Showing papers in "Journal of Imaging Science and Technology in 2004"

A Theory of Image Quality: The Image Quality Circle

Abstract: A theory of image quality, called the Image Quality Circle, is presented and described in detail. It includes many of the well recognized elements of image quality, but arranges them in a way that makes the subject of image quality both understandable and complete. This tutorial report provides a comprehensive description of the underlying assumptions and rationale of the theory. An organizational view of the Image Quality Circle is also provided to facilitate the establishment of image quality goals and specifications.

Six band HDTV camera system for spectrum-based color reproduction

Abstract: We present the architecture of a six band high-definition television (HDTV) camera system newly developed for accurate color reproduction of motion pictures based on spectral information. The camera has an optical component connecting the objective lens to two conventional HDTV camera heads whose spectral sensitivities are individually adjusted by placing different interference filters between the objective lens and each camera. Evaluation of the accuracy of the color estimation obtained using the six band camera and a conventional RGB, i.e., three band, HDTV camera using simulated and experimentally obtained camera signals showed that the six band camera achieves accurate color estimation. In the experimental evaluation, the average color differences for the 24 color patches of the GretagMacbeth Color Checker for the six band and RGB camera signals were 1.43 and 4.12 delta-E a b . This system should thus be suitable for a wide variety of applications.

Six color printer characterization using an optimized cellular Yule-Nielsen spectral neugebauer model

Abstract: A printer characterization attempts to map, in both directions, corresponding points in colorant and colorimetric spaces. Two limiting approaches are used: analytical models based on a small number of samples, and direct measurement and interpolation requiring many samples. For six-color printers, the former approach often has insufficient accuracy whereas the latter approach requires an excessive number of samples. An intermediate approach was used to characterize a CMYKGO ink jet printer, the Cellular-Yule-Nielsen-Spectral-Neugebauer (CYNSN) model. This model included an optimized Yule-Nielsen n value and one-dimensional look-up tables between digital data and effective area coverage for each colorant. Each colorant was divided into three subspaces, or cells, requiring the selection of two intermediate values and fixed endpoints of 0% and 100% effective area coverage. An optimization was performed that determined these intermediate values by minimizing the maximum spectral error when one-colorant CYNSN models were used to predict 256-step ramps. This technique enabled a considerable reduction of the total number of required samples from several hundreds of thousands to 4,096, the required number of cellular Neugebauer primaries. Of these colors, only 1,024 could be printed; the remainder was non-printable due to inkblots. A third optimization synthesized the spectral properties of the non-printable cellular primaries using weighted spectral regression, the weighting a function of colorant-space location. The CYNSN model based on these three optimizations was able to predict 600 random colors sampling the colorimetric gamut to an average spectral RMS error of less than 0.5% and ΔE 0 0 of less than 1.0. The color gamut achievable using the synthesized spectra was 54% larger in colorant space and 15% larger in CIELAB space than that achievable when limiting the CYNSN model to printable cellular primaries.

Classical print reflection models: a radiometric approach

Abstract: The present contribution reviews the classical print prediction models of Williams and Clapper and of Clapper and Yule by pursuing a radiometric approach. The relationship between these models is established and the related respective contributions of Judd, Saunderson and Shore and Spoonhower are highlighted. Thanks to the radiometric approach, variants of the Williams and Clapper and Clapper and Yule models are derived in order to account for different measuring geometries (integrated sphere, 45°/0° geometry or other bistatic geometries) and for different reference white reflectors (coated paper, perfect white diffuser). The radiometric approach presented here gives a more profound understanding of these models and of the corresponding physical phenomena

Traveling wave transport of particles and particle size classification

Abstract: Experimental research was carried out on transport of particles and particle size classification in a traveling electrostatic field. Particle conveyors which consisted of parallel electrodes were constructed and a four phase traveling electrostatic wave was applied to the electrodes to transport particles on the conveyor. The following points were clarified by the experiment: (1) Particles were transported almost linearly with time. Transport rate was also linear with applied voltage but a threshold existed due to adhesion force. (2) The direction of particle transport did not always coincide with that of the traveling wave but it was in part changed depending on the frequency of the traveling wave, the particle diameter, and the electric field. Motion of particles at low frequency was nearly synchronized with the traveling wave but at medium frequency it was opposite to and slower than the wave. Particles were vibrated but not transported at high field frequency. (3) Particles were efficiently transported under conditions of high electrostatic field with a rectangular waveform. (4) Particles essentially moved along the electric flux line, but electrostatic interaction and particle-particle and particle-conveyor collisions made trajectories complex. (5) Particles were classified according to size under application of voltages of appropriate frequency.

