Search

This study investigates the classification performance of elastic and collagen fibers using H&E stained hyperspectral images. As many as 1200 sample pixels were trained by using Linear discriminant analysis (LDA) and support vector machine (SVM) methods for reduction and classification. The classification result both using LDA and SVM show that H&E stained hyperspectral images performed better classification than H&E stained RGB image by comparing the classification of EVG stained images visually and quantitatively.

In dealing with medical information of patients through the network, high-level security is required. In Japan, the Ministry of Health, Labor and Welfare has issued the Security Guidelines for Health Information Systems which stipulates that proper measures should be taken for the network to preserve the patient data. The guideline also state that the management of the network must clarify demarcation of responsibility between network service providers including user. To keep the security regulation in communicating with external organizations, many medical institutions use several managed-VPN service lines separately according to the usages. It is because to avoid the risk of forming a virtual detour which can be used to leak the patient data. This makes the hospital bear the cost and inconvenience. To solve this problem, we propose a new secure network implementation method using single-line VPN network service for small-scale medical institution by arranging OpenFlow control technology with Healthcare PKI. The network works over a dynamic on-demand VPN with control based on the user certification responding their qualification, such as a medical doctor, written in the HPKI digital certificates. We developed a simple network system adopting our new method, and confirmed the validity and the effectiveness of our idea through the experiment.

Recently, the application field of biometric authentication is expanding from small-scale service to wide-scale one such as big event venue, theme park, and so on. In these situations, high-speed calculation and smooth operation are required because of a large number of users. In order to apply finger vein authentication to above services, a walkthrough type finger vein authentication is effective. In our previous study, a hand waving finger vein authentication system was proposed [1]. However, it is hard for this system to achieve high-speed calculation and highly accurate authentication simultaneously. In this study, we propose an improved pattern matching method that make it possible to accelerate the calculation of the finger vein authentication by applying Scale-Invariant Feature Transform (SIFT).

It is important to ensure the security of biometric authentication information, because its leakage causes serious risks, such as replay attacks using the stolen biometric data, and also because it is almost impossible to replace raw biometric information. In this paper, we propose a secure biometric authentication scheme that protects such information by employing an optical data ciphering technique based on compressed sensing. The proposed scheme is based on two-factor authentication, the biometric information being supplemented by secret information that is used as a random seed for a cipher key. In this scheme, a biometric image is optically encrypted at the time of image capture, and a pair of restored biometric images for enrollment and verification are verified in the authentication server. If any of the biometric information is exposed to risk, it can be reenrolled by changing the secret information. Through numerical experiments, we confirm that finger vein images can be restored from the compressed sensing measurement data. We also present results that verify the accuracy of the scheme.

Hybrid-resolution multispectral imaging is a framework to acquire multispectral images through a reconstruction procedure using two types of measurement data with different spatial and spectral resolutions. In this paper, we propose a new method for such a framework on the basis of a full-resolution RGB image and the data obtained from an image sensor with a multispectral filter array (MSFA). In the proposed method, a small region of each image band is reconstructed as a linear combination of RGB images, where the weighting coefficients are determined using MSFA data. The effectiveness of the proposed approach is shown by simulations using spectral images of natural scenes.

This paper presents multispectral image coding methods based on visible component separation techniques. By coding the visible components separately from other spectral components, the color information can be used without decoding the whole multispectral image. Two types of separation techniques are introduced: one is a direct separation of visible and invisible layers, while the other is a scalable separation. In the scalable separation, visible components are first compressed to produce the base layer bitstream. The enhancement layer data include not only invisible components but also the information on the distortion of the coded visible components. Such a separation technique improves the accuracy of decoded multispectral images especially when visible components are highly compressed. The performance of the proposed methods was evaluated for different bit allocations to visible components. In general, the coding performance reduces when data are divided and coded separately to incorporate scalabilities. Nonetheless, it is found that the proposed scalable approach realizes a peak signal-to-noise ratio PSNR value comparable to that achieved by employing JPEG2000, which incorporates no scalability.

