An audio-graphics teleconferencing system has been developed that uses ordinary personal computers interconnected over a basic rate (2B+ D) ISDN line. The system supports high-speed transmission of 200-dpi resolution documents read in by an optical
scanner and presented on the displays of the conference participants. While looking at the same material, the conferees can interactively converse and make handwritten notations on the document via an LCD tablet for all the participants to see. We describe the configuration and performance of the system, focusing mainly on the ISDN based multimedia transmission method and the method for reducing and enlarging binary images.
The Hughes Aircraft Company CRC-365 is a hybrid infrared focal plane array that features a 256 x 256 array of platinum silicide photodiodes. The device was tested in a laboratory environment to determine its performance limits, in the context of medium
wave IR imaging. Test results included the dc transfer function of the multiplexer, power dissipation, dark current, thermal barrier height, Schottky barrier height, blackbody response, saturation charge, signal contrast, Schottky quantum yield, temporal noise, noise-equivalent temperature difference, response nonuniformity, temporal drift, and minimum resolvable temperature difference. Overall, performance was found to be excellent. Using two-point nonuniformity correction, background-noise-limited photodetection was achieved for a 40-deg target temperature range. Notable problem areas included bandwidth limitations and electrical crosstalk.
We present a new adaptive thresholding algorithm that
uses the theories of human visual perception to select a global threshold value. The algorithm is based on the total contour information contained within an image and selects a threshold value that
maximizes the edge features within the binarized image.
Simple optimization procedures are used to obtain suboptimal filters based on the Matheron expansion for binary images corrupted by subtractive noise. The Matheron expansion allows us to represent all morphological filters as either a union of erosions or an intersection of dilations. We perform a search over a subset of the filters represented by the union-of-erosions version of the Matheron expansion. We present results for a set of simple binary images with
regular borders and a more complex set with very irregular borders. Results are compared against results for binary closings.
We survey the framework of morphological edge detection. Morphological gradients are hybrid operators: they are constructed with set and arithmetic operations. After a short introduction
to gradients in digital images, we present the gradients available in mathematical morphology: morphological gradients, half gradients, and directional gradients. These gradients are based on dilations and erosions. We present a new directional gradient based on graytone thinning/thickening and a new multiscale gradient called the regularized gradient. Morphological gradients have a considerable
advantage with respect to classical edge detection paradigms: they are easier to generalize to any type of space in which dilation can be defined. We describe the gradient operators in image sequences,
3-D images, and graphs. We propose a new operator on graphs, the mosaic gradient.
A new approach for multilevel thresholding based on two interesting properties of the histogram is proposed. First, a peakfinding algorithm is presented based on the symmetry of a histogram, for which the hiilsides of each hill are symmetrical about the
central curve. Next, the duality that the peaks and valleys are opposite is presented in order to identify the valleys of the histogram. Based on this property, we propose a vailey-finding algorithm to construct the hierarchical order of the various vaileys. The proposed method is computationaily faster than traditional ones such as the variance-based and entropy-based methods. Compared with recent
work proposed by Lim, the proposed method has some advantages in constructing the hierarchical order of various vaileys.
A novel image encryption technique is proposed. The encrypted image is obtained by randomly changing the phase spectra of the original image. Therefore, the resulting image is unrecognized and the image encryption is achieved. The phase spectra of the
original image are added with the binary phase spectra of a pseudonoise. This type of image encryption is similar to the private-key
cryptographic system. The attack study for the proposed image encryption scheme is also conducted. It is shown that the possibility of a successful attack for a 512 x 512 encrypted image is, at most,
1.25 x 10 which requires 84 years by the state-of-the-art technology. The application of this type of encryption for the progressive transmission is also conducted. It is found that any part of the encrypted image can be used to reconstruct a meaningful original image. In addition, the encrypted image is insensitive to the presence of data loss. Therefore, in avoiding the network congestion it is suitable
for the multimedia communications.
The error diffusion algorithm for binarization of continuous-tone images suffers from artifacts in the low- and high-input regions. These artifacts can be reduced by introducing inputdependent weighting coefficients in the form of multiple error diffusion matrices into the error diffusion algorithm.
A model was derived and tested that predicted the spectral reflectance factor from digital data for a dye diffusion thermal transfer printer. The model was based on the Kubelka-Munk turbid media theory and included terms to account for back diffusion onto
the donor supply and dye transfer inhibition. The average colonmetric characterization accuracy of the printer's color gamut based on 45 samples to estimate the model parameters was 3.0 Eb with a maximum of 7.4. The spectral model had similar performance to look-up table and multidimensional interpolation methods and supenior performance to multiple-linear regression methods based on tristimulus data.