, Karl Ridgeway, Renjie Liao, Brett D. Roads, Michael C. Mozer, Richard S. Zemel
2017 2017 IEEE International Conference on Image Processing (ICIP)  
For three different architectures, we collected human judgments of the quality of image reconstructions.  ...  We propose instead to use a loss function that is better calibrated to human perceptual judgments of image quality: the multiscale structural-similarity score (MS-SSIM) [31] .  ...  A city-block (L 1 ) distance is sometimes used as well, referred to as the mean absolute error or MAE.  ... 
doi:10.1109/icip.2017.8297089 dblp:conf/icip/SnellRLRMZ17 fatcat:ykczprlhjbhzbax5btzs7vnjai

Texture Synthesis Using Shallow Convolutional Networks with Random Filters [article]

Ivan Ustyuzhaninov, Wieland Brendel, Leon A. Gatys, Matthias Bethge
2016 arXiv   pre-print
Here we demonstrate that the feature space of random shallow convolutional neural networks (CNNs) can serve as a surprisingly good model of natural textures.  ...  Samples synthesized from the model capture spatial correlations on scales much larger then the receptive field size, and sometimes even rival or surpass the perceptual quality of state of the art texture  ...  Our results clearly demonstrate that Gram matrices computed from the feature maps of convolutional neural networks generically lead to useful summary statistics for texture synthesis.  ... 
arXiv:1606.00021v1 fatcat:caytzmpkhfey5izblocfxesmq4

Gibson Env: Real-World Perception for Embodied Agents [article]

Fei Xia, Amir Zamir, Zhi-Yang He, Alexander Sax, Jitendra Malik, Silvio Savarese
2018 arXiv   pre-print
This has given rise to learning-in-simulation which consequently casts a question on whether the results transfer to real-world.  ...  In this paper, we are concerned with the problem of developing real-world perception for active agents, propose Gibson Virtual Environment for this purpose, and showcase sample perceptual tasks learned  ...  Acknowledgement: We gratefully acknowledge the support of Facebook, Toyota (1186781-31-UDARO), ONR MURI (N00014-14-1-0671), ONR (1165419-10-TDAUZ); Nvidia, CloudMinds, Panasonic (1192707-1-GWMSX).  ... 
arXiv:1808.10654v1 fatcat:24pkus2ksfarrmiquodjqtxcqe

Learning to Generate Images with Perceptual Similarity Metrics [article]

Jake Snell, Karl Ridgeway, Renjie Liao, Brett D. Roads, Michael C. Mozer, Richard S. Zemel
2017 arXiv   pre-print
For three different architectures, we collected human judgments of the quality of image reconstructions.  ...  We propose instead to use a loss function that is better calibrated to human perceptual judgments of image quality: the multiscale structural-similarity score (MS-SSIM).  ...  A city-block (L 1 ) distance is sometimes used as well, referred to as the mean absolute error or MAE.  ... 
arXiv:1511.06409v3 fatcat:ig2kqqvucjcihbogroo5djy3bi

3D visual saliency and convolutional neural network for blind mesh quality assessment

Ilyass Abouelaziz, Aladine Chetouani, Mohammed El Hassouni, Longin Jan Latecki, Hocine Cherifi
2019 Neural computing & applications (Print)  
In this work, we propose a noreference convolutional neural network (CNN) framework to estimate the perceived visual quality of 3D meshes.  ...  Finally, a CNN is used for the feature learning and the quality score estimation.  ...  To demonstrate the influence of this parameter, we test the ability of our network in predicting the visual quality by using a variety of convolution kernels while fixing the other parameters.  ... 
doi:10.1007/s00521-019-04521-1 fatcat:zxk3nmtlbza5pbfyhohz7vt5zu

Image Quality Assessment without Reference by Combining Deep Learning-Based Features and Viewing Distance

Aladine Chetouani, Marius Pedersen
2021 Applied Sciences  
We introduce in this study a novel image quality metric able to estimate the quality of a given image without reference for different viewing distances between the image and the observer.  ...  For each patch, a feature vector is extracted from a convolutional neural network model and concatenated at the viewing distance, for which the quality is predicted.  ...  Conflicts of Interest: The authors declare no conflict of interest.  ... 
doi:10.3390/app11104661 fatcat:rtkvff5e7ngxllvr4knr7ypim4

Blind Quality Estimation by Disentangling Perceptual and Noisy Features in High Dynamic Range Images

Navaneeth Kamballur Kottayil, Giuseppe Valenzise, Frederic Dufaux, Irene Cheng
2018 IEEE Transactions on Image Processing  
We propose a new convolutional neural network based model for No reference image quality assessment(NR-IQA) on HDR data.  ...  This model predicts the amount and location of noise, perceptual influence of image pixels on the noise, and the perceived quality, of a distorted image without any reference image.  ...  We use a convolutional neural network based architecture that computes perceptual features to derive the quality of an image. The contributions of this paper are as follows: 1.  ... 
doi:10.1109/tip.2017.2778570 pmid:29990064 fatcat:ikew6youx5hfzclw4aamzk4m4e

Deep convolutional reconstruction for gradient-domain rendering

Markus Kettunen, Erik Härkönen, Jaakko Lehtinen
2019 ACM Transactions on Graphics  
We optimize our network to minimize a perceptual image distance metric calibrated to the human visual system.  ...  Drawing on the power of modern convolutional neural networks, we propose a novel reconstruction method for gradient-domain rendering.  ...  for converting many of them to Mitsuba.  ... 
doi:10.1145/3306346.3323038 fatcat:5d3alsh4hneqzgoiasg6myyldi

Perceptual Image Super-Resolution with Progressive Adversarial Network [article]

Lone Wong, Deli Zhao, Shaohua Wan, Bo Zhang
2020 arXiv   pre-print
Single Image Super-Resolution (SISR) aims to improve resolution of small-size low-quality image from a single one.  ...  To address this issue, we propose Progressive Adversarial Network (PAN) that is capable of coping with this difficulty for domain-specific image super-resolution.  ...  Super-Identity Convolutional Neural Network (SICNN) [24] uses a super-identity loss function to recover the person identity.  ... 
arXiv:2003.03756v4 fatcat:dg32vyec5ndhrhmpin7kp4uhwi
« Previous Showing results 1 — 15 out of 3,852 results