With the widespread use of commercialized wide-gamut displays, the demand for wide-gamut image content is increasing. To acquire wide-gamut image content using camera systems, color information should be accurately reconstructed from recorded image signals for a wide range of colors. However, it is difficult to obtain color information accurately, especially for saturated colors, if conventional color cameras are used. Spectrum-based color image reproduction can solve this problem; however, bulky spectral imaging systems are required for this purpose. To acquire spectral images more conveniently, a new spectral imaging scheme has been proposed that uses two types of data: high spatial-resolution red, green, and blue (RGB) images and low spatial-resolution spectral data measured from the same scene. Although this method estimates spectral images with high overall accuracy, the error becomes relatively large when multiple different colors, especially those with high saturation, are arranged in a small region. The main reason for this error is that the spectral data are utilized as low-order spectral statistics of local spectra in this method. To solve this problem, in this study, a nonlinear estimation method based on sparse and redundant dictionaries was used for spectral image estimation—where the dictionary contains a number of spectra—without loss of information from the low spatial-resolution spectral data. The estimated spectra are represented by a mixture of a few spectra included in the dictionary. Therefore, the respective feature of every spectrum is expected to be preserved in the estimation, and the color saturation is also preserved for any region. Experiments performed using the simulated data showed that the dictionary-based estimation can be used to obtain saturated colors accurately, even when multiple colors are arranged in a small region. © 2011 Wiley Periodicals, Inc. Col Res Appl, 2013

This paper proposes a class-based spectral estimation method for high-resolution red, green, and blue (RGB) images and corresponding low-resolution spectral data. Each spectrum in the low-resolution data is assumed to be a mixture of spectra of different classes. Then, the spectral estimation matrix for every class is derived using a regression approach, in which the clustering results of the high-resolution RGB image are used to incorporate spectral unmixing. Experiments confirm reduced normalized root mean squared error for the spectral images if the number of classes in the clustering is appropriately selected. In addition, the experimental results show that the spectra are accurately reconstructed even when they are observed as mixed spectra in the low-resolution data.

We have shown that the application of double random phase encoding (DRPE) to biometrics enables the use of biometrics as cipher keys for binary data encryption. However, DRPE is reported to be vulnerable to known-plaintext attacks (KPAs) using a phase recovery algorithm. In this study, we investigated the vulnerability of DRPE using fingerprints as cipher keys to the KPAs. By means of computational experiments, we estimated the encryption key and restored the fingerprint image using the estimated key. Further, we propose a method for avoiding the KPA on the DRPE that employs the phase retrieval algorithm. The proposed method makes the amplitude component of the encrypted image constant in order to prevent the amplitude component of the encrypted image from being used as a clue for phase retrieval. Computational experiments showed that the proposed method not only avoids revealing the cipher key and the fingerprint but also serves as a sufficiently accurate verification system.

We are currently developing an imaging system that combines simultaneous positron emission tomography (PET) with near-infrared (NIR) optical tomography, thus supporting two different types of molecular imaging. For this system, we are considering whether to use depth of interaction (DOI) PET detectors as simultaneous detectors of gamma rays and NIR light by changing the original upper reflectors to dichroic mirrors. The DOI-PET detector has very low spatial resolution for NIR light compared to the charge-coupled device cameras that are normally used. However, it is possible to reconstruct images of comparable value from the data acquired by low-resolution devices because the light is scattered by biological tissues and high-resolution devices are not necessarily effective at improving image quality. In this study, we demonstrate the feasibility of 3D NIR fluorescence tomography imaging by employing DOI-PET detectors in computer simulations. In the simulations, we used a 40 mm x 40 mm x 40 mm cubic phantom, a square detector geometry, and an optical diffusion equation to approximate the light propagation. We then evaluated imaging systems for 3D fluorescence tomography with different detector resolutions and excitation light arrangements using singular-value analysis and imaging simulation. We confirmed that the reconstructed images from low-resolution detectors (8 x 8 pixels for an area of 40 mm x 40 mm) are the same as those from high-resolution detectors (16 x 16 pixels for the same area).

