We studied the preference judgment of pictorial images by image experts and naive observers. We first asked image experts to improve pictorial images the way they preferred. Then, we showed the different versions of each image to naive observers and asked them which version they preferred. To enhance an image, an expert divides it into large areas of interest, which mainly correspond to natural colors. To assess their preference judgments, naive observers principally focus on natural colors like sky, skin, or grass when present. A closer analysis of the digital image files showed that the segmentation process used by the experts allows to apply different corrections on the different objects. We used the previous work on memory colors by Yendrikhovskij and we showed that, to enhance an image, an expert moves the color space coordinates of identified zones towards those of memory colors corresponding to the objects being represented. The expert also follows some rules: the corrections must be plausible inside each segment and for the whole image, in relation with the illuminant of the scene. The images are accepted by observers in relation with the presence of memory colors and when the treatment of the whole image seems coherent.

The field of radiology dates back to 1896 with the first direct x-ray exposure of film by Roentgen. For the next 80 years the exposure of black-and-white film by x-rays, and later, from calculated images from computed tomography scans, ultrasound scans, and magnetic resonance imaging scans would dominate the hardcopy world of medical imaging. The use of color was largely experimental and limited until the development of real-time color Doppler imaging in the early 1980s. Since that time, the use of color has grown rapidly as three-dimensional visualizations and multimodality or multispectral images become widely utilized. The rapid growth of imaging techniques that combine anatomical information with additional functional or molecular information is driving color to the forefront, since additional information needs to be fused in the renderings. Thus, 100 years after Roentgen's experiment, a century of monochrome imaging is giving way to an emerging need for color displays of medical images.

Traditionally, hardware for additive color displays, including projection devices, has been built from a set of only three primaries: a red, a green, and a blue. Recently, some manufacturers of projector displays have designed their hardware to project a fourth primary, a white. This fourth primary has been helpful in increasing the luminous output possible from these displays. Because interdevice color communication infrastructure is based on red, green, and blue channels (RGB), the four-primary devices accept RGB digits and internally convert to red, green, blue, and white channels (RGBW). From a color management viewpoint, the four-color projectors look like RGB devices, but the typical color characterization models fail owing to the complexity introduced by the hidden RGB to RGBW conversion. Several four-primary digital light processing projectors were investigated and a new characterization model is proposed that approximately accounts for the relationship between RGB digital counts and resultant projected colorimetry.

This paper proposes a method of six-color separation using additional colorants and a quantitative granularity metric to reduce the color difference and graininess. In conventional six-color separation, the use of light magenta and light cyan is generally an effective way of reducing the visibility of dots in bilevel printing devices, i.e., light magenta and light cyan are used in a bright region instead of magenta and cyan. However, the hue values for light magenta and light cyan differ from those for magenta and cyan in CIELAB space, making the colorimetric reproduction less accurate. Therefore, to minimize this inaccuracy, the proposed method uses yellow and light magenta colorants as additional colorants. In a bright region, magenta is replaced with light magenta and yellow, while cyan is replaced with light cyan and light magenta. These combinations reduce the hue difference, as they create colors with a similar hue to magenta and cyan. In addition, a smooth image can simultaneously be obtained owing to the lower dot visibility of the additional colorants. In a middle region, magenta is replaced with light magenta and magenta, while cyan is replaced with light cyan and cyan. Since the use of these two colorants with different concentrations makes a coarse dot pattern, this phenomenon is reflected based on a quantitative granularity metric. Finally, in a dark region, only the magenta and cyan colorants are used as usual. Experiments show that the proposed method can simultaneously produce a colorimetric and smooth tone reproduction.

Multispectral printer characterization requires an effective model to map the inputs to the printer (i.e., the digital counts of the inks) into reflectance spectra and vice versa. Most of the methods for printer modeling are based on the color mixing model of Neugebauer, but this model, in its original formulation, is a rather poor predictor of the printer's output, since it fails to take into account many of the relevant phenomena that take place in the printing process. These phenomena, which include light scattering within the substrate, internal and surface reflection, and ink spreading, determine an enlargement of ink drops called dot gain, which differs on the basis of the substrate condition. This paper presents a novel strategy to model dot gain and interaction among inks in the definition of a printer model based on the Yule-Nielsen spectral Neugebauer equation. The method proposed has been designed for a four-ink ink jet printer, but its formulation is general and may be extended to the characterization of devices having more than four inks. Our method requires the definition of a relatively large number of parameters, that we estimate using genetic algorithms. The model has been tested on two different printers: An Epson StylusColor™ 740 ink jet printer and an Epson StylusPhoto™ 890 ink jet printer. Using a data set consisting of 777 samples, regularly distributed in the HSV color space, we have obtained an accuracy in terms of mean root mean squared error of 0.59% and of 1.54 ΔE^*_ab for the first printer and of 1.02% and of 2.04 ΔE^*_ab for the second printer. With respect to an approach based on a single dot gain function for each ink, our approach based on many dot gain functions reduced the average root mean square error on the test set of about 40% on average.