Filter selection for multispectral color image acquisition

Abstract: The quality of a multispectral color image acquisition system depends on many factors, the spectral sensitivity of the different channels being one of them. In a relatively common setup, a multispectral camera is being implemented by coupling a monochrome digital camera with a set of optical filters, typically mounted on a filter wheel. The properties of these filters is an important component of the system design. Different methods have been proposed for the design or selection of appropriate filters. In this article we review several methods used for selection of an optimal subset of filters from a set of available filters. The different filter selection methods are subjected to a comprehensive evaluation procedure, in which their quality is evaluated mainly in terms of the ability of the resulting system to reconstruct scene spectral reflectances.

Water fastness of ink jet prints on modified conventional coatings

Abstract: In order to produce color ink jet prints with good fastness properties, the use of special, coated papers is generally essential. The use of modified conventional coating pigments has been suggested as a means to obtain amatte ink jet coating that would combine the good properties of silica-based ink jet coatings and conventional printing papers. This study examined the water fastness of ink jet prints using experimental coated papers containing modified PCC and kaolin pigments, and model inks with known compositions. The properties of the coatings were altered by using different ratios of coating pigments, and types of binder and dispersant. The water fastness of the prints was analyzed using conventional methods for measuring paper properties and print quality, combined with FTIR and Raman spectroscopy and multivariate data analysis methods. The results indicated that within the studied sample sets, primarily the chemical paper-ink interactions contributed to water fastness on cationic PVA-poly-DADMAC coatings, whereas on weakly cationic styrene-acrylate latex-starch coatings, structural paper properties were relevant as well. In general, increased impermeability of the coating appeared to be advantageous. With regard to the chemical paper-ink interactions, ionic bonding between the dye and the coating proved to be beneficial for water fastness, provided that the interacting coating component was insoluble to water.

Proximity Theory of Toner Adhesion

Abstract: Measurements of toner adhesion are usually at least one order of magnitude larger than predicted by image force calculations that model the toner charge by locating it in the center of the toner particle. In order to account for this discrepancy, it has been suggested that either the toner charge is not uniformly distributed on the surface or that van der Waals forces dominate toner adhesion. We propose an alternative model of toner adhesion, which takes into account a recent result that shows that there is an electrostatic proximity force of adhesion at the contact point between a toner particle and a conductive plane.

Why Multispectral Imaging In Medicine

Abstract: Multispectral imaging will certainly provide an excellent solution to color problems in medicine, and may have significant impacts in many aspects of medicine, so that the realization of individual improvement forms a complicated network. Specific reports on the concrete medical problems that will be solved by a specific multispectral technology will therefore be indispensable. It is also important to understand some specific principles are appropriate from a marketing viewpoint for the chaotic medical field. Based on these considerations, two promising medical applications of multispectral imaging are proposed; digital images with spectral reflectance for each pixel, and digital images that are very accurate reproductions of real objects. The first technology will lead to new morphological diagnostic methods more powerful than human visual perception alone, and possibly even to the discovery of the mechanism of human color recognition, while the second advance will lead to a major improvement in the diagnostic reliability of digital color images and wider medical adoption of digital technology. The former will require considerable investment but will provide significant improvements in diagnostic ability, particularly for rare diseases, while the latter will provide practical and general improvements in medicine at relatively low cost.

Colorimetric and spectral characterization of a color scanner using local statistics

Abstract: The accuracy of scanner characterization is usually affected by two factors. First, the statistics may vary for individual color samples and, second, the behavior of the scanner may depart from the linear reflectance model. This article proposes a method of colorimetric and spectral characterization for a color scanner using local statistics to deal with these problems. The experimental results showed that the proposed method significantly outperforms one using global statistics. The color accuracy of the colorimetric characterization is slightly better than that of the spectral one, while cross-media metamerism exists for both methods.