This study proposes a piecewise Wiener estimation method to reconstruct a spectral reflectance image from a three-band image by multipoint spectral information collected simultaneously with image acquisition. A three-band image is divided into several blocks and the spectral estimation is carried out using the Wiener estimation matrix assigned to each block. Each Wiener estimation matrix is constructed on the basis of spectral measurement data. The experimental results show that the proposed method reduces the average estimation error monotonically as the number of spectral measurements increases. In addition, the computational time of the piecewise Wiener estimation costs only severalfold of the computational time of the conventional single-matrix method.

IT Strategic Headquarters of the Japanese government compiled the Priority Policy Program 2007, in which “Establishment of the structure for every citizen to be able to manage and utilize his health information by himself” and “Foundation of the e‐ Post‐Office box for the realization of the social security service in aspects of people” are declared. For this purpose, a health information system is considered that delivers healthcare data to the server, where the data is to be individually self‐administered by the owner. A patient can register his data, and download or reference it from any medical institution or home when necessary. We made a prototype system to realize such a personal health data referring system based on the e‐ post‐office box concept. The system is to be used in field trial experiment with the staffs and students of Tokyo Institute of Technology using their ID Card. This prototype system is expected to be available for the policy suggestion in the realization of the e‐P.O.Box stated in the Priority Policy Program of the government.

The article proposes a multispectral image compression scheme using nonlinear spectral transform for better colorimetric and spectral reproducibility. In the method, we show the reduction of colorimetric error under a defined viewing illuminant and also that spectral accuracy can be improved simultaneously using a nonlinear spectral transform called Labplus, which takes into account the nonlinearity of human color vision. Moreover, we show that the addition of diagonal matrices to Labplus can further preserve the spectral accuracy and has a generalized effect of improving the colorimetric accuracy under other viewing illuminants than the defined one. Finally, we discuss the usage of the first-order Markov model to form the analysis vectors for the higher order channels in Labplus to reduce the computational complexity. We implement a multispectral image compression system that integrates Labplus with JPEG2000 for high colorimetric and spectral reproducibility. Experimental results for a 16-band multispectral image show the effectiveness of the proposed scheme.

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 high-performance brain PET scanner, jPET-D4, which provides four-layer depth-of-interaction (DOI) information, is being developed to achieve not only high spatial resolution, but also high scanner sensitivity. One technical issue to be dealt with is the data dimensions which increase in proportion to the square of the number of DOI layers. It is, therefore, difficult to apply algebraic or statistical image reconstruction methods directly to DOI-PET, though they improve image quality through accurate system modelling. The process that requires the most computational time and storage space is the calculation of the huge number of system matrix elements. The DOI compression (DOIC) method, which we have previously proposed, reduces data dimensions by a factor of 1/5. In this paper, we propose a transaxial imaging system model optimized for jPET-D4 with the DOIC method. The proposed model assumes that detector response functions (DRFs) are uniform along line-of-responses (LORs). Then each element of the system matrix is calculated as the summed intersection lengths between a pixel and sub-LORs weighted by a value from the DRF look-up-table. 2D numerical simulation results showed that the proposed model cut the calculation time by a factor of several hundred while keeping image quality, compared with the accurate system model. A 3D image reconstruction with the on-the-fly calculation of the system matrix is within the practical limitations by incorporating the proposed model and the DOIC method with one-pass accelerated iterative methods.

Pathological images are color images of stained tissue slides, the color of which varies depending on staining conditions. For reliable diagnosis, the color variation must be corrected in these images. This paper proposes a color correction method for hematoxylin and eosin (H&E) stained pathological images in which the amounts of H&E dyes are estimated based on multispectral imaging technique and Beer Lambert law, and the color image is generated corresponding to the adjusted amount of dyes. This enables us to correct an image to an arbitrary or specified optimal staining-condition image. Through experiments using H&E stained human liver slide images, the effectiveness of the proposed method was confirmed.

