Jesús Malo

Nuria Alabau-Bosque, Paula Daudén-Oliver, Jorge Vila-Tomás et al.

Subjective image quality metrics are usually evaluated according to the correlation with human opinion in databases with distortions that may appear in digital media. However, these oversee affine transformations which may represent better the changes in the images actually happening in natural conditions. Humans can be particularly invariant to these natural transformations, as opposed to the digital ones. In this work, we propose a methodology to evaluate any image quality metric by assessing their invariance to affine transformations, specifically: rotation, translation, scaling, and changes in spectral illumination. Here, invariance refers to the fact that certain distances should be neglected if their values are below a threshold. This is what we call invisibility threshold of a metric. Our methodology consists of two elements: (1) the determination of a visibility threshold in a subjective representation common to every metric, and (2) a transduction from the distance values of the metric and this common representation. This common representation is based on subjective ratings of readily available image quality databases. We determine the threshold in such common representation (the first element) using accurate psychophysics. Then, the transduction (the second element) can be trivially fitted for any metric: with the provided threshold extension of the method to any metric is straightforward. We test our methodology with some well-established metrics and find that none of them show human-like invisibility thresholds. This means that tuning the models exclusively to predict the visibility of generic distortions may disregard other properties of human vision as for instance invariances or invisibility thresholds. The data and code are publicly available to test other metrics.

Valero Laparra, Alexander Hepburn, J. Emmanuel Johnson et al.

In this paper we elaborate an extension of rotation-based iterative Gaussianization, RBIG, which makes image Gaussianization possible. Although RBIG has been successfully applied to many tasks, it is limited to medium dimensionality data (on the order of a thousand dimensions). In images its application has been restricted to small image patches or isolated pixels, because rotation in RBIG is based on principal or independent component analysis and these transformations are difficult to learn and scale. Here we present the \emph{Convolutional RBIG}: an extension that alleviates this issue by imposing that the rotation in RBIG is a convolution. We propose to learn convolutional rotations (i.e. orthonormal convolutions) by optimising for the reconstruction loss between the input and an approximate inverse of the transformation using the transposed convolution operation. Additionally, we suggest different regularizers in learning these orthonormal convolutions. For example, imposing sparsity in the activations leads to a transformation that extends convolutional independent component analysis to multilayer architectures. We also highlight how statistical properties of the data, such as multivariate mutual information, can be obtained from \emph{Convolutional RBIG}. We illustrate the behavior of the transform with a simple example of texture synthesis, and analyze its properties by visualizing the stimuli that maximize the response in certain feature and layer.

Pablo Hernández-Cámara, Jorge Vila-Tomás, Valero Laparra et al.

Subjective image quality measures based on deep neural networks are very related to models of visual neuroscience. This connection benefits engineering but, more interestingly, the freedom to optimize deep networks in different ways, make them an excellent tool to explore the principles behind visual perception (both human and artificial). Recently, a myriad of networks have been successfully optimized for many interesting visual tasks. Although these nets were not specifically designed to predict image quality or other psychophysics, they have shown surprising human-like behavior. The reasons for this remain unclear. In this work, we perform a thorough analysis of the perceptual properties of pre-trained nets (particularly their ability to predict image quality) by isolating different factors: the goal (the function), the data (learning environment), the architecture, and the readout: selected layer(s), fine-tuning of channel relevance, and use of statistical descriptors as opposed to plain readout of responses. Several conclusions can be drawn. All the models correlate better with human opinion than SSIM. More importantly, some of the nets are in pair of state-of-the-art with no extra refinement or perceptual information. Nets trained for supervised tasks such as classification correlate substantially better with humans than LPIPS (a net specifically tuned for image quality). Interestingly, self-supervised tasks such as jigsaw also perform better than LPIPS. Simpler architectures are better than very deep nets. In simpler nets, correlation with humans increases with depth as if deeper layers were closer to human judgement. This is not true in very deep nets. Consistently with reports on illusions and contrast sensitivity, small changes in the image environment does not make a big difference. Finally, the explored statistical descriptors and concatenations had no major impact.