A micro-goniophotometer and the measurement of print gloss

Abstract: The focus of this report is on the development of an instrument for measuring gloss light. The word "gloss" is intuitively easy to understand, but making an optical measurement that correlates well with the perception of gloss remains an unsolved challenge. The hypothesis behind this project is that both spatial resolution (micro-) and angular resolution (gonio-) are required of an instrument in order to correlate meaningfully with visual perceptions of gloss and also with underlying causes of gloss. Thus, a micro-goniophotometric instrument was developed. The instrument employed plain polarized light to separate bulk, diffuse light from specular light. Specular light was defined instrumentally as the difference between the light captured with a polarizing filter orthogonal to and parallel to the direction of polarization of the incident light. In addition, it was assumed that the instrument must be able to account quantitatively for all of the gloss light from the surface distributed in all directions around the specular (equal/opposite) angle. The instrument developed in this project measures specular and diffuse light as a surface radiance distributed over angle a from the specular direction in the plane defined by the sample, detector, and illuminator. The orthogonal angle, β, is not scanned. Rather, all of the light distributed in the β direction is integrated at each angle a. The area under the bidirectional reflectance factor friction, BDRF, generated in this way accounts for all of the specular light distributed over both angles a, and β. Using this instrument, it was shown that the effects of surface roughness could be differentiated from the effects of variations in refractive index. In addition, the color of the light and the ink were measured and found to vary the amount of specular reflectance by as much as a factor of two. A mechanism is proposed to account for this.

Printing of dye and pigment-based aqueous inks onto porous substrates

Abstract: In this article, the dynamics of droplet impingement and absorption into microporous materials for pigment based aqueous inks and dye based aqueous inks are compared. For dye based inks it was shown earlier that three main phases could be resolved: inertial spreading, absorption, and evaporation of the liquid, leading to the final equilibrium condition on which the typical customer is evaluating the image quality. For the inertial spreading phase it could be shown that the spreading behavior is largely determined by the hydrodynamic properties, and is easily amenable to dimensionless analysis. The absorption phase could be well described by a capillary wicking process according to imbibition models. Evaporation is the slowest process only being finalized after many seconds. These results are now compared with droplet impingement and absorption of pigment based inks on microporous receivers. It is shown that these inks behave totally differently from dye based inks. Immediately after impingement and initial spreading the pigment particles start to coagulate on the surface of the microporous layer, creating a filter cake limiting the passage of carrier liquid. As a result much longer absorption times are observed and the equilibrium dot stays on top of the microporous layer. Most polymer stabilizers in the pigment based inks create a colored polymer layer having polymeric blend characteristics limiting considerably the penetration of water compared to the capillary wicking process. The capillary imbibition models are not valid any more because now the build-up of the filter cake changes not only the receding contact angle but also introduces a diffusion process changing as a function of time during the drying of the wet ink.

Application of image-based skin chromophore analysis to cosmetics

Abstract: The spatial distributions of melanin and hemoglobin in human skin can be determined by image-based skin chromophore analysis including independent component analysis (ICA) of a skin color image. The separation is based onthe skin color model in the optical density domain to quantify the change in the chromophores. In this paper, the analysis technique developed by Tsumura et al. was applied to many skin images, and the distribution of skin chromophores, such as melanin and hemoglobin, agreed well with the physiological knowledge. The effectiveness of cosmetic products was also evaluated by observing the changes in the amount of each chromophore. Finally a simulation to synthesize the changes in skin chromophores was performed to demonstrate its validity.

A Short Image Quality Model Taxonomy

Abstract: A short taxonomy of image quality models is proposed. Although the taxonomy includes medical and security image quality, its primary focus is commercial imaging, where the image quality judgment by the user is often cast as a "beauty contest" selection from images produced by competing products. By taking a broad view of "image quality models," it seems possible to arrange them into categories based on how they are applied in practice. This taxonomy illustrates and clarifies an array image of image quality models applicable to many practical applications.

Non-rigid registration based active appearance models for 3D medical image segmentation

Abstract: Active shape models and active appearance models are getting increasingly popular in medical image segmentation applications. However, they are not suitable for three-dimensional (3D) images in their original form. This is due to the underlying shape representation (a point distribution model, PDM), which becomes impractical in 3D. Recently, it was shown that nonlinear registration algorithms can assist in the automatic creation of a 3D PDM. Based on this idea, we built a 3D active appearance model of brain structures. The model extracts the mean texture and the image deformation variation information from the training set of images. A special benefit is the inclusion of an extended region of interest into the model, making it suitable for segmentation of structures with poorly defined edges. We evaluated the model by applying it to the task of automatic segmentation of the hippocampi from magnetic resonance brain images. We found high accuracy of the model, which is comparable to the accuracy of the underlying registration method. The main benefit of the model-based segmentation over the registration-based segmentation is time, which is reduced from many hours (for registering an atlas to the image) to only a few minutes (for fitting the model to the image).