We made a tissue phantom with double-peak particle size distribution, which has polystyrene particles of cell nuclear size and mitochondrial size, and measured the spectrum from the tissue phantom using a single optical fiber. In this paper we investigate the characterization method for the tissue phantom with double-peak particle size distribution by comparing the measured spectra with the calculated ones using the Monte Carlo (MC) method. It is first shown that the Mie phase function characterizes better than the Henyey-Greenstein (H-G) phase function in MC calculation. Next, we compare the measurement spectra with those obtained by modeling as single-peak, conventional modeling for particle size distribution, and for double-peak particle size distribution. The single-peak modeling is found to cause considerable error for the tissue phantom with double-peak particle size distribution, which seems to simulate a biological tissue. We suggest that if one simulates the particle size distribution of a biological tissue by conventional modeling, the accuracy of estimation will be lower.

Recently, multispectral images have been utilized for high fidelity color reproduction through visual telecommunication systems and its compression schemes have been required. However multispectral image coding considering color degradation has not been established. This article presents a comparison of four spectral transforms presented before to find out which transform is more suitable to a spectral decorrelation scheme in transform coding of multispectral images. Transforms dealt with in this paper are Karhunen-Loeve transform (KLT), weighted KLT (WKLT), one mode analysis (OMA) and a transform realizing compatibility to the conventional color systems, latter three of which are designed to represent spectral reflectance functions with small number of coefficients having small colorimetric errors. Through the theoretical and experimental comparisons, it is found that WKLT and OMA reduced colorimetric error compared to the others, where OMA is suitable when the rendering illuminations in the color reproduction can be assumed in advance and WKLT is applicable to other situations.

We propose a new method for the color enhancement of multispectral image in the visible wavelength region. The purpose of the proposed method is to explore the weak features contained in a specific wavelength by discounting the major color distribution. Such examination will be valuable in visual inspection applications, for example, a medical examination using color image to find a small spectral change of an abnormal part. In this method, Karhunen-Loeve (KL) transform is applied to multispectral data, and specific wavelength components of only high-order KL coefficients are amplified while low-order coefficients are not changed to retain the major color distribution. In the experiment, this method was applied to multispectral images: a printed test image and a human skin image of a bruised arm were captured by a 16-band multispectral camera. The resultant images were compared with the images obtained by saturation enhancement and that obtained by applying the proposed method to the 3-band image. The method successfully visualized the features, which are almost invisible in natural color images, with less change in background color than saturation enhancement.

Staining of tissue specimens is a classical procedure in pathological diagnosis to enhance the contrast between tissue components such that identification and classification of these components can be easily performed. In this paper, a framework for digital staining of pathological specimens using the information derived from the L-band spectral transmittance of various pathological tissue components is introduced, particularly the transformation of a Hematoxylin and Eosin (HE) stained specimen to its Masson-Trichrome (MT) stained counterpart. The digital staining framework involves the classification of tissue components, which are highlighted when the specimen is actually stained with MT stain, e.g. fibrosis, from the HE-stained image; and the linear mapping between specific sets of HE and MT stained transmittance spectra through pseudo-inverse procedure to produce the LxL transformation matrices that will be used to transform the HE stained transmittance to its equivalent MT stained transmittance configuration. To generate the digitally stained image, the decisions of multiple quadratic classifiers are pooled to form the weighting factors for the transformation matrices. Initial results of our experiments on liver specimens show the viability of multispectral imaging (MSI) for the implementation of digital staining in the pathological context.

