Showing papers presented at "Computational Color Imaging Workshop in 2017"

Hand-Crafted vs Learned Descriptors for Color Texture Classification

Abstract: The paper presents a comparison between hand-crafted and learned descriptors for color texture classification. The comparison is performed on five color texture databases that include images under varying imaging conditions: scales, camera orientations, light orientations, light color temperatures, etc. Results demonstrate that learned descriptors, on average, significantly outperform hand-crafted descriptors. However, results obtained on the individual databases show that in the case of Outex 14, that includes training and test images taken under varying illuminant conditions, hand-crafted descriptors perform better than learned descriptors.

A Multidistortion Database for Image Quality

Abstract: In this paper we introduce a multidistortion database, where 10 pristine color images have been simultaneously distorted by two types of distortions: blur and JPEG and noise and JPEG. The two datasets consist of respectively 350 and 400 images, and have been subjectively evaluated within two psycho-physical experiments. We here also propose two no reference multidistortion metrics, one for each of the two datasets, as linear combinations of no reference single distortion ones. The optimized weights of the combinations are obtained using particle swarm optimization. The different combinations proposed show good performance when correlated with the subjective scores of the multidistortion database.

Improved opponent colour local binary patterns for colour texture classification

Abstract: In this paper we introduce Improved Opponent Colour Local Binary Patterns (IOCLBP), a conceptually simple yet effective descriptor for colour texture classification. The method was experimentally validated over eight datasets of colour texture images. The results show that IOCLBP outperformed other LBP variants and was at least as effective as last generation features from Convolutional Neural Networks.

Analytical Survey of Highlight Detection in Color and Spectral Images

Abstract: Detection of highlights is a prominent issue in computer vision, graphics and image processing. Applications which require object properties measurement or rendering are affected by specular reflection since the models assume matte diffusing surfaces most of the time. Hence, detection, and sometimes removal, of specular reflection (highlights) in an image may be critical. Several methods are proposed for addressing this issue. In this paper, we present a review and analysis of these techniques in color and spectral images.

Image Contrast Measure as a Gloss Material Descriptor

Abstract: Bidirectional reflectance distribution function provides a physical description of material appearance. In particular, it helps to describe the gloss. We suggest that, at least, one attribute of gloss: Contrast gloss (luster), may be described directly from an image by using local image contrast measurement. In this article, we investigate the relation between image contrast measures, gloss perception and bidirectional reflectance distribution function based on the Ward’s \(\alpha \) model parameter. Although more investigation is required to provide stronger conclusions, it seems that image related contrast measures may provide an indication of gloss perception.

Artistic photo filtering recognition using cnns

Abstract: In this paper we propose an approach based on deep Convolutional Neural Networks (CNNs) to recognize artistic photo filters applied to images. A total of 22 types of Instagram-like filters is considered. Different CNN architectures taken from the image recognition literature are compared on a dataset of more than 0.46 M images from the Places-205 dataset. Experimental results show that not only it is possible to reliably determine whether or not one of these filters has been applied, but also which one. Differently from other tasks, where the fine-tuning of a CNN trained on a different problem is usually good enough, here the fine-tuned AlexNet obtains an accuracy of only 67.5%. We show, instead, that an accuracy of about 99.0% can be obtained by training a CNN from scratch for this specific problem.

T-Rex: A Milano Retinex Implementation Based on Intensity Thresholding

Abstract: We present T-Rex (from the words Threshold and REtineX), a new Milano Retinex implementation, based on an intensity thresholding strategy. Like all the algorithms of the Retinex family, T-Rex takes as input a color image and processes its channels separately. For each channel, T-Rex re-scales the chromatic intensity of each pixel x by the average of a set of pixels whose intensity, weighted by a function of the distance from x, exceeds the intensity of x. The main novelty of this approach is devised by the usage of the pixel intensity as a threshold for selecting the pixels relevant to Retinex. Here we show an application of T-Rex as image enhancer, showing that, as a member of the Retinex family, it equalizes the dynamic range of any input picture and makes its details more evident.