Pablo Hernández-Cámara, Jorge Vila-Tomás, Paula Dauden-Oliver et al.

Autonomous driving is a challenging scenario for image segmentation due to the presence of uncontrolled environmental conditions and the eventually catastrophic consequences of failures. Previous work suggested that a biologically motivated computation, the so-called Divisive Normalization, could be useful to deal with image variability, but its effects have not been systematically studied over different data sources and environmental factors. Here we put segmentation U-nets augmented with Divisive Normalization to work far from training conditions to find where this adaptation is more critical. We categorize the scenes according to their radiance level and dynamic range (day/night), and according to their achromatic/chromatic contrasts. We also consider video game (synthetic) images to broaden the range of environments. We check the performance in the extreme percentiles of such categorization. Then, we push the limits further by artificially modifying the images in perceptually/environmentally relevant dimensions: luminance, contrasts and spectral radiance. Results show that neural networks with Divisive Normalization get better results in all the scenarios and their performance remains more stable with regard to the considered environmental factors and nature of the source. Finally, we explain the improvements in segmentation performance in two ways: (1) by quantifying the invariance of the responses that incorporate Divisive Normalization, and (2) by illustrating the adaptive nonlinearity of the different layers that depends on the local activity.

Pablo Hernández-Cámara, Valero Laparra, Jesús Malo

One of the key problems in computer vision is adaptation: models are too rigid to follow the variability of the inputs. The canonical computation that explains adaptation in sensory neuroscience is divisive normalization, and it has appealing effects on image manifolds. In this work we show that including divisive normalization in current deep networks makes them more invariant to non-informative changes in the images. In particular, we focus on U-Net architectures for image segmentation. Experiments show that the inclusion of divisive normalization in the U-Net architecture leads to better segmentation results with respect to conventional U-Net. The gain increases steadily when dealing with images acquired in bad weather conditions. In addition to the results on the Cityscapes and Foggy Cityscapes datasets, we explain these advantages through visualization of the responses: the equalization induced by the divisive normalization leads to more invariant features to local changes in contrast and illumination.

Alexander Hepburn, Valero Laparra, Raúl Santos-Rodriguez et al.

Previously, Barlow and Attneave hypothesised a link between biological vision and information maximisation. Following Shannon, information was defined using the probability of natural images. Several physiological and psychophysical phenomena have been derived from principles like info-max, efficient coding, or optimal denoising. However, it remains unclear how this link is expressed in mathematical terms from image probability. Classical derivations were subjected to strong assumptions on the probability models and on the behaviour of the sensors. Moreover, the direct evaluation of the hypothesis was limited by the inability of classical image models to deliver accurate estimates of the probability. Here, we directly evaluate image probabilities using a generative model for natural images, and analyse how probability-related factors can be combined to predict the sensitivity of state-of-the-art subjective image quality metrics, a proxy for human perception. We use information theory and regression analysis to find a simple model that when combining just two probability-related factors achieves 0.77 correlation with subjective metrics. This probability-based model is validated in two ways: through direct comparison with the opinion of real observers in a subjective quality experiment, and by reproducing basic trends of classical psychophysical facts such as the Contrast Sensitivity Function, the Weber-law, and contrast masking.

Nuria Alabau-Bosque, Jorge Vila-Tomás, Paula Daudén-Oliver et al.

Playing games is inherently human, and a lot of games are created to challenge different human characteristics. However, these tasks are often left out when evaluating the human-like nature of artificial models. The objective of this work is proposing a new approach to evaluate artificial models via board games. To this effect, we test the color perception and color naming capabilities of CLIP by playing the board game Hues & Cues and assess its alignment with humans. Our experiments show that CLIP is generally well aligned with human observers, but our approach brings to light certain cultural biases and inconsistencies when dealing with different abstraction levels that are hard to identify with other testing strategies. Our findings indicate that assessing models with different tasks like board games can make certain deficiencies in the models stand out in ways that are difficult to test with the commonly used benchmarks.

Jorge Vila-Tomás, Pablo Hernández-Cámara, Valero Laparra et al.