To preserve museum collections of works of art, these collections are often photographed for display in digital museums. However, conventional photography cannot capture spectral characteristics of objects. In this paper, we describe a 16-band camera system designed to produce spectral images of ancient paintings. Results of color reproduction of captured images and results of spectral analysis of images of ancient paintings are also presented. The camera consists of a 2000×2000-pixel CCD, a rotational filter turret with 16 interference filters, and a PC-based image capturing and displaying unit. The camera's lens is interchangeable, and it enables two or more different view sizes. Each band image of the camera can be focused independently, and it reduces longitudinal chromatic aberration. A stroboscope is used for lighting, and the rotational filter turret and electrical shutter of the CCD have been synchronized with it. An electric motor-driven photographic platform is used to enable photographing large objects in several shots. We evaluated the results of color estimation for an image taken by this camera using the GretagMacbeth ColorChecker 24-color chart. The average ΔEab was 2.09 (maximum ΔEab was 4.03). Spectral reflectance were used to analyze a degraded area on an ancient painting.

In the conventional color reproduction based on the colorimetric match for a standard observer, color mismatch can be perceived if the color matching functions of the observer deviate from those of the standard observer; this phenomenon is known as observer metamerism. Recently, multi-primary display, using more than three-primary colors, has attracted attention as a color reproduction media because of its expanded gamut and its possibility to reduce the color mismatch caused by observer metamerism. In this paper, a new color conversion method for multi-primary display that reduces the observer metamerism is proposed. The proposed method gives the multi-dimensional control value of a display device to minimize the spectral approximation error under the constraints of tristimulus match. Reproduced spectrum by a seven-primary display is simulated and evaluated by the color matching functions of Stiles's 20 observers. The results confirmed that the proposed method reduces the color reproduction error caused by observer variability compared to the other seven-primary reproduction and conventional three-primary reproduction. The preliminary visual evaluation results with a seven-primary display using light-emitting diodes are also introduced.

Color reproduction systems using multispectral imaging techniques make it possible to accurately reproduce the color of the original object under various viewing illuminants. In this paper, a multispectral image compression method for high fidelity color reproduction is proposed in consideration of color degradation. In the proposed method, a spectral transform and a nonlinear quantization designed to reduce colorimetric error are combined with the discrete wavelet transform in JPEG2000. Through the experiments using some 16-band multispectral images, it is confirmed that the proposed method reduces the average and the maximum color differences in L*a*b* color space in comparison with the conventional methods.

Multi-primary displays, which enable to expand their color gamut, have a degree of freedom in the selection of the device control signals to reproduce a set of tristimulus values. To reproduce a given tristimulus values on a multi-primary display, we need a color conversion algorithm which gives a unique set of device control signals; several algorithms with different characteristics have been proposed. However, they do not guarantee that control signals change smoothly even when tristimulus values change smoothly. Such signal discontinuity is pointed out as a cause of pseudo contour appeared in smooth tonal change when color matching functions of observers are deviated from the standard one or when device profiles include some error. In this article, first, the relation between signal discontinuity and pseudo contour is theoretically analyzed. Then a comparative evaluation is performed on the smoothness of the gradation patterns among three color conversion algorithms. As a result of the comparison, it was confirmed that a method that generates smooth signal all over the color gamut gave the smoothest gradation both objectively and subjectively.

A high-performance positron emission tomography (PET) scanner, which measures depth-of-interaction (DOI) information, is under development at the National Institute of Radiological Sciences in Japan. Image reconstruction methods with accurate modeling of the system response functions have been successfully used to improve PET image quality. It is, however, difficult to apply these methods to the DOI-PET scanner because the dimension of DOI-PET data increases in proportion to the square of the number of DOI layers. In this paper, we propose a compressed imaging system model for DOI-PET image reconstruction, in order to reduce computational cost while keeping image quality. The basic idea of the proposed method is that the DOI-PET imaging system is highly redundant. First, DOI-PET data is transformed into compact data so that data bins with highly correlating sensitivity functions are combined. Then image reconstruction methods based on accurate system modeling, such as the maximum likelihood expectation maximization (ML-EM), are applied. The proposed method was applied to simulated data for the DOI-PET scanner operated in 2-D mode. Then the tradeoff between the background noise and the spatial resolution was investigated. Numerical simulation results show that the proposed method followed by ML-EM reduces computational cost effectively while keeping the advantages of the accurate system modeling and DOI information.