Similarities and Differences in the Mathematical Formalizations of the Retinex Model and Its Variants

Abstract: Edwin H. Land and John J. McCann introduced the Retinex model as a computational theory of color vision. However, they specified the details of Retinex rather algorithmically and not mathematically and this opened the way to a multitude of different interpretations of their model, many times even contradicting ones. The aim of this paper is to present a systematic and self-contained overview about these different interpretations and the corresponding mathematical formalizations in terms of variational principles and partial differential equations.

A Complexity-Based Image Analysis to Investigate Interference Between Distortions and Image Contents in Image Quality Assessment

Abstract: In this paper we investigate how distortion and image content interfere within image quality assessment. To this end we analyze how full reference metrics behave within three different groups of images. Given a dataset of images, these are first classified as high, medium or low complexity and the FR methods are applied within each group separately. We consider images from LIVE, CSIQ and LIVE multi-distorted databases. We evaluate 17 full reference quality metrics available in the literature on each of these the high, medium and low complexity groups. We observe that within these groups the metrics better correlate subjective data. In particular, the signal based metrics are the ones that show the highest improvements. Moreover for the LIVE multi-distorted database the gain in performance is evident for all the metrics considered.

A Database of Spectral Filter Array Images that Combine Visible and NIR

Abstract: Spectral filter array emerges as a multispectral imaging technology, which could benefit several applications. Although several instantiations are prototyped and commercialized, there are yet only a few raw data available that could serve research and help to evaluate and design adequate related image processing and algorithms. This document presents a freely available spectral filter array database of images that combine visible and near infra-red information.

Characterization by Hyperspectral Imaging and Hypercolor Gamut Estimation for Structural Color Prints

Abstract: A recently developed color printing system on glass plates, based on dot-by-dot laser irradiation generating the growth of metallic nanoparticles in a special coating, produces structural colors depending strongly on the illumination and observation configuration. The difficulty for an exhaustive color characterization of the printing technology comes not only from the goniochromaticity of the samples, but also from their very high specularity, to which classical measurement instruments are not adapted. Moreover, as the light-matter interaction relies on a number of optical phenomena (surface plasmon resonance, interferences , diffraction, effects of polarization of light) for which no predictive model is available today, their characterization requires measurement of many printed samples. In this paper, we present a characterization method based on multispec-tral imaging and on spectral prediction for halftone colors that permitted a first gamut estimation in three specific illumination/viewing configurations. Recent progresses in nanotechnologies enable the coloration of glass with interesting visual rendering. This is for example the case of the technology developed by the la-boratoire Hubert Curien, called PICSULP [1], where a coating containing silver [2] is deposited on the glass plate, then irradiated by a laser beam in order to anneal the coating and cluster the metallic ions into metallic nanoparticles (NPs). Goniochromatic col-oration of the glass plate surface is thus obtained thanks to various optical phenomena: the presence of silver NPs generates surface plasmon resonance, therefore wavelength-selective absorption as in stained glass [3]; the organization of the NPs along one plane parallel to the coating-air interface generates interferences as in thin films; the NPs can even be aligned along parallel lines, as shown in Figure 1-a, which produces diffraction effects visible at grazing angles, and also gives to the sample a dichroic spectral behavior , i.e. polarization sensitive colors [4,5]. These optical effects are influenced by several physical parameters: the nanoparticle shape, size and spatial organization, as well

When It Is Not Only About Color: The Importance of Hyperspectral Imaging Applied to the Investigation of Paintings

Abstract: This paper illustrates some of the developments achieved in the field of non-contact analytical tools for two-dimensional polychrome artworks. It reports significant advantages within the application of hyperspectral imaging for high-quality documentation, accurate color reproduction, study of artists’ materials and techniques, and identification of past conservation treatments. In particular, Durer’s oil painting on panel, Adoration of the Magi (1504), was analyzed with a pushbroom hyperspectral imaging system in the visible-near-infrared range (Vis-NIR, 400–900 nm). The results obtained, including high-resolution color-accurate and false-color images, as well as high-resolution reflectance spectra, are reported and discussed, and the importance of hyperspectral imaging applied to the investigation of paintings is shown.