Human vision models are at the core of image processing. For instance, classical approaches to the problem of image quality are based on models that include knowledge about human vision. However, nowadays, deep learning approaches have obtained competitive results by simply approaching this problem as regression of human decisions, and training an standard network on human-rated datasets. These approaches have the advantages of being easily adaptable to a particular problem and they fit very efficiently when data is available. However, mainly due to the excess of parameters, they have the problems of lack of interpretability, and over-fitting. Here we propose a vision model that combines the best of both worlds by using a parametric neural network architecture. We parameterize the layers to have bioplausible functionality, and provide a set of bioplausible parameters. We analyzed different versions of the model and compared it with the non-parametric version. The parametric models achieve a three orders of magnitude reduction in the number of parameters without suffering in regression performance. Furthermore, we show that the parametric models behave better during training and are easier to interpret as vision models. Interestingly, we find that, even initialized with bioplausible trained for regression using human rated datasets, which we call the feature-spreading problem. This suggests that the deep learning approach is inherently flawed, and emphasizes the need to evaluate and train models beyond regression.

Paula Daudén-Oliver, David Agost-Beltran, Emilio Sansano-Sansano et al.

Image quality databases are used to train models for predicting subjective human perception. However, most existing databases focus on distortions commonly found in digital media and not in natural conditions. Affine transformations are particularly relevant to study, as they are among the most commonly encountered by human observers in everyday life. This Data Descriptor presents a set of human responses to suprathreshold affine image transforms (rotation, translation, scaling) and Gaussian noise as convenient reference to compare with previously existing image quality databases. The responses were measured using well established psychophysics: the Maximum Likelihood Difference Scaling method. The set contains responses to 864 distorted images. The experiments involved 105 observers and more than 20000 comparisons of quadruples of images. The quality of the dataset is ensured because (a) it reproduces the classical Piéron's law, (b) it reproduces classical absolute detection thresholds, and (c) it is consistent with conventional image quality databases but improves them according to Group-MAD experiments.

Alexander Hepburn, Valero Laparra, Raul Santos-Rodriguez et al.

It has been demonstrated many times that the behavior of the human visual system is connected to the statistics of natural images. Since machine learning relies on the statistics of training data as well, the above connection has interesting implications when using perceptual distances (which mimic the behavior of the human visual system) as a loss function. In this paper, we aim to unravel the non-trivial relationships between the probability distribution of the data, perceptual distances, and unsupervised machine learning. To this end, we show that perceptual sensitivity is correlated with the probability of an image in its close neighborhood. We also explore the relation between distances induced by autoencoders and the probability distribution of the training data, as well as how these induced distances are correlated with human perception. Finally, we find perceptual distances do not always lead to noticeable gains in performance over Euclidean distance in common image processing tasks, except when data is scarce and the perceptual distance provides regularization. We propose this may be due to a \emph{double-counting} effect of the image statistics, once in the perceptual distance and once in the training procedure.

J. Emmanuel Johnson, Valero Laparra, Gustau Camps-Valls et al.

Density destructors are differentiable and invertible transforms that map multivariate PDFs of arbitrary structure (low entropy) into non-structured PDFs (maximum entropy). Multivariate Gaussianization and multivariate equalization are specific examples of this family, which break down the complexity of the original PDF through a set of elementary transforms that progressively remove the structure of the data. We demonstrate how this property of density destructive flows is connected to classical information theory, and how density destructors can be used to get more accurate estimates of information theoretic quantities. Experiments with total correlation and mutual information inmultivariate sets illustrate the ability of density destructors compared to competing methods. These results suggest that information theoretic measures may be an alternative optimization criteria when learning density destructive flows.

Alexander Gomez-Villa, Adrian Martín, Javier Vazquez-Corral et al.

Visual illusions are a very useful tool for vision scientists, because they allow them to better probe the limits, thresholds and errors of the visual system. In this work we introduce the first ever framework to generate novel visual illusions with an artificial neural network (ANN). It takes the form of a generative adversarial network, with a generator of visual illusion candidates and two discriminator modules, one for the inducer background and another that decides whether or not the candidate is indeed an illusion. The generality of the model is exemplified by synthesizing illusions of different types, and validated with psychophysical experiments that corroborate that the outputs of our ANN are indeed visual illusions to human observers. Apart from synthesizing new visual illusions, which may help vision researchers, the proposed model has the potential to open new ways to study the similarities and differences between ANN and human visual perception.