Print Quality Test Page

Abstract: A print quality (PQ) test page is a diagnostic tool which is designed for accurate and effcient diagnosis of PQ defects We develop a set of test pages for the customer to check the condition of a printer without using any other tools We design test features which clearly reveal the common PQ defects that occur with color laser printers, such as failure of color plane registration, ghosting, and repetitive artifacts We then incorporate these test features into a set of test pages for a specific print mechanism architecture We introduce the concept of imperfections within specification which are PQ defects that fall within the PQ speci- fication of the printer The PQ test pages should be designed to not reveal such defects to the customer For the constant tone fields on the test page, we consider the specific case of designing halftone patterns that will not show fine pitch banding that is within specification Finally, we describe the design of a set of instruction pages that walk the customer through the process of identifying on the test pages PQ defects that can be remedied by customer replaceable parts

Crosslinkable branched hydrazones as hole transporting materials

Abstract: The new hole transport materials which comprise molecules consisting of two hydrazone branches linked by a central bridge containing a flexible thiophenyl sulfide and two hydroxyl groups were synthesized and investigated.These transporting materials are low molecular weight glasses and allow preparation of layers stable to crystallization. Ionization potential of the materials is in the range 5.03-5.38 eV. The highest hole mobility, reaching 10 - 4 cm 2 /Vs at 6 x 10 5 V/cm electric field, was observed in the transport material with dimethyltriphenylamine or triphenylamine moieties. These transport materials can be used with or, in the case of a solid substrate, without polymer binder. They can be chemically crosslinked in the layer by reaction of the hydroxyl groups with polyisocyanates.

Experimental verification of the proximity theory of toner adhesion

Abstract: Recently a new theory of toner adhesion was suggested. In this new theory an electrostatic proximity force of adhesion is taken into account. The proximity force is due to the attraction of charges on the toner particle in close proximity to a conductive plane to their respective image charges in the conductive plane. In this paper experimental verification of this new theory is presented using a 16 μm diameter, ground toner with silica additives. The theory is fit to complete curves of the developed and residual (1) mass per unit area, (2) charge-to-mass ratio, and (3) size distribution in an electric field detachment experiment, a much more stringent test of an adhesion theory than has been published before (to the authors' knowledge). The observed adhesion and ratio of the measured toner adhesion to theory are the lowest ever observed (to the authors' knowledge).

Spectral redundancy in a six-ink ink jet printer

Abstract: We demonstrate for the multi-ink printers investigated that there are many spectral reflectances that can be produced approximately by a single printer through a large variety of different ink combinations. This spectral redundancy was evaluated for a pair of six-ink ink jet CMYKGO printers. Through use of the lookup tables, density maps were built illustrating the six-dimensional distribution of redundancy throughout colorant space. A tolerance of 0.01 RMS showed none of the inks in our CMYKGO systems to be fully replaceable by combinations of the other inks. However, when the tolerance was relaxed to 0.02 RMS, the degrees of freedom for matching spectra in the systems fall to five because the five chromatic inks cover the entire spectral gamut without the need of the black ink. Systematic relationships among the inks are reported, detailing the likelihood that combinations of printer digital counts may be replaced by largely different ones while preserving spectral reflectance to within an RMS spectral reflectance factor tolerance.

A unified model of optical and physical dot gain in print color reproduction

Abstract: A unified model coping with both physical and optical dot gains on print color tone reproduction is proposed. The physical dot gain, is approximated by a quadratic function of nominal dot percentages. The function, for each color, is characterized by a single parameter depending on printing technologies as well as ink-paper interactions, and has a symmetric form around where the nominal dot percentage is 50%. The parameter can be derived from the best fit for the model to measured data, such as spectral reflectance values or CIEXYZ tristimulus values. Applications to a color laser printer (on office copy paper) reproduces the experimental dot gain curves fairly well. Dependence on physical dot percentage, a summation of the nominal dot percentage and the corresponding physical dot gain, results in the optical dot gain and in turn the overall dot gain asymmetric forms, plotted against the nominal dot percentages. Furthermore, theoretical analysis reveals fundamental differences between physical and optical dot gains. Therefore, effects of optical dot gain can not be accurately represented by any physical extension in printing practices.