Fast-Calibration Reflectance-Transmittance Model to Compute Multiview Recto-Verso Prints

Abstract: Predicting simultaneously the spectral reflectance and transmittance of recto-verso prints is made easier thanks to a flux transfer matrix model. In the case where the printing support is symmetrical, i.e., its two sides are similar, the model can be calibrated from 44 halftone colors printed on the recto side, whose spectral reflectances and transmittance are measured. Color predictions are then allowed for any recto-verso halftone print illuminated on either side, and the inverse approach can be addressed: we can compute the digital layout for the recto and verso side that, once printed, can display different images according to the illuminated side.

iFAS: Image Fidelity Assessment

Abstract: image Fidelity Assessment (iFAS) is a software tool designed to assist image quality researchers providing easy access to a range of state-of-the-art measures which can be applied on a single pair of images and/or in a full database, as well as intuitive visualizations that aid data analysis, e.g., images and histograms of pixel-wise image differences, scatter plots and correlation analysis. The software is freely available for non-commercial use.

Video Smoke Removal Based on Smoke Imaging Model and Space-Time Pixel Compensation

Abstract: This paper presents a novel video smoke removal method based on a smoke imaging model and space-time pixel compensation. First, we develop an optical imaging model for natural scenes that contain smoke. Then, we remove the smoke in a video, frame-by-frame, based on the smoke imaging model and conventional dehazing approaches. Next, we align the smoke-removed frames using corresponding pixels. To obtain the corresponding pixels, we use SIFT and color features with distance constraints. Finally, to reproduce clear video appearance, we compensate pixel values by utilizing the space-time weightings of the corresponding pixels between the smoke-removed frames. Validation experiments show our method can provide effective smoke removal resulting in dynamic scenes.

Fourier Multispectral Imaging: Measuring Spectra, One Sinusoid at a Time

Abstract: We recently introduced the notion of Fourier Multispectral Imaging (Fourier MSI), a novel technique for undersampling spectral images without significant sacrifices to the spectral features that may be useful for material identification. The idea originated in Fourier transform spectroscopy, where multiple interferometric measurements are used to take spectral samples while varying optical path difference (OPD). Since interference is equivalent to spectral modulation by a filter with sinusoidally shaped transmittance function, we designed a sinusoidal filter using thin film Fabry-Perot to acquire OPD samples at every pixel. Owing to the rapid decay of OPD samples, Fourier MSI is an ideal multispectral imaging modality for preserving spectral features with the fewest spectral samples. We detail our prototyping efforts and demonstrate the advantages of such multispectral imaging configurations.

Color Vision Is a Spatial Process: The Retinex Theory

Abstract: Born through the work of Edwin H. Land and John J. McCann more than 40 years ago, Retinex theory proposes a computational model to explain and estimate the human color sensation, i.e. the color perception that human vision system produces when oberving a scene. Retinex is founded on a series of experiments, evidencing that the human color sensation at any observed point does not depend merely on the photometric cues of that point, but also on those of the surrounding regions and on their spatial arrangement. Indeed, human color vision is a spatial process. This paper presents the conceptual framework of Retinex, the main challenges it faced and solved, and some algorithmic procedures implementing it.

A Simple Scanner for High Resolution Imaging of Wall Paintings

Abstract: The increasing demand for high-resolution digital color imaging of cultural heritage assets such as mural paintings makes it necessary to build image acquisition systems that are easy to use in different conditions and sometimes in environments without sources of electric power. Here, we propose a simple structure of a scanning system in which the motion of the scanning head utilizes gravity as its motion mechanism instead of electricity. This system enables an area image sensor camera to move along the rail at constant speed without an electrical motor, but uses gravity instead. The camera mounted on the camera stage moves slowly at constant speed with continuous shooting mode. Using this method, it is possible to acquire many images and stitch them together to obtain a panoramic image of a big object at high resolution. This paper discusses the structure and application of this system that can acquire high-resolution images.

