Objective methods for assessing perceptual image quality traditionally attempted to quantify the visibility of errors (differences) between a distorted image and a reference image using a variety of known properties of the human visual system. Under the assumption that human visual perception is highly adapted for extracting structural information from a scene, we introduce an alternative complementary framework for quality assessment based on the degradation of structural information. As a specific example of this concept, we develop a structural similarity index and demonstrate its promise through a set of intuitive examples, as well as comparison to both subjective ratings and state-of-the-art objective methods on a database of images compressed with JPEG and JPEG2000. A MATLAB implementation of the proposed algorithm is available online at http://www.cns.nyu.edu//spl sim/lcv/ssim/.

/pdf/image-quality-assessment-from-error-visibility-to-structural-1rlwcqe34t.pdf

Image quality assessment: from error visibility to structural similarity

Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Despite the breakthroughs in accuracy and speed of single image super-resolution using faster and deeper convolutional neural networks, one central problem remains largely unsolved: how do we recover the finer texture details when we super-resolve at large upscaling factors? The behavior of optimization-based super-resolution methods is principally driven by the choice of the objective function. Recent work has largely focused on minimizing the mean squared reconstruction error. The resulting estimates have high peak signal-to-noise ratios, but they are often lacking high-frequency details and are perceptually unsatisfying in the sense that they fail to match the fidelity expected at the higher resolution. In this paper, we present SRGAN, a generative adversarial network (GAN) for image super-resolution (SR). To our knowledge, it is the first framework capable of inferring photo-realistic natural images for 4x upscaling factors. To achieve this, we propose a perceptual loss function which consists of an adversarial loss and a content loss. The adversarial loss pushes our solution to the natural image manifold using a discriminator network that is trained to differentiate between the super-resolved images and original photo-realistic images. In addition, we use a content loss motivated by perceptual similarity instead of similarity in pixel space. Our deep residual network is able to recover photo-realistic textures from heavily downsampled images on public benchmarks. An extensive mean-opinion-score (MOS) test shows hugely significant gains in perceptual quality using SRGAN. The MOS scores obtained with SRGAN are closer to those of the original high-resolution images than to those obtained with any state-of-the-art method.

/pdf/photo-realistic-single-image-super-resolution-using-a-2y0438rlmd.pdf

Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network

We propose a deep learning method for single image super-resolution (SR). Our method directly learns an end-to-end mapping between the low/high-resolution images. The mapping is represented as a deep convolutional neural network (CNN) that takes the low-resolution image as the input and outputs the high-resolution one. We further show that traditional sparse-coding-based SR methods can also be viewed as a deep convolutional network. But unlike traditional methods that handle each component separately, our method jointly optimizes all layers. Our deep CNN has a lightweight structure, yet demonstrates state-of-the-art restoration quality, and achieves fast speed for practical on-line usage. We explore different network structures and parameter settings to achieve trade-offs between performance and speed. Moreover, we extend our network to cope with three color channels simultaneously, and show better overall reconstruction quality.

Image Super-Resolution Using Deep Convolutional Networks

We consider image transformation problems, where an input image is transformed into an output image. Recent methods for such problems typically train feed-forward convolutional neural networks using a \emph{per-pixel} loss between the output and ground-truth images. Parallel work has shown that high-quality images can be generated by defining and optimizing \emph{perceptual} loss functions based on high-level features extracted from pretrained networks. We combine the benefits of both approaches, and propose the use of perceptual loss functions for training feed-forward networks for image transformation tasks. We show results on image style transfer, where a feed-forward network is trained to solve the optimization problem proposed by Gatys et al in real-time. Compared to the optimization-based method, our network gives similar qualitative results but is three orders of magnitude faster. We also experiment with single-image super-resolution, where replacing a per-pixel loss with a perceptual loss gives visually pleasing results.

Perceptual Losses for Real-Time Style Transfer and Super-Resolution

Reduced-reference (RR) image quality measures aim to predict the visual quality of distorted images with only partial information about the reference images. In this paper, we propose an RR quality assessment method based on a natural image statistic model in the wavelet transform domain. In particular, we observe that the marginal distribution of wavelet coefficients changes in different ways for different types of image distortions. To quantify such changes, we estimate the Kullback-Leibler distance between the marginal distributions of wavelet coefficients of the reference and distorted images. A generalized Gaussian model is employed to summarize the marginal distribution of wavelet coefficients of the reference image, so that only a relatively small number of RR features are needed for the evaluation of image quality. The proposed method is easy to implement and computationally efficient. In addition, we find that many well-known types of image distortion lead to significant changes in wavelet coefficient histograms, and thus are readily detectable by our measure. The algorithm is tested with subjective ratings of a large image database that contains images corrupted with a wide variety of distortion types.

