The usefulness of different distance measures in the training phase of self-organizing map (SOM) for color histogram generation for spectral image retrieval purposes is examined. The calculation of the best-matching unit (BMU) in the training phase of SOM is done by using Euclidean distance, Kullback–Leibler distance, Jeffrey divergence, and CIEL*a*b* color difference as distance measures. One-dimensional SOMs are generated for two different data sets consisting of 1269 Munsell color chips and 1, 440, 000 color spectra collected from a real spectral image database. The suitability of the introduced measures is first evaluated by calculating the average color differences between the Munsell data set and its BMUs in the SOMs trained by Munsell data. The achieved results are validated by a practical application, in which the queries from a real spectral image database are performed. Furthermore, the ability of SOMs trained by different distance measures to distinguish between spectral images of real human skin and magazine prints of human skin is examined. The achieved results are promising and indicate that two-dimensional self-organizing maps, which are trained by using Euclidean distance and Jeffrey divergence as distance measure and color histograms that correspond the spectral images as training data, could be used for classifying spectral images.

Multiple-beam scanning optical systems are widely used for high-speed and high-resolution printers. One of the requirements that the printer with several levels of print dot density has is conversion of intervals of scanning lines formed by multiple beams. For this purpose, we developed a converting lens system, which consists of two cylindrical lenses and a spherical lens, and designed a four-beam optical system with a converting function of seven levels in print dot density.

An experimental study on the deposition of micro-size droplets (∼39 μm in diameter) of molten wax ink on an aluminum surface is presented. Effects of initial temperature of droplets, substrate temperature and distance from printhead to substrate on the deposited droplet shape and textures were investigated. Depending on impact conditions, droplets may have either smooth or irregular edges, and the final shape may be either regular or two tiered. Analysis was conducted to compare the time scales for solidification, viscous damping and oscillation. A simple heat transfer model was developed, and temperature dependences of viscosity and surface tension were taken into account. The Ohnesorge number of droplets was investigated as a function of time to compare the transient effects of viscous damping and oscillation of the droplets after impact. The number of oscillations completed before the Ohnesorge number reaches unity agrees with the number of tiers formed. The height of the first tier was related to the value of the Ohnesorge number during the first oscillation. The thermal capillary effect was evaluated by defining and examining two Marangoni numbers for the spreading and post-spreading phases of the droplet impact. Splashing of droplets occurred and produced fingers around the droplet peripheries, which was mainly determined by local solidification and spreading dynamics in the vicinity of contact line.

One of the ways to improve the heating efficiency of a heater is to eliminate the passivation layer of the heater by using TaSiO for the heater material, whereby the heater surface becomes thermally oxidized. Rapid boiling of water occurs at around 300°C; therefore, it is necessary for the fabrication condition of the oxidation layer to correspond to a temperature higher than the boiling temperature. In a pulse drive during printing, it is also necessary to keep the heat at a temperature lower than that of thermal oxidization to prevent deterioration. As a result of observing the generation of bubbles and measuring the heater temperature, it was confirmed that within the range of heating velocity in this experiment, boiling starts at around 300°C, and it was also possible to understand pulse control in thermal ink jet printheads to maintain the heater at an appropriate temperature.

The term "image quality" is often used to describe the performance of an imaging system. Recent research showed however that image quality may not be the most appropriate term to capture the evaluative processes associated with experiencing three-dimensional (3D) images. The added value of depth in 3D images is clearly recognized when viewers judge image quality of unimpaired 3D images against their two-dimensional (2D) counterparts. However, when viewers are asked to rate image quality of impaired 2D and 3D images, the image quality results for both 2D and 3D images are mainly determined by the introduced artifacts, and the addition of depth in the 3D images is hardly accounted for. In this article we describe an experiment where we applied and tested two alternative evaluative concepts: naturalness and viewing experience. It was hypothesized that these concepts would be more sensitive to the added value of depth in 3D images. Four scenes were used, varying in dimension (2D and 3D) and noise level (six levels of white Gaussian noise). Results showed that both viewing experience and naturalness were rated higher in 3D than in 2D when the same noise level was applied. Thus, the added value of depth is clearly demonstrated when the concepts of viewing experience and naturalness are being evaluated in contrast to earlier results found using image quality. The added value of 3D over 2D, expressed in noise level, was 2 dB for viewing experience and 4 dB for naturalness, indicating that naturalness appears the more sensitive evaluative concept for demonstrating the psychological impact of 3D displays.

This article presents a new method to track and measure retinal blood vessel with sub-pixel accuracy. The method is based on a modified Canny edge detection method and a Gaussian model of the vessel profile and zero cross of second order derivative of the Gaussian model. In this method, Canny edge detection method is used at first to detect edges from an input retinal image. The edge orientation is then refined and used to obtain vessel cross-sectional profile. Vessel sample profile is searched using local maximum and local minimum from the smoothed vessel cross-sectional profile. Gaussian model is then used to fit the vessel sample profile. The zeros cross of the second order derivatives of the Gaussian fit vessel sample profile are then used to represent the boundaries of the blood vessels. The peaks of the Gaussian fits are then used as positions of vessel center lines and to calculate the widths of the vessel at those points. The method outputs vessel wall positions, the center lines of the vessels with the widths of the vessels with respect to center points in sub-pixel accuracy.

The need for color constancy and the effect of illuminant changes on the outputs of spectral image queries are examined in this study. The queries are performed in five spectral image databases which are similar in content but different in lighting condition. One of the used databases consists of reflectance images whereas the images of the other databases are under the International Commission on Illumination illuminants A, D₆₅, F8 and F11. Color and texture features are utilized in the performed queries and the similarity of two query outputs is evaluated by calculating the Hamming distance and intersection number between the outputs. The results indicate the importance of the correct lighting condition, the need of color constancy, and the robustness of the local binary pattern-based texture features under illuminant changes.