A Transmission Type Scanning System for Ultra High Resolution Scanning

Abstract: In an exceptional photography project, wall paintings of golden hall in Horyu-ji Temple were photographed on a one-to-one scale on glass dry plates 80 years ago. Unfortunately the temple burned in a fire and major parts of the wall paintings were destroyed. Because of the size of the glass plates, it is difficult to find a proper image grabbing systems to scan them with high resolution. Moreover, there were no color reference charts when the wall paintings were photographed, so it is not possible to create a reference data using a color chart. The method of creating the reference data without using color chart has not been systematized. Therefore, this paper gives a brief record of our project to develop a scanning system that is capable of digitizing these large glass plates with ultra high resolution and a method to reproduce the colours with high color definition from available information.

Computational Print Control

Abstract: Printing may seem like a dinosaur among today’s imaging technologies, since its roots stretch back to Becquerel’s work that lead to the first color photographs and the first mechanical color reproduction at the end of the nineteenth century. Ten years ago, we then made the fundamental discovery of a new print control domain, where instead of choices about colorant amounts that are akin to the effect of color filters used since the beginning of color printing, print can be specified by the probabilities of colorant combinations, the Neugebauer Primaries. This has led to the ability to print patterns that were previously inaccessible and consequently, by using large-scale computational optimization, to delivering more color gamut, greater ink use efficiency and greater sharpness and detail in print, while using the same materials and printing system as before. This keynote will present the basic principles of the HANS print control paradigm, review the highlights of results obtained using it to date and indicate its potential future developments.

Visualization of Subsurface Features in Oil Paintings Using High-Resolution Visible and Near Infrared Scanned Images

Abstract: High-resolution imaging is on the rise in the field of digital archiving of cultural heritage. Conventionally, this is accomplished by capturing images in the visible region. However, the visible region has a very narrow spectrum. In this study, around 100 oil paintings belonging to the Bridgestone Museum of Art in Tokyo, Japan have been digitized at both visible and near infrared region (i.e. ~400–700 nm and ~850 nm respectively) at 1000 dpi scanning resolution. Since materials behave differently when irradiated by a source with different wavelengths, the resulting images from visible and near infrared scans could reveal some under-drawings which are not visible from the naked eye. By applying false color image composition, it was possible to visualize subsurface features more easily. Different false color images were investigated by substituting the individual RGB channels of the visible images with NIR image to increase the optical contrast. Promenade (1926) by George Grosz was selected as a test case.

Visualizing Lost Designs in Degraded Early Modern Tapestry Using Infra-red Image

Abstract: This paper shows how to experimentally visualize lost designs in damaged early modern tapestries used in the Kyoto Gion festival. Unlike cloth weaving, tapestry is weft-faced weaving. As the surface welt threads become worn or turn over time, the design in a tapestry is gradually lost. On the other hand, weft threads hidden by warp threads still remain. In the tapestries of the Kyoto Gion festival, gold and silver threads were often used as weft, and they reflect infrared radiation. In experiments, a tapestry woven in the seventeenth century was used. Six-band images were taken for accurate color reproduction and infrared images were taken for visualizing the lost design. The viewing angle and image resolution of both types of images were the same. Superimposing the infrared image on color image after correcting registration errors revealed the original design of the tapestry.

A Novel Scanning Technique for Imaging of Gold and Silver Foils Used in Art Works

Abstract: Digital archiving has seen rapid growth in the recent years, accompanied by ever-increasing demands for high resolution and high-color definition color images. Despite the advances in imaging techniques, gold leaves and golden objects remain as some of the most difficult-to-image materials due to their highly reflective surfaces. Oversaturation commonly occurs when direct reflection is captured by an imaging element.