Showing papers presented at "Computational Color Imaging Workshop in 2015"

Data Fusion of Objects Using Techniques Such as Laser Scanning, Structured Light and Photogrammetry for Cultural Heritage Applications

Abstract: In this paper we present a semi-automatic 2D-3D local registration pipeline capable of coloring 3D models obtained from 3D scanners by using uncalibrated images. The proposed pipeline exploits the Structure from Motion (SfM) technique in order to reconstruct a sparse representation of the 3D object and obtain the camera parameters from image feature matches. We then coarsely register the reconstructed 3D model to the scanned one through the Scale Iterative Closest Point (SICP) algorithm. SICP provides the global scale, rotation and translation parameters, using minimal manual user intervention. In the final processing stage, a local registration refinement algorithm optimizes the color projection of the aligned photos on the 3D object removing the blurring/ghosting artefacts introduced due to small inaccuracies during the registration. The proposed pipeline is capable of handling real world cases with a range of characteristics from objects with low level geometric features to complex ones.

Does Color Influence Image Complexity Perception

Abstract: In this paper we investigate if color influences the perception of image complexity. To this end we perform two different types of psycho-physical experiments on color and grayscale images. In the first experiment, images are ranked based on their complexity (image ranking), while in the second experiment the complexity of each image is assessed on a continuous scale (image scaling). Moreover, we investigate if ten image features, that measure colors as well as other spatial properties of the images, correlate with the collected subjective data. The performance of these correlations are evaluated in terms of Pearson correlation coefficients and Spearman rank-order correlation coefficients. We observe that for each type of experiment, subjective scores for color images are highly correlated with those of the corresponding grayscale versions suggesting that color is not a relevant attribute in evaluating image complexity. Moreover none of the tested simple image features seem to be adapt to predict the image complexity according to the human judgments.

An Overview of Color Name Applications in Computer Vision

Abstract: In this article we provide an overview of color name applications in computer vision. Color names are linguistic labels which humans use to communicate color. Computational color naming learns a mapping from pixels values to color names. In recent years color names have been applied to a wide variety of computer vision applications, including image classification, object recognition, texture classification, visual tracking and action recognition. Here we provide an overview of these results which show that in general color names outperform photometric invariants as a color representation.

Descriptor-Based Image Colorization and Regularization

Abstract: We propose a new, fully automatic method for example-based image colorization and a robust color artifact regularization solution. To determine correspondences between the two images, we supplement the PatchMatch algorithm with rich statistical image descriptors. Based on detected matches, our method transfers colors from the reference to the target grayscale image. In addition, we propose a general regularization scheme that can smooth artifacts typical to color manipulation algorithms. Our regularization approach propagates the major colors in image regions, as determined through superpixel-based segmentation of the original image. We evaluate the effectiveness of our colorization for a varied set of images and demonstrate our regularization scheme for both colorization and color transfer applications.

Adaptive Filters for Color Images: Median Filtering and Its Extensions

Abstract: In this paper we are concerned with robust structure-preserving denoising filters for color images We build on a recently proposed transformation from the RGB color space to the space of symmetric \(2\times 2\) matrices that has already been used to transfer morphological dilation and erosion concepts from matrix-valued data to color images We investigate the applicability of this framework to the construction of color-valued median filters Additionally, we introduce spatial adaptivity into our approach by morphological amoebas that offer excellent capabilities for structure-preserving filtering Furthermore, we define color-valued amoeba M-smoothers as a generalization of the median-based concepts Our experiments confirm that all these methods work well with color images They demonstrate the potential of our approach to define color processing tools based on matrix field techniques

First Stage of a Human Visual System Simulator: The Retina

Abstract: We propose a configurable simulation platform that reproduces the analog neural behavior of different models of the Human Visual System at the early stages. Our software can simulate efficiently many of the biological mechanisms found in retina cells, such as chromatic opponency in the red-green and blue-yellow pathways, signal gathering through chemical synapses and gap junctions or variations in the neuron density and the receptive field size with eccentricity. Based on an image-processing approach, simulated neurons can perform spatiotemporal and color processing of the input visual stimuli generating the visual maps of every intermediate stage, which correspond to membrane potentials and synaptic currents. An interface with neural network simulators has been implemented, which allows to reproduce the spiking output of some specific cells, such as ganglion cells, and integrate the platform with models of higher brain areas. Simulations of different retina models related to the color opponent mechanisms, obtained from electro-physiological experiments, show the capability of the platform to reproduce their neural response.