Alexander Hepburn, Valero Laparra, Jesús Malo et al.

Traditionally, the vision community has devised algorithms to estimate the distance between an original image and images that have been subject to perturbations. Inspiration was usually taken from the human visual perceptual system and how the system processes different perturbations in order to replicate to what extent it determines our ability to judge image quality. While recent works have presented deep neural networks trained to predict human perceptual quality, very few borrow any intuitions from the human visual system. To address this, we present PerceptNet, a convolutional neural network where the architecture has been chosen to reflect the structure and various stages in the human visual system. We evaluate PerceptNet on various traditional perception datasets and note strong performance on a number of them as compared with traditional image quality metrics. We also show that including a nonlinearity inspired by the human visual system in classical deep neural networks architectures can increase their ability to judge perceptual similarity. Compared to similar deep learning methods, the performance is similar, although our network has a number of parameters that is several orders of magnitude less.

Valero Laparra, Sandra Jiménez, Gustavo Camps-Valls et al.

Mechanisms of human color vision are characterized by two phenomenological aspects: the system is nonlinear and adaptive to changing environments. Conventional attempts to derive these features from statistics use separate arguments for each aspect. The few statistical approaches that do consider both phenomena simultaneously follow parametric formulations based on empirical models. Therefore, it may be argued that the behavior does not come directly from the color statistics but from the convenient functional form adopted. In addition, many times the whole statistical analysis is based on simplified databases that disregard relevant physical effects in the input signal, as for instance by assuming flat Lambertian surfaces. Here we address the simultaneous statistical explanation of (i) the nonlinear behavior of achromatic and chromatic mechanisms in a fixed adaptation state, and (ii) the change of such behavior. Both phenomena emerge directly from the samples through a single data-driven method: the Sequential Principal Curves Analysis (SPCA) with local metric. SPCA is a new manifold learning technique to derive a set of sensors adapted to the manifold using different optimality criteria. A new database of colorimetrically calibrated images of natural objects under these illuminants was collected. The results obtained by applying SPCA show that the psychophysical behavior on color discrimination thresholds, discount of the illuminant and corresponding pairs in asymmetric color matching, emerge directly from realistic data regularities assuming no a priori functional form. These results provide stronger evidence for the hypothesis of a statistically driven organization of color sensors. Moreover, the obtained results suggest that color perception at this low abstraction level may be guided by an error minimization strategy rather than by the information maximization principle.

Valero Laparra, Gustavo Camps-Valls, Jesús Malo

Most signal processing problems involve the challenging task of multidimensional probability density function (PDF) estimation. In this work, we propose a solution to this problem by using a family of Rotation-based Iterative Gaussianization (RBIG) transforms. The general framework consists of the sequential application of a univariate marginal Gaussianization transform followed by an orthonormal transform. The proposed procedure looks for differentiable transforms to a known PDF so that the unknown PDF can be estimated at any point of the original domain. In particular, we aim at a zero mean unit covariance Gaussian for convenience. RBIG is formally similar to classical iterative Projection Pursuit (PP) algorithms. However, we show that, unlike in PP methods, the particular class of rotations used has no special qualitative relevance in this context, since looking for interestingness is not a critical issue for PDF estimation. The key difference is that our approach focuses on the univariate part (marginal Gaussianization) of the problem rather than on the multivariate part (rotation). This difference implies that one may select the most convenient rotation suited to each practical application. The differentiability, invertibility and convergence of RBIG are theoretically and experimentally analyzed. Relation to other methods, such as Radial Gaussianization (RG), one-class support vector domain description (SVDD), and deep neural networks (DNN) is also pointed out. The practical performance of RBIG is successfully illustrated in a number of multidimensional problems such as image synthesis, classification, denoising, and multi-information estimation.

Valero Laparra, Sandra Jiménez, Devis Tuia et al.

