This PDF file contains the editorial “Editorial: What is the value of the journal Optical Engineering” for OE Vol. 32 Issue 07

This PDF file contains the editorial “Guest Editorial: Visual Communications and Image Processing IV” for OE Vol. 32 Issue 07

In many still-image and video processing applications, the time-frequency localization properties of the decomposition technique are an important consideration. While bandwidth compression of the image requires operators with good localization in frequency, spatial features such as edge preservation demand a high degree of localization in time (or the spatial variable). These requirements compete with each other and one is secured at the expense of the other. The classical "uncertainty" principle in the continuous-time domain provides the back drop for this trade-off. Our purpose is to review recent extensions of this principle to the discrete-time case and to develop optimum wave forms. We review common features of block transforms, subband filter banks, and wavelets, and demonstrate how the discrete uncertainty can be used to evaluate these decomposition methods. In particular, we evaluate the trade-off between localization in time and in frequency for several proposed signal decomposition structures.

A hexagonally sampled image is split into a low passband and nine passbands of one octave width and 60 deg angular orientation. The conditions to be satisfied by the filter banks for perlect reconstruction are presented. The human visual system's response to stimuli at differing spatial frequencies is then employed to shape the coding noise spectrum. Rate is allocated under a frequency-weighted mean-square-error distortion measure. A framework is presented employing either the power spectral density of the image or the variance of the subbands. Both adaptive and nonadaptive entropy coding simulations are carried out under a Laplacian source distribution. Transparent coding results are presented at rates below 1 bit/pixel.

A method for image compression based on subband decomposition is presented. We describe a new filter bank design method for image coding applications and a new entropy coding algorithm for the compression of subband images. A set of relevant optimization criteria is defined for the filter bank design. For the compression, a composite source model is defined by combining vector quantization (VQ) and scalar quantization (SQ) with entropy coding. In the proposed scheme, VQ exploits the remaining statistical dependencies among the subband samples, while SQ allows an optimal control on local distortions. The system is based on a statistical model that uses VQ information to generate low entropy probability tables for an arithmetic coder. The bit rate can be shared between the VQ rate and the SQ rate, allowing many possible configurations in terms of performance and implementation complexity. The proposed system shows improved performance when compared with other existing methods.

Many conventional video coding schemes, such as the CCITT H.261 recommendation, are based on the independent processing of nonoverlapping image blocks. An important disadvantage with this approach is that blocking artifacts may be visible in the decoded frames. We propose a coding scheme based entirely on the processing of overlapping, windowed data blocks, thus eliminating blocking eftects. Motion estimation and, in part, compensation are performed in the frequency domain using a complex lapped transform (CLT), which can be viewed as a complex extension of the lapped orthogonal transform (LOT). The motion compensation algorithm is equivalent to overlapped compensation in the spatial domain, but also allows image interpolation for subpixel displacements and sophisticated loop filters to be conveniently applied in the frequency domain. For inter- and intraframe coding, we define the modified fast lapped transform (MFLT). This is a modified form of the LOT that entirely eliminates blocking artifacts in the reconstructed data. The transform is applied in a hierarchical structure, and performs better than the discrete cosine transform (DCT) for both coding modes. The proposed coder is compared with the H.261 scheme and is found to have significantly improved performance.

A general approach to block-matching motion estimation is introduced. It very well handles the complex motion found in broadcasting television signals by comparing each block of the current frame with a deformed quadrilateral of the previous one. Calculation of the extra motion information requires additional operations that increase the computational load but the improved prediction reduces the bit rate. It is shown that the proposed method outperforms the conventional full search method by improving the overall SNR of the prediction error by at least 1 dB and reducing the bit rate by 10%.

A motion field estimation method for image sequence coding is presented. Motion vector field is estimated to remove the temporal redundancy between two successive images of a sequence. Motion estimation is an ill-posed inverse problem. Usually, the solution has been stabilized by regularization, as proposed by Tikhonov in 1963, i.e., by assuming a priori the smoothness of the solution. Here, discontinuities of the motion field are taken into account by using a Markov random field (MRF) model. Discontinuities, which unavoidably appear at the edges of a moving object, can be modeled by a continuous line process, as introduced by Geman and Reynolds in 1992, via a regularization function that belongs to the Φ function family. This line process leads to solutions less sensitive to noise than an all-or-nothing Boolean line process. Taking discontinuities into account leads to the minimization of a nonconvex functional to get the maximum a posteriori (MAP) optimal solution. We derive a new deterministic relaxation algorithm associated with the Φ function, to minimize the nonconvex criterion. We apply this algorithm in a coarse-to-fine multiresolution scheme, leading to more accurate results. We show results on synthetic and real-life sequences.

