

This paper presents a method for reconstructing the original high dynamic range (HDR) image from a saturated low dynamic range (LDR) image with missing physical information, specifically for single dielectric objects. A deep neural network approach is employed to map an 8-bit LDR image directly to its corresponding HDR representation. We begin by analyzing the reflection and saturation characteristics of dielectric materials and then construct an HDR image database using a diverse set of dielectric objects. Each HDR image is clipped to generate a set of 8-bit LDR images. All HDR-LDR image pairs are normalized to a fixed resolution and used for training and validation. A deep convolutional neural network (CNN) is designed in the form of an autoencoder architecture with skip connections. The entire network is implemented using MATLAB’s machine learning toolbox, with the ADAM optimizer employed for training. The performance of the proposed method is evaluated using a separate validation set. Comparative experiments with existing methods demonstrate that our approach achieves significantly higher reconstruction accuracy and better histogram fitting.

One central challenge in modeling material appearance perception is the creation of an explainable and navigable representation space. In this study, we address this by training a StyleGAN2-ADA deep generative model on a large-scale, physically based rendered dataset containing translucent and glossy objects with varying intrinsic optical parameters. The resulting latent vectors are analyzed through dimensionality reduction, and their perceptual validity is assessed via psychophysical experiments. Furthermore, we evaluate the generalization capabilities of StyleGAN2-ADA on unseen materials. We also explore inverse mapping techniques from latent vectors reduced by principal component analysis back to original optical parameters, highlighting both the potential and the limitations of generative models for explicit, parameter-based image synthesis. A comprehensive analysis provides significant insights into the latent structure of gloss and translucency perception and advances the practical application of generative models for controlled material appearance generation.

One central challenge in modeling material appearance perception is the creation of an explainable and navigable representation space. In this study, we address this by training a StyleGAN2-ADA deep generative model on a large-scale, physically based rendered dataset containing translucent and glossy objects with varying intrinsic optical parameters. The resulting latent vectors are analyzed through dimensionality reduction, and their perceptual validity is assessed via psychophysical experiments. Furthermore, we evaluate the generalization capabilities of StyleGAN2-ADA on unseen materials. We also explore inverse mapping techniques from latent vectors reduced by principal component analysis back to original optical parameters, highlighting both the potential and the limitations of generative models for explicit, parameter-based image synthesis. A comprehensive analysis provides significant insights into the latent structure of gloss and translucency perception and advances the practical application of generative models for controlled material appearance generation.

Naturalness is a complex appearance attribute that is dependent on multiple visual appearance attributes like color, gloss, roughness, and their interaction. It impacts the perceived quality of an object and should therefore be reproduced correctly. In recent years, the use of color 3D printing technology has seen considerable growth in different fields like cultural heritage, medical, entertainment, and fashion for producing 3D objects with the correct appearance. This paper investigates the reproduction of naturalness attribute using a color 3D printing technology and the naturalness perception of the 3D printed objects. Results indicate that naturalness perception of 3D printed objects is highly subjective but is found to be objectively dependent mainly on a printed object’s surface elevation and roughness.

We consider a method for reconstructing the original HDR image from a single LDR image suffering from saturation for metallic objects. A deep neural network approach is adopted for directly mapping from 8-bit LDR image to an HDR image. An HDR image database is first constructed using a large number of objects with different shapes and made of various metal materials. Each captured HDR image is clipped to create a set of 8-bit LDR images. The whole pairs of HDR and LDR images are separated and used to train and test the network. Next, we design a deep CNN in the form of a deep auto-encoder architecture. The network was also equipped with skip connections to keep high image resolution. The CNN algorithm is constructed using MATLAB's machine-learning functions. The entire network consists of 32 layers and 85,900 learnable parameters. The performances of the proposed method are examined in experiments using a test image set. We also compare our method with other methods. It is confirmed that our method is significantly superior in reconstruction accuracy and the good histogram fitting.

The measurement of specular gloss using a glossmeter is normalized in the ISO 2813 Standard, which is widely used for many industrial applications. In practice, the principle of the measurement relies on using a primary standard that approximates a perfectly polished back glass surface and an optical design where rectangular diaphragms are used for the source and detection apertures. Any deviation in the refractive index or the polishing level of the standard artefact, or in the machining of the rectangular diaphragms ends in measurement uncertainties. To tackle these issues, we propose to calculate the specular gloss from the bidirectional reflectance distribution function (BRDF) measured using a goniospectrophotometer equipped with a conoscopic detection. With such an instrument, no calibration sample is needed anymore, and the geometry of measurement given in the standard can be applied with good accuracy. The method has been implemented and tested on samples of various gloss values.

Translucency is an appearance attribute, which primarily results from subsurface scattering of light. The visual perception of translucency has gained attention in the past two decades. However, the studies mostly address thick and complex 3D objects that completely occlude the background. On the other hand, the perception of transparency of flat and thin see-through filters has been studied more extensively. Despite this, perception of translucency in see-through filters that do not completely occlude the background remains understudied. In this work, we manipulated the sharpness and contrast of black-and-white checkerboard patterns to simulate the impression of see-through filters. Afterward, we conducted paired-comparison psychophysical experiments to measure how the amount of background blur and contrast relates to perceived translucency. We found that while both blur and contrast affect translucency, the relationship is neither monotonic, nor straightforward.

The visual mechanisms behind our ability to distinguish translucent and opaque materials is not fully understood. Disentanglement of the contributions of surface reflectance and subsurface light transport to the still image structure is an ill-posed problem. While the overwhelming majority of the works addressing translucency perception use static stimuli, behavioral studies show that human observers tend to move objects to assess their translucency. Therefore, we hypothesize that translucent objects appear more translucent and less opaque when observed in motion than when shown as still images. In this manuscript, we report two psychophysical experiments that we conducted using static and dynamic visual stimuli to investigate how motion affects perceived translucency.

Color imaging has historically been treated as a phenomenon sufficiently described by three independent parameters. Recent advances in computational resources and in the understanding of the human aspects are leading to new approaches that extend the purely metrological view of color towards a perceptual approach describing the appearance of objects, documents and displays. Part of this perceptual view is the incorporation of spatial aspects, adaptive color processing based on image content, and the automation of color tasks, to name a few. This dynamic nature applies to all output modalities, including hardcopy devices, but to an even larger extent to soft-copy displays with their even larger options of dynamic processing. Spatially adaptive gamut and tone mapping, dynamic contrast, and color management continue to support the unprecedented development of display hardware covering everything from mobile displays to standard monitors, and all the way to large size screens and emerging technologies. The scope of inquiry is also broadened by the desire to match not only color, but complete appearance perceived by the user. This conference provides an opportunity to present, to interact, and to learn about the most recent developments in color imaging and material appearance researches, technologies and applications. Focus of the conference is on color basic research and testing, color image input, dynamic color image output and rendering, color image automation, emphasizing color in context and color in images, and reproduction of images across local and remote devices. The conference covers also software, media, and systems related to color and material appearance. Special attention is given to applications and requirements created by and for multidisciplinary fields involving color and/or vision.