Representation of Cultural Objects by Image Sets with Directional Illumination

Abstract: In a dome illumination system, many different images can be captured in pixel register from the same viewpoint, each illuminated from a different direction. This is a much richer representation than a single image, and has many applications in cultural heritage for the digitising and display of objects that are flattish with surface relief, such as coins, medals, fossils, rock art, incised tablets, bas reliefs, engravings, canvas paintings, etc. The image sets can be used in three ways: (1) visualisation by interactive movement of a virtual light source over the enclosing hemisphere; (2) 3D reconstruction of the object surface; (3) modelling of the specular highlights from the surface and hence realistic rendering.

Evaluation of the Colorimetric Performance of Single-Sensor Image Acquisition Systems Employing Colour and Multispectral Filter Array

Abstract: Single-sensor colour imaging systems mostly employ a colour filter array (CFA). This enables the acquisition of a colour image by a single sensor at one exposure at the cost of reduced spatial resolution. The idea of CFA fit itself well with multispectral purposes by incorporating more than three types of filters into the array which results in multispectral filter array (MSFA). In comparison with a CFA, an MSFA trades spatial resolution for spectral resolution. A simulation was performed to evaluate the colorimetric performance of such CFA/MSFA imaging systems and investigate the trade-off between spatial resolution and spectral resolution by comparing CFA and MSFA systems utilising various filter characteristics and demosaicking methods including intra- and inter-channel bilinear interpolation as well as discrete wavelet transformed based techniques. In general, 4-band and 8-band MSFAs provide better or comparable performance than the CFA setup in terms of CIEDE2000 and S-CIELAB colour difference. This indicates that MSFA would be favourable for colorimetric purposes.

Accuracy in Colour Reproduction: Using a ColorChecker Chart to Assess the Usefulness and Comparability of Data Acquired with Two Hyper-Spectral Systems

Abstract: Hyper-spectral imaging has been applied as an in situ technique for the study and accurate digital documentation of coloured artworks. Providing spectral and colorimetric characterisation across the entire surface of an object, it can be used to identify the coloured materials, measure colour changes, and document it with high fidelity. However, depending on the system used, data accuracy and reliability may vary. In this work, developed within the Round Robin Test being carried out by COSCH Working Group 1, an X-Rite® ColorChecker Classic chart was analysed with two push-broom hyper-spectral systems developed by different groups (IFAC-CNR and IP-UEF), in the 400-1000 nm range, and the data obtained were compared. This comparison allowed to assess the accuracy of colour reproduction processes performed by the two systems. The results obtained are satisfactory in terms of spectral and colorimetric accuracy for some colours, but show differences at both ends of the visible range.

Color and Spectral Mixings in Printed Surfaces

Abstract: The present paper discusses the concept of subtractive color mixing widely used in color hardcopy applications and shows that a more realistic concept would be “spectral mixing”: the physical description of the coloration of light by printed surfaces comes from the mixing of light components selectively absorbed by inks or dyes during their patch within the printing materials. Some classical reflectance equations for continuous tone and halftone prints are reviewed and considered as spectral mixing laws. The challenge of extending these models to new inkless printing processes based on laser radiation is also addressed.

High-Resolution Multispectral Scanning for Mesoscopic Investigation of Discoloration of Traditional Japanese Pigments

Abstract: This study describes a non-invasive analytical imaging scanning technique utilizing multispectral images to study discoloration and degradation of pigments used in traditional Japanese artworks. The images have high spatial resolution which can achieve mesoscopic resolution (typically 0.1mm-10mm). Since the images are being scanned line by line instead of being recorded frame by frame, this enables accurate color and spectral recording of the material response from visible and near infrared irradiation. The multispectral images were used to reconstruct color information and spectral reflectance. The mathematical model is based on the Moore-Penrose pseuodoinverse. Using mesoscopically resolvable images, it is possible to measure the spectral reflectance of pigments ranging from μm-mm ROI unlike conventional spectrometers that requires big sampling area. The significance of mesoscopically-resolved analysis is demonstrated by investigating the discoloration and degradation of natural and artificial Japanese pigments. The pigments were heated in air at 300°C and sampled every 10 minutes. It was observed that the pigments discolored at seemingly random clusters. The reconstruction of the spectral reflectance at different sizes of ROI reveals strong correlation with background reflection. The size of the initial discoloration sites makes it impossible to measure using conventional spectrometers. It was observed that by using sub-mm ROI, it is possible to observe reflection and absorption patterns in the pigments which does not register with mm-scale ROI. The results have shown that mesoscopically resolvable multispectral images can be used effectively to study degradation and discoloration in pigments.

Estimating the Colors of Paintings

Abstract: Observers can adjust the spectrum of illumination on paintings for optimal viewing experience But can they adjust the colors of paintings for the best visual impression? In an experiment carried out on a calibrated color monitor images of four abstract paintings obtained from hyperspectral data were shown to observers that were unfamiliar with the paintings The color volume of the images could be manipulated by rotating the volume around the axis through the average (a*, b*) point for each painting in CIELAB color space The task of the observers was to adjust the angle of rotation to produce the best subjective impression from the paintings It was found that the distribution of angles selected for data pooled across paintings and observers could be described by a Gaussian function centered at 10o, ie very close to the original colors of the paintings This result suggest that painters are able to predict well what compositions of colors observers prefer

Can Linear Data Projection Improve Hyperspectral Face Recognition

Abstract: This paper investigates if the performance of hyperspectral face recognition algorithms can be improved by considering 1D projections of the whole spectral data along the spectral dimension. Three different projections are investigated: single spectral band selection, non-negative spectral band combination, and unbounded spectral band combination. Experiments are performed on a standard hyperspectral dataset and the obtained results outperform seven existing hyperspectral face recognition algorithms.

Development of a Dynamic Relighting System for Moving Planar Objects with Unknown Reflectance

Abstract: Relighting is a technique to modify an image to account for alternate illumination conditions. Conventionally, the reflectance characteristics of an object are provided, and a relighted image is calculated using input light source information. In this paper, we propose a dynamic relighting system for moving planar objects with unknown reflectance. By acquiring the surface spectral reflectance of moving objects, our system is able to reproduce accurate colors on a display device. In the reflectance acquisition, we use a programmable light source that can produce any spectral curve. The surface spectral reflectance of an object in a darkroom is obtained based on the lighting technique with five spectral basis functions that are generated by the programmable light source. The acquired reflectance and user-input illumination information are used to calculate accurate CIEXYZ values of the relighted image. Finally, CIEXYZ values are accurately transformed to RGB values. In the experiment, illuminant A and D65 are used as the illuminants for relighting. As a result, by comparing a computer simulation with actual experiments with real objects, we observe an average color difference \( \Delta {\text{E}}^{*}_{\text{ab}} \) of approximately 7. This system operates at a rate of 1 frame per second. In addition, in this study, we have implemented another relighting system for objects under an environmental lighting condition by determining the spectral power distribution of the illumination source.

The display gamut available to simulate colors perceived by anomalous trichromats

Abstract: The aim of this work was to investigate the effect on a display gamut of varying the optical density and the position of the maximum sensitivity of the cones spectra of anomalous trichromatic observers. The anomalous cone spectral sensitivities were estimated for a set of varying optical density and maximum sensitivity spectra conditions and used to compute the display color gamut. The computed display gamut simulated for normal observers the chromatic diversity perceived by anomalous observers. It was found that even small variations on the optical density and on the position of the maximum sensitivity spectra have an impact on the simulations of the display gamut for anomalous observers. It was also found that simulations for deuteroanomalous observers are the ones with greater impact if the estimation of the corresponding color display gamut is not carefully adjusted for the observer.

Spectral Impulse Noise Model for Spectral Image Processing

Abstract: The performance of an image processing algorithm can be assessed through its resulting images. However, in order to do so, both ground truth image and noisy target image with known properties are typically required. In the context of hyperspectral image processing, another constraint is introduced, i.e. apart from its mathematical properties, an artificial signal, noise, or variations should be physically correct. Deciding to work in an intermediate level, between real spectral images and mathematical model of noise, we develop an approach for obtaining suitable spectral impulse signals. The model is followed by construction of target images corrupted by impulse signals and these images will later on be used to evaluate the performance of a filtering algorithm.

The Color Logarithmic Image Processing (CoLIP) antagonist space and chromaticity diagram

Abstract: The CoLIP framework defines a vectorial space for color images It is in accordance with the theories of visual perception (Weber, Fechner) as well as Hering’s trichromacy theory It is mathematically well defined and computationally usable This article recalls the fundamentals of the LIP framework for graytone images, and introduces the elementary operations of the vectorial structure for color images It illutrates the representation of the chromaticity diagram with color modification application, namely white balance correction and color transfer The results show that the hull of the diagram are not modified, but the colors are

Optics and Computational Methods for Hybrid Resolution Spectral Imaging

Abstract: The concept and computational methods for hybrid resolution spectral imaging (HRSI) are introduced. In HRSI, a high-resolution spectral image is reconstructed with combining a high-resolution RGB image and a low-resolution spectral image. An important difficulty in high-resolution spectral imaging is that the light-energy is reduced at the image sensor. Such problem can be solved by the hybrid resolution approach, since the image resolution and quality are mostly determined by the high-resolution RGB image, which can be captured by commercial high-performance cameras. Different reconstruction methods suitable for a hybrid resolution system are reviewed and the performance of those methods is discussed. The hybrid resolution spectral video system is also demonstrated.

Relative Pose Estimation and Fusion of 2D Spectral and 3D Lidar Images

Abstract: This paper presents a unified approach for the relative pose estimation of a spectral camera - 3D Lidar pair without the use of any special calibration pattern or explicit point correspondence. The method works without specific setup and calibration targets, using only a pair of 2D-3D data. Pose estimation is formulated as a 2D-3D nonlinear shape registration task which is solved without point correspondences or complex similarity metrics. The registration is then traced back to the solution of a non-linear system of equations which directly provides the calibration parameters between the bases of the two sensors. The method has been extended both for perspective and omnidirectional central cameras and was tested on a large set of synthetic lidar-camera image pairs as well as on real data acquired in outdoor environment.

Experimental Analysis for Modeling Color of Halftone Images

Abstract: Reflectance models such as the monochrome Murray–Davies (MD) and the Neugebauer color equations make inaccurate predictions owing to changes in reflectance or tristimulus values (TSVs) of halftone dots and the paper between the dots. In this paper, we characterize the change of micro-TSVs as a function of printed area in spectral halftone image by a power function and compare its prediction efficiency using theoretically and experimentally measured limiting TSVs assuming dots of uniform thickness. We found that experimentally accounting for dot thickness variations as solid and mixed areas more precisely explained the single-model parameter that captured the observed lateral light scattering effect. The results showed that incorporating empirically modeled TSVs of the dots and the paper between dots, as well as introducing a new term addressing mixed area in the MD equation, produced CIE ∆\( {\text{E}}_{\text{ab}}^{*} \) in the range 1.22–1.76, and the overall gain was more than 1 ∆\( {\text{E}}_{\text{ab}}^{*} \).

Visual Search for Normal Color and Dichromatic Observers Using a Unique Distracter Color

Abstract: It is well known that color coding facilitates search and identification in real-life tasks. The aim of this work was to compare reaction times for normal color and dichromatic observers in a visual search experiment. A unique distracter color was used to avoid abnormal color vision vulnerability to background complexity. Reaction times for normal color observers and dichromats were estimated for 2\(^\circ \) central vision at 48 directions around a white point in CIE \(L^*a^*b^*\) color space for systematic examination on the mechanisms of dichromatic color perception. The results show that mean search times for dichromats were twice larger compared to the normal color observers and for all directions. The difference between the copunctual confusion lines and the confusion direction measure experimentally was 5.5\(^\circ \) for protanopes and 7.5\(^\circ \) for deuteranopes.

Collection, Analysis and Representation of Memory Color Information

Abstract: Memory color plays an important role in the perceptual process. The aim of this research is to collect, analyze and represent memory color data for certain natural scenes objects: sky, grass and tree leaves. To emphasize reliable data collection, we consider several sources: (a) psychophysical experiment; (b) multispectral image; (c) standard image database and (d) random image collection. Moreover, we consider different daylight conditions and locations. We perform an in-depth analysis of the collected information in the CIE-xy chromaticity space and present the natural scene objects as a memory color ellipse or polygon. Finally, we demonstrate a potential use of the collected information for natural image segmentation and enhancement.