This paper presents a new framework for manifold learning based on a sequence of principal polynomials that capture the possibly nonlinear nature of the data. The proposed Principal Polynomial Analysis (PPA) generalizes PCA by modeling the directions of maximal variance by means of curves, instead of straight lines. Contrarily to previous approaches, PPA reduces to performing simple univariate regressions, which makes it computationally feasible and robust. Moreover, PPA shows a number of interesting analytical properties. First, PPA is a volume-preserving map, which in turn guarantees the existence of the inverse. Second, such an inverse can be obtained in closed form. Invertibility is an important advantage over other learning methods, because it permits to understand the identified features in the input domain where the data has physical meaning. Moreover, it allows to evaluate the performance of dimensionality reduction in sensible (input-domain) units. Volume preservation also allows an easy computation of information theoretic quantities, such as the reduction in multi-information after the transform. Third, the analytical nature of PPA leads to a clear geometrical interpretation of the manifold: it allows the computation of Frenet-Serret frames (local features) and of generalized curvatures at any point of the space. And fourth, the analytical Jacobian allows the computation of the metric induced by the data, thus generalizing the Mahalanobis distance. These properties are demonstrated theoretically and illustrated experimentally. The performance of PPA is evaluated in dimensionality and redundancy reduction, in both synthetic and real datasets from the UCI repository.

Valero Laparra, Juan Gutiérrez, Gustavo Camps-Valls et al.

A successful class of image denoising methods is based on Bayesian approaches working in wavelet representations. However, analytical estimates can be obtained only for particular combinations of analytical models of signal and noise, thus precluding its straightforward extension to deal with other arbitrary noise sources. In this paper, we propose an alternative non-explicit way to take into account the relations among natural image wavelet coefficients for denoising: we use support vector regression (SVR) in the wavelet domain to enforce these relations in the estimated signal. Since relations among the coefficients are specific to the signal, the regularization property of SVR is exploited to remove the noise, which does not share this feature. The specific signal relations are encoded in an anisotropic kernel obtained from mutual information measures computed on a representative image database. Training considers minimizing the Kullback-Leibler divergence (KLD) between the estimated and actual probability functions of signal and noise in order to enforce similarity. Due to its non-parametric nature, the method can eventually cope with different noise sources without the need of an explicit re-formulation, as it is strictly necessary under parametric Bayesian formalisms. Results under several noise levels and noise sources show that: (1) the proposed method outperforms conventional wavelet methods that assume coefficient independence, (2) it is similar to state-of-the-art methods that do explicitly include these relations when the noise source is Gaussian, and (3) it gives better numerical and visual performance when more complex, realistic noise sources are considered. Therefore, the proposed machine learning approach can be seen as a more flexible (model-free) alternative to the explicit description of wavelet coefficient relations for image denoising.

Gustavo Camps-Valls, Juan Gutiérrez, Gabriel Gómez-Pérez et al.

Conventional SVM-based image coding methods are founded on independently restricting the distortion in every image coefficient at some particular image representation. Geometrically, this implies allowing arbitrary signal distortions in an $n$-dimensional rectangle defined by the $\varepsilon$-insensitivity zone in each dimension of the selected image representation domain. Unfortunately, not every image representation domain is well-suited for such a simple, scalar-wise, approach because statistical and/or perceptual interactions between the coefficients may exist. These interactions imply that scalar approaches may induce distortions that do not follow the image statistics and/or are perceptually annoying. Taking into account these relations would imply using non-rectangular $\varepsilon$-insensitivity regions (allowing coupled distortions in different coefficients), which is beyond the conventional SVM formulation. In this paper, we report a condition on the suitable domain for developing efficient SVM image coding schemes. We analytically demonstrate that no linear domain fulfills this condition because of the statistical and perceptual inter-coefficient relations that exist in these domains. This theoretical result is experimentally confirmed by comparing SVM learning in previously reported linear domains and in a recently proposed non-linear perceptual domain that simultaneously reduces the statistical and perceptual relations (so it is closer to fulfilling the proposed condition). These results highlight the relevance of an appropriate choice of the image representation before SVM learning.