We propose a new illumination method for a medical endoscope: narrow band imaging (NBI), in which the spectral bandwidth of the filtered light is narrowed. To confirm how the spectral specifications of the filtered light influence a reproduced image, an experiment was conducted observing the endoscopic images of the back mucosa of a human tongue. In addition, the effect of NBI on endoscopic images was investigated through preliminary clinical tests in colonoscopy and upper gastrointestinal endoscopy. It has been shown that NBI can enhance the capillary pattern and the crypt pattern on the mucosa. These patterns are useful features for diagnosing an early cancer.

The architecture of color reproduction system based on spectral information is addressed. This system aims to realize the accurate color reproduction under arbitrary illuminants and color matching functions and anticipates the ultimate reproduction of spectral reflectance and spectrum of objects. This system is like an expanded version of that proposed by the International Color Consortium (ICC). One of the features of this system is the usage of the spectral space for the profile connection space alternative to the colorimetric space. It retains compatibility with the indigenous color management functions of the current system. The proposed system is further expanded to construct the spectrum-based color reproduction system for motion picture. Six-band HDTV camera and six-primary projection display system, which is realized in our laboratory as experimental spectrum-based color reproduction system for motion picture, is briefly introduced.

A method for estimating a spectral reflectance image of an object is presented, in which the image of a color chart is used instead of the spectral characteristics of the camera and the illumination. Since the spectral characteristics of the color chart are the essential issue in this method, this paper shows the requirement for the color chart based on the spectral characteristics of the set of target objects. Computer simulations are performed to confirm the effectiveness of the proposed method for typical multi-band imaging systems and illuminations. In addition, an investigation is carried out to roughly estimate the amount of the error when the color chart does not satisfy the requirement completely.

A novel color reproduction method that is independent of the viewing angle is proposed. In this method two sets of tristimulus values corresponding to the CIE 1931 and the CIE 1964 color matching functions are reproduced using a six- primary display. The digital counts of the six-primary display to reproduce a given set of six-stimulus values can be uniquely determined by using a 6 by 6 matrix which is calculated from the two color matching functions and the primary spectra of the six-primary display. In this paper we report the result of the simulation that examines the color reproduction accuracy using this method when a spectrum of the object is given. Comparing the color reproduction accuracy achieved by the 6-primary display using our method with that by a current RGB display it is verified that the proposed method can improve the accuracy.

Algebraic reconstruction methods have been successfully used to improve quality of positron emission tomography (PET) images by accurate modeling of measurement system, but they have computational burden. The authors propose a fast PET image reconstruction method based on an algebraic technique. In this method, reconstruction operators are pre-computed approximately using subsets of sensitivity functions. The subsets contain the sensitivity functions that contribute significantly to each point to be reconstructed. The proposed method was applied to simulated data and experimental data for the ECAT EXACT HR+ (Siemens/CTI) scanner operating in 2D mode. These results show that the proposed method produces images with almost the same quality as the conventional algebraic methods do and has a similar computation time to the filtered backprojection method.

In the analysis of a medical image database aimed at formulating useful knowledge for image diagnosis requires an unsupervised image processing technique without preconceived knowledge. In this paper, we propose a method for unsupervised image segmentation, which is suitable for finding the features contained in an image. A small region around each pixel is considered as a pattern vector, and the set of pattern vectors acquired from the whole image is classified using the hierarchical clustering technique. In hierarchical clustering, the classification of pattern vectors is divided into two clusters at each node according to the statistical criterion based on the entropy in thermodynamics. Results of the test image generated by the Markov random field (MRF) model and real medical images photomicrographs of a colon tumor are shown.