/pdf/reduced-reference-image-quality-assessment-using-a-wavelet-2l0op8qn4d.pdf

Reduced-reference image quality assessment using a wavelet-domain natural image statistic model

The objective measurement of blocking artifacts plays an important role in the design, optimization, and assessment of image and video coding systems. We propose a new approach that can blindly measure blocking artifacts in images without reference to the originals. The key idea is to model the blocky image as a non-blocky image interfered with a pure blocky signal. The task of the blocking effect measurement algorithm is then to detect and evaluate the power of the blocky signal. The proposed approach has the flexibility to integrate human visual system features such as the luminance and the texture masking effects.

https://users.ece.utexas.edu/~bevans/courses/ee381k/projects/fall98/zhou/report.pdf

Blind measurement of blocking artifacts in images

We propose a multi-task end-to-end optimized deep neural network (MEON) for blind image quality assessment (BIQA). MEON consists of two sub-networks—a distortion identification network and a quality prediction network—sharing the early layers. Unlike traditional methods used for training multi-task networks, our training process is performed in two steps. In the first step, we train a distortion type identification sub-network, for which large-scale training samples are readily available. In the second step, starting from the pre-trained early layers and the outputs of the first sub-network, we train a quality prediction sub-network using a variant of the stochastic gradient descent method. Different from most deep neural networks, we choose biologically inspired generalized divisive normalization (GDN) instead of rectified linear unit as the activation function. We empirically demonstrate that GDN is effective at reducing model parameters/layers while achieving similar quality prediction performance. With modest model complexity, the proposed MEON index achieves state-of-the-art performance on four publicly available benchmarks. Moreover, we demonstrate the strong competitiveness of MEON against state-of-the-art BIQA models using the group maximum differentiation competition methodology.

End-to-End Blind Image Quality Assessment Using Deep Neural Networks

We propose a deep bilinear model for blind image quality assessment that works for both synthetically and authentically distorted images. Our model constitutes two streams of deep convolutional neural networks (CNNs), specializing in two distortion scenarios separately. For synthetic distortions, we first pre-train a CNN to classify the distortion type and the level of an input image, whose ground truth label is readily available at a large scale. For authentic distortions, we make use of a pre-train CNN (VGG-16) for the image classification task. The two feature sets are bilinearly pooled into one representation for a final quality prediction. We fine-tune the whole network on the target databases using a variant of stochastic gradient descent. The extensive experimental results show that the proposed model achieves state-of-the-art performance on both synthetic and authentic IQA databases. Furthermore, we verify the generalizability of our method on the large-scale Waterloo Exploration Database, and demonstrate its competitiveness using the group maximum differentiation competition methodology.

Blind Image Quality Assessment Using a Deep Bilinear Convolutional Neural Network

Since its introduction in 2004, the structural similarity (SSIM) index has gained widespread popularity as a tool to assess the quality of images and to evaluate the performance of image processing algorithms and systems. There has been also a growing interest of using SSIM as an objective function in optimization problems in a variety of image processing applications. One major issue that could strongly impede the progress of such efforts is the lack of understanding of the mathematical properties of the SSIM measure. For example, some highly desirable properties such as convexity and triangular inequality that are possessed by the mean squared error may not hold. In this paper, we first construct a series of normalized and generalized (vector-valued) metrics based on the important ingredients of SSIM. We then show that such modified measures are valid distance metrics and have many useful properties, among which the most significant ones include quasi-convexity, a region of convexity around the minimizer, and distance preservation under orthogonal or unitary transformations. The groundwork laid here extends the potentials of SSIM in both theoretical development and practical applications.

/pdf/on-the-mathematical-properties-of-the-structural-similarity-l94e02oe66.pdf

Zhou Wang

Papers

Reduced-reference image quality assessment using a wavelet-domain natural image statistic model

Blind measurement of blocking artifacts in images

End-to-End Blind Image Quality Assessment Using Deep Neural Networks

Blind Image Quality Assessment Using a Deep Bilinear Convolutional Neural Network

On the Mathematical Properties of the Structural Similarity Index