A psychophysical experiment evaluating the color and spatial image quality of several multispectral image capture techniques

Abstract: Theoretically, there can be a compromise between color and spatial image quality for multispectral imaging: Increasing the number of channels increases spectral and colorimetric accuracy, that is, color quality increases;decreasing the number of channels reduces image noise and other spatial artifacts, that is, spatial image quality increases. Two paired comparison psychophysical experiments were performed to scale color and spatial image quality in order to better understand this compromise. Test targets, a watercolor painting, and several dioramas were imaged using three-, six-, and 31-channel image acquisition systems. One of the three-channel systems was a professional grade trichromatic digital camera; the other systems used the identical research grade sensor. For the six- and 31-channel images, both direct pseudoinverse based transformations and the use of principal component analysis were used to convert from digital to spectral data. The spectral data were used to render colorimetric images. Pseudoinverse transformations were used to convert the three-channel images to colorimetry. Twenty-seven observers judged, successively, color and spatial image quality of colorimetric images rendered for an LCD display compared with objects viewed in a light booth. The targets were evaluated under simulated daylight (6800K) and incandescent (2700K) illumination and the visual data were transformed to quality scales using Thurstone's law of comparative judgments, Case V. The first experiment evaluated color image quality. Under simulated daylight, the subjects judged all of the images to have the same color accuracy, except the professional camera image that was significantly worse. Under incandescent illumination, all the images, including the professional camera, had equivalent performance. The second experiment evaluated spatial image quality. The results of this experiment were highly target dependent. A subsequent image registration experiment showed that the results of the spatial image quality experiment were affected by image registration to some degree. For both experiments, there was high observer uncertainty and poor data normality. Dual scaling and a graphical analysis of observer response data were used as alternate techniques to Thurstone's Law. These techniques yielded similar results to the Thurstone-based quality scales. The uncertainty was caused by insufficient ambiguity between images. A simultaneous analysis of the color and spatial image quality results for the research grade sensor indicated that the most preferred image types were the 31-channel images. Thus, it is possible for multispectral images with many channels to achieve similar color and spatial image quality to systems with just a few channels. The theoretical compromise between color and spatial image quality as the number of channels increased was not observed under these experimental conditions.

Dependent color halftoning: Better quality with less ink

Abstract: When a color image is halftoned its color channels are normally halftoned independently. The dots in different channels are placed independent of each other and consequently the final result may not be of high quality even if a. well performing monochromatic halftoning method has been used. In this article we propose a method that halftones the channels of the color image in a context dependent manner. Since the yellow ink on a white paper is hardly visible, only cyan and magenta separations need to be halftoned dependently. We also show that dependent color halftoning not only increases the halftone image quality but also decreases the amount of ink needed to reproduce different colors.

Measurements of the Modulation Transfer Function of image displays

Abstract: Measurements of the Modulation Transfer Function (MTF) of image displays are often required for objective image quality assessments, but are difficult to carry out due to the need for specialized apparatus. This article presents a simple method for the measurement of the MTF of a sample CRT display system which involves the use of a still digital camera for the acquisition of displayed test targets. Measurements are carried out using, first, the sine wave method, where a number of artificial sine wave images of discrete spatial frequency and constant modulation are captured from a close distance. Fourier techniques are employed to extract the amplitude of the display signal from the resulting macroimages. In a second phase, displayed artificial step edges are captured, and the ISO 12333 SFR (Spatial Frequency Response) Slanted Edge plug-in is used for automatic edge analysis. The display MTF, in both cases, is cascaded from the closed-loop system MTF. The two measuring techniques produced matching results, indicating that under controlled test conditions accurate measurements of the display MTF can be achieved with the use of relatively simple equipment.

A review of the mechanism of photographic sensitivity and roles of silver clusters

Abstract: Recent progress in the mechanism of latent image formation and roles of silver clusters in photographic sensitivity is reviewed with respect to classification of silver clusters: comparison between light and reduction clusters; preferential formation of reduction sensitization R centers to P centers; difference between latent sub-image centers and P centers; characterization of R and P centers by photoconductivity measurements; utility of R centers with two electron sensitization; and direct observation of latent image and fog centers.