A motion model is developed to determine the velocity and orientation of motion in video sequences. It is able to classify the motion into eight velocity classes, and for each velocity class it is further classified into three orientations of horizontal, vertical, and diagonal motions. The motion model was tested with real-life video sequences and the results were verified by informal subjective evaluations.

We present an improved decoding paradigm for vector quantization (VQ) of images. In this new decoding method, the dimension of the code vectors at the decoder is higher than the dimension of the input vectors at the encoder, so that the area covered by each output vector extends beyond the input block of pixels into its neighborhood. The image is reconstructed as an overlapping patchwork of output code vectors, where the pixel values in the lapped region are obtained by summing the corresponding elements of the overlapping code vectors. With a properly designed decoder code book, this lapped block-decoding technique is able to improve the performance of VQ by exploiting the interblock correlation at the decoder. We have developed a recursive algorithm for designing a locally optimal decoder code book from a training set of images, given a fixed VQ encoder. Computer simulation with both full-search VQ and pruned-tree-structured VQ encoders demonstrate that, compared to conventional VQ decoding, this new decoding technique reproduces images with not only higher SNR but also better perceptual quality.

Vector quantization (VQ) is a powerful technique for low-bit-rate image coding. However, initial studies of image coding with VQ have revealed that VQ causes degradations, most notably around edges. Moreover, the computational complexity is high. Although a few algorithms have been developed to reduce edge degradation, such as block truncation coding (BTC) with VQ (BTCNQ) or classified vector quantization, their compression ratios are not satisfactory. Discrete cosine transformation with VQ (DCTNQ) has been applied to image compression, showing a high compression ratio, but the edge degradation problem still exists. We present an image compression algorithm that takes advantage of the merits of DCTNQ and BTCNQ to achieve a high-quality and low-bit-rate compression of images. High quality images can be achieved at rates of 0.34 to 0.46 bit/pixel.

A new predictive vector quantization (PVQ) technique capable of exploring the nonlinear dependencies in addition to the linear dependencies that exist between adjacent blocks (vectors) of pixels is introduced. The two components of the PVQ scheme, the vector predictor and the vector quantizer, are implemented by two different classes of neural networks. A multilayer perceptron is used for the predictive cornponent and Kohonen self-organizing feature maps are used to design the codebook for the vector quantizer. The multilayer perceptron uses the nonlinearity condition associated with its processing units to perform a nonlinear vector prediction. The second component of the PVQ scheme vector quantizes the residual vector that is formed by subtracting the output of the perceptron from the original input vector. The joint-optimization task of designing the two components of the PVQ scheme is also achieved. Simulation results are presented for still images with high visual quality.

A new algorithm for designing differential pulse code modulation (DPCM) systems is presented for image data compression. When transmitting images over noiseless channels, the distortion between the original and reconstructed images results primarily from quantization noise. This is true when optimal predictor structures are employed. The quantization error becomes severe at low bit rates. This is because of the large quantization error being directly fed back into the predictor and used in subsequent estimation of future pixels. The DPCM scheme developed attempts to balance between nonoptimal predictor designs and significantly reduced feedback effects resulting from quantization errors with the objective of maximizing reconstructed image quality. DPCM system performance using the algorithm is about 2.5 dB greater than that obtained from an optimally designed conventional system. In addition, the algorithm is robust. Thus, the DPCM predictor does not need to be redesigned using exact statistics of the input image data for each image to be transmitted.

A progressive image transmission (PIT) scheme based on the classified transform vector quantization (CVQ) technique using the lapped orthogonal transform (LOT) and human visual system (HVS) weighting is proposed. Conventional block transform coding of images using the discrete cosine transform (DCT) produces, in general, undesirable blocking artifacts at low bit rates. Here image blocks are transformed using the LOT and classified into four classes based on their structural properties and further subdivided adaptively into subvectors depending on the LOT coefficient statistics with HVS weighting to improve the reconstructed image quality by adaptive bit allocation. The subvectors are vector quantized and transmitted progressively. Coding tests using computer simulations show that the LOT/CVQ-based PIT of images is an effective coding scheme. The results are also compared with those obtained using PIT/DCTVQ. The LOT/CVQ-based PIT reduces the blocking artifact significantly.

A neural network approach to finding trajectories of feature points in a monocular image sequence is proposed. In conventional methods, this problem is formulated as an optimization problem and solved using heuristic algorithms. The problem usually involves lengthy computations, making it computationally difficult. We apply the Hopfield neural network to image sequence correspondence. The design and development of the Lyapunov function for this problem are discussed in detail. Furthermore, the neural-network-based image correspondence scheme is extended to the case of successive image frames, in which some feature points are allowed to be occluded. Examples and simulation results are presented to illustrate the design process and the convergence characteristics of the proposed neural network. By using the massive parallel-processing power of neural networks, a real-time and accurate solution can be obtained.

We develop a bidirectional associative memory (BAM)-based neural network to achieve high-speed partial shape recognition. To recognize objects that are partially occluded, we represent each object by a set of landmarks. The landmarks of an object are points of interest relative to the object that have important shape attributes. To achieve recognition, feature values (landmark values) of each model object are trained and stored in the network. Each memory cell is trained to store landmark values of a model object for all possible positions. Given a scene that may consist of several objects, landmarks in the scene are first extracted, and their corresponding landmark values are computed. Scene landmark values are entered to each trained memory cell. The memory cell is shown to be able to recall the position of the model object in the scene. A heuristic measure is then computed to validate the recognition.

A new scheme is proposed for the efficient coding of highdefinition television (HDTV) sequences for digital video tape recording. This scheme allows the variable speed playback of recorded material at speeds up to 100 times normal in both the forward and reverse directions. An 8:1 level of compression of the original HDTV data is obtained by exploiting both intraframe and interframe redundancies during the coding process, although the perceptual quality of the reconstructed sequences is uncompromised. To allow the high-quality playback of the recorded data at very high speeds, a hierarchical data structure is employed and the head-positioning capabilities of conventional digital video tape recorders (DVTRs) are exploited to enable the recovery of perceptually vital information at high tape speeds. The performance of the resuIting scheme is examined by simulating the compression and reconstruction of three HDTV sequences at 1, 10, 20, 50, and 100 times normal playback speed.

We present a generic video codec, which is able to compress image sequences efticiently regardless of their input formats. In addition, this codec supports a wide range of bit rates, without significant changes in its main architecture. A multiresolution representation of the data is generated by a Gabor-like wavelet transform. The motion estimation is performed by a locally adaptive multigrid block-matching technique. This codec can handle both interlaced and progressive input sequences, the temporal redundancies being exploited by interframe/interfield coding. A perceptual quantization of the resulting coefficients is then performed, followed by adaptive entropy coding. Simulations using different test sequences demonstrate a reconstructed signal of good quality for a wide range of bit rates. Thereby demonstrating that this codec can perform generic coding with reduced complexity and high efficiency.

A preliminary version of a feature point extractor is proposed. In existing model-based coding systems, the feature points of the human face are extracted manually. To make model-based image coding more practical, a coder that can automatically extract the feature points must be developed. We define an ideal feature point extractor and discuss both the requirements and problems. A preliminary solution has been implemented. The proposed feature point extractor can automatically extract the feature points of eyebrows, eyes, the outline of a face, and some other useful information. It can be combined with model-based image coding and used in the video phone/conference systems.

A segmentation-based image coding technique is described. Both uniform and textured region extraction algorithms are used for segmentation. Textured regions are reconstructed using 2-D noncausal Gaussian-Markov random field models. Uniform regions are reconstructed using polynomial expansions. An arithmetic coder is used for coding the boundaries of regions. Reasonable quality images are obtained at compression factors of 85:1.

The dominant image transformation used in the existing fractal coding schemes is the affine function. Although an affine transformation is easy to compute and understand, its linear approximation ability limits the employment of larger range blocks, that is, it limits further improvement in compression efficiency. We generalize the image transformation from the usual affine form to the more general form, e.g., quadratic form, and provide theoretical requirements for the generalized transformation to be contractive. Based on the self-transformation system (STS) model, an image sequence coding scheme-fractal-based image sequence coding-is proposed. In this coding scheme, our generalized transformation is used to model the self-transformation from the domain block to its range blocks. Experimental results on a real image sequence show that for the same size of blocks, the SNR can be improved by 10 dB, or, for the same SNR of the decoded image sequence, the compression ratio is raised twofold when the new generalized transformation is used to replace the usual affine transformation. In addition, due to the utilization of the STS model, the computational complexity is only linearly related to the size of the 3-D blocks. This provides for fast encoding and decoding.

In contrast to the numerous edge-detection techniques that detect edges either point by point or using overlapping circular windows, an edge detector using nonoverlapping rectangular windows is proposed. The detector examines the pixels within each rectangular window of an image, and decides whether an edge element is present or not in the window. Based on the gray and mass moment-preserving principles, the step edge is estimated locally to subpixel accuracy using analytical formulas. To apply the edge detection results to image compression, the detected edge elements are then tracked and grouped based on proximity and orientation. Using the line parameters of the grouped edge elements, region boundaries are approximated in a piecewise linear manner. This reduces the amount of data required to describe region shapes and is useful for compressing some types of images. Good experimental results of compressing character and trademark images are also included to show the feasibility of the proposed approach.

Automatic tissue characterization systems are in great demand by pathologists. However, the existing methods are either too simple to classify a complicated liver tissue image or are dependent on heavy human intervention and very time consuming. We have developed a highly parallel and effective system based on color image segmentation to analyze liver tissue images. To simplify the tissue classification problem, the system first utilizes the achromatic information (the intensity) to segment the tissue image coarsely, then makes use of the chromatic information to classify the segmented regions into four different tissue classes. Thus, the proposed method includes an unsupervised probabilistic relaxation segmentation process and a supervised Bayes classification process. Because the invariant gray level and color properties of the liver tissue image are fully utilized, the difficult classification problem can be fulfilled well at a reasonable computational cost. The proposed method also shows reliable liver tissue classification results from different test sample sets.

We describe a complex log mapping approach for 2-D color object recognition. First, a 3-D vector color image is transformed into a 2-D vector image by projection onto a color plane, then a complex log mapping transform is applied to this 2-D vector color image for invariant pattern recognition. Using extensive computer simulations, this method can effectively recognize a scale- and rotation-changed color object among objects of similar shape but with different color contents and also among objects of different shape and color.

A VLSI implementation of the optical Chinese character recognition (OCCR) system with pipelined and parallel structure is presented. We also propose an efficient method for performing block classification and character segmentation as well as an effective and adaptive feature extraction algorithm for recognizing multifont printed Chinese characters. With the complex and huge amount of data involved in Chinese characters, their recognition requires numerous complex computations. Therefore, to improve the recognition efficiency for practical applications, a VLSI chip is designed and fabricated. To preserve a certain degree of flexibility so that various recognition algorithms can be implemented with the system, only the most time-consumig parts are implemented into the VLSI circuit. By combining the VLSI technology and the effective Chinese character recognition algorithm, a practical OCCR system with high speed, high-recognition rate, and accumulated learning capability is developed. Based on the experimental results, the VLSI chip can process up to 200 characters/s, which is one hundred times faster than the original software algorithm. The recognition rates of three different test conditions are also given.

The generalized Huygens/Fresnel integrals in the time domain are derived and then formulated in the canonical operator form with the aid of the Green function and matrix/operator methods. We show in typical applications that the derivation of the ABCD law in the time domain and temporal imaging with a time lens can be simply treated by using the canonical operator representation of the diffraction integrals.

Axial strain on an optical fiber can be determined by monitoring the phase shift of a variety of optical fiber sensors. The exact analytical solutions for optical fields that propagate in a circular core optical fiber are used here to determine the phase shifts that occur for TE and TM modes due to axial strain. Whenever an optical fiber is stressed, the optical path length, the index of refraction, and the propagation constant of each fiber mode change. In consequence, the modal phase term βz of the fields is shifted by an amount ΔΦ. In certain cases, it is desirable to control the phase sensitivity to make fibers that are either more or less sensitive to strain. We show that it can be accomplished by choosing appropriate fiber parameters.

The effects of magnetic fields on the microchannel plate (MCP) image intensifiers to be used in a novel biplanar fluoroscope are studied along with the system's overall contrast as a function of beam energy. For a second-generation device with wrap-around power supply, B_-3dB values for the gain roll-off were found to be approximately 0.08 T (axial field) and 0.06 T (transverse field). The maximum image shift resuiting from a 0.0035-T transverse field is found to be 0.065 mm, limited by centroid location error resulting from low-dose x-ray noise. The results of x-ray contrast studies suggests that the presently estimated 0.1-rad dose delivered to the patient (2-h magnetic stereotaxis procedure; 12% "fluoro-on" duty cycle) might be reduced by increasing the x-ray energy.

The folded perfect shuffle was proposed to implement the 1-D perfect shuffle by folding the 1-D input data array into a 2-D array, which has the inherent advantage of wholly using the space-bandwidth product of the 2-D optical system. We suggest several methods to implement the folded perfect shuffle using quadrant-encoded gratings and proper spatial filters. The general principles of this interconnection network are described and the experimental results are given.

The contrast and resolution of contemporary ferroelectric liquid crystal (FLC) optically addressed spatial light modulators (OASLM) were measured in time-average and instantaneous modes while the devices were cycled at multikilohertz rates. The inherent, full cycle, write/read/erase response time was 270 μs. The time-average contrast peaked at 54:1 at 3.5 kHz when a 9 lp/mm resolution was used for evaluation. Instantaneous resolutions of 57 and 32 p/mm were obtained when the FLC-OASLM was operated at repetition rates of 900 and 2700 Hz, respectively.

We propose using the phase-shift technique to improve such shortcomings inherent to conventional shadow moire topography as low sensitivity and difficulties in computational processing. Because in shadow moire the reference grating and the deformed grating are mutually dependent, it is not possible to produce phase shifts of the fringe by moving the grating. However, fringe shifting that is actually constant regardless of fringe orders is achieved by changing both the gap between a specimen and the grating and the angle of illumination. This proposal is justified by a simulation