The surface reflectance functions of natural and manmade surfaces are invariably smooth. It is desirable to exploit this smoothness in a multispectral imaging system by using as few sensors as possible to capture and reconstruct the data. In this paper we investigate the minimum number of sensors to use, while also minimizing reconstruction error. We do this by deriving different numbers of optimized sensors, constructed by transforming the characteristic vectors of the data, and simulating reflectance recovery with these sensors in the presence of noise. We find an upper limit to the number of optimized sensors one should use, above which the noise prevents decreases in error. For a set of Munsell reflectances, captured under educated levels of noise, we find that this limit occurs at approximately nine sensors. We also demonstrate that this level is both noise and dataset dependent, by providing results for different magnitudes of noise and different reflectance datasets.

A method is presented for deriving a conversion from reflectance spectra to a convenient intermediate form introduced as LabPQR. LabPQR is designed for use within a spectral color management system as an interim connection space (ICS) between a reflectance based profile connection space (PCS) and output device digit counts. The LabPQR ICS makes use of a spectral encoding that explicitly incorporates colorimetry. The initial three dimensions of a LabPQR ICS provide a colorimetric representation of reflectance spectra under an illuminant. Additional dimensions define spectral corrections that allow inverse transformation to approximately the original spectra. Consistent with a previously defined spectral color management transformation chain, the PQR dimensions of the ICS can be optimally formulated to suit any specific output device’s spectral rendering capabilities. After describing a method for distilling spectra into, and reconstituting spectra, from LabPQR coordinates, several sample transformations are demonstrated using various output devices. Visualizations follow of some sample LabPQR gamuts. Observations are made, and from these observations, a possible method for performing spectral gamut mapping is proposed. Current methods of color management (specifically using ICC profiles) are discussed relative to LabPQR visualization. In conclusion, proposals are made for future research and possibilities.

The article aims to provide a solution for multispectral image compression for high color reproducibility with preservation to spectral accuracy. In the method previously proposed to reduce the colorimetric error of the reconstructed multispectral image, a weighting matrix is incorporated to Karhunen-Loeve transform (KLT) as the spectral transform for multispectral image compression, which accounts for the color matching functions of human observers as well as the viewing illuminants. However, the colorimetric improvements are obtained on the cost of degradation of spectral accuracy. In this paper, we show that the reduction of colorimetric error and the preservation of spectral accuracy is a tradeoff that can be controlled by adding a diagonal matrix that is composed of a scalar multiple of an identity matrix to the weighting matrix of KLT. As the result, the small values in the weighting matrix can be lifted up, thus reduce the spectral errors in the corresponding reconstructed multispectral image bands. We implement a multispectral image compression system that integrates the proposed spectral transforms with the addition of diagonal matrix and JPEG2000 for high colorimetric and spectral reproducibility. Experimental results for three 16-band multispectral images show that spectral accuracy can be improved without loss of substantial color reproducibility if the magnitude of the scalar in the diagonal matrix is chosen appropriately

A parameter optimized simple matrix (POSM) model is proposed for characterizing the liquid crystal display (LCD) monitor with variant primary chromaticity and primary crosstalk. This method is the same as the conventional simple matrix (CSM) model except that black point, white point, one-primary measurement data, and two primary measurement data are taken as the training samples for optimizing the signal nonlinear transformations (SNTs) and chromaticity matrix. As the configuration of the POSM model is the same as the CSM model, its backward model is simple and can be implemented for video applications. Polynomial function is taken as the SNT. The optimization process comprises the simulated annealing method for optimizing the coefficients of SNTs and the regression method for calculating the chromaticity matrix. The results show that the average color difference (CIEDE2000) of 224 random test samples for two characterized LCD monitors are 2.02 and 1.33 with a forward POSM model, and are 2.32 and 1.27 with a backward POSM model. The primary crosstalk of the LCD monitor with a higher average color difference is more serious than the other monitor. The performance of the POSM model is much better than the CSM model. For the monitor with lower crosstalk, the POSM model is better than the three-dimensional look-up table (3D-LUT) model with a 5 × 5 × 5 lattice but is worse than the 3D-LUT model with a 8 × 8 × 8 lattice.

When analog red, green, blue (RGB) signals of a video/graphic image need to be displayed on pixelated display devices, graphic digitizers have to be utilized to convert the analog signals to digital signals. In such an application choosing correct sampling frequency for the digitizers is essential since an incorrect frequency will impair the recovery of images encoded in the analog signals. This article presents an algorithm that searches automatically for the sampling frequency. The phase of the sampling clock is also critical since inadequate phase control can create undesirable visible artifacts. Thus, a second algorithm for automatically searching for the appropriate sampling phase is presented in this article as well. These algorithms can be applied in pixelated display applications.