Illuminant influence on the reconstruction of near-infrared spectra

Abstract: In this study we analyze the influence of the illuminant on the reconstruction of spectral reflectance curves in the near infrared region of the spectrum. We have tested several methods based on multispectral imaging thatpermit us to obtain the spectral reflectance of samples from measurements performed with a CCD camera. Using numerical simulation, we have analyzed the number and shape of the optimum filters that need to be used in the acquisition channels in order to obtain good spectral reconstructions under a great number of different lighting conditions. Finally, we have studied the influence of the illuminant on the quality of the reconstruction using a set of commercially available filters which are similar to the optimum filters obtained from the simulations. Our results show that the reconstruction does not depend strongly on the illuminant used. This indicates that with the same set of filters we can obtain good reconstructions for different types of illuminant.

Training set selection for multispectral imaging system Characterisation

Abstract: Deriving the actual multispectral data from the output of the acquisition system is a key problem in the field of multispectral imaging. Solving it requires a characterization method and the training set (if any) on whichthe method relies. In this paper we propose three novel approaches in selecting a training set to be used for the characterization of a multispectral acquisition system. The first approach, which we call the Hue Analysis Method, is based on colorimetric considerations; the second and third approaches, which we call the Camera Output Analysis Method and the Linear Distance Maximization Method respectively, are mainly based on algebraic and geometrical facts. In all three cases the selected training sets will have relatively low numerosity and broad applicability. We also test our three approaches, as well as an approach from another author and a random selection method, on the data obtained from a real acquisition. We then compare the reconstructed reflectances with the measurements obtained using a spectrophotometer. Our results indicate that all our methods can be substantial improvements compared to a random selection of the training set, and that the performances of the Linear Distance Maximization Method make it the best choice among all the methods tried for application in a general context.

Electrophotographic printing on textiles and non-planar substrates

Abstract: Use of electrophotographic techniques to digitally produce images on textiles and non-planar substrates can be accomplished if the adhesion forces between the toner and the photoconductor are sufficiently reduced and/or that the forces between the toner to the receiver are greater than those adhering the toner to the photoconductor.

A continuous ink jet device on the basis of electrohydrodynamic mechanism

Abstract: A new type of ink jet system is proposed on the basis of electrohydrodynamic (EHD) effect. On the application of high DC electric fields to some insulating fluids, a fluid jet with a velocity of about 1 ms - 1 is generated from the positive electrode as a bulk flow. By controlling the EHD flow to issue across the free surface, the ink jet device is developed, of which the nozzle unit is simply composed of a needle electrode and capillary. When the fluids are subjected to high electric fields, the ink jet is generated above a critical threshold and the height increases with increasing voltage. The critical voltage strongly depends on the nozzle geometry. The height of ink jet at high voltages increases with decreasing viscosity and conductivity of fluids. The performance of ink jet based on the EHD jet can be improved by a combination of designs of nozzle geometry and material functions.

Verification of the predicting model and characteristics of dye-based ink jet printer

Abstract: For analyzing color reproduction of an ink jet printer, it is important to fully understand the relationship between amounts of dye placed on the paper and resultant colors. This insight will enable one to develop a simulation that can be used to design optimum subtractive color dyes. The purpose of the present study is to verify the predicting models that can predict tristimulus values from printed dye amounts, for use in evaluating the performance of dye-based ink jet printers (IJs). In the present study, five predicting models were compared for dye-based IJ, including the Neugebauer model, the Yule-Nielsen Neugebauer model, the Kubelka-Munk model, the Cellular Neugebauer model, and the Cellular Kubelka-Munk model. Further, the comparison between a dye-based IJ and photographic color paper was studied by means of a computer simulation. This investigation was carried out from the viewpoints of the stability of selective grays for illumination metamerism and of maximizing color gamut volumes. For a dye-base IJ, the effect of ink dilution was also studied. This study is an important step toward the development of simulations for use in improving image quality for dye-based IJ.

PCA transform in watermarking spectral images

Abstract: In this study we present a technique to embed a digital watermark containing copyright information into a spectral image. The watermark is embedded in a transform domain of the spectral dimension of the image. The transform domain is derived by performing the principal component analysis (PCA) transform on the original image. The watermark is embedded by modifying the coefficients of the eigenvectors from the PCA transform. After modification the image is reconstructed by the inverse PCA transform, thus containing the watermark. We provide analysis of watermark's imperceptibility and robustness against attacks with various parameter values in embedding and in attacks. The attacks include lossy image compression by the wavelet transform, median filtering and mean filtering. Experimental results indicate that the watermarked image is very similar to the original one and the watermark can be extracted with reasonable visual fidelity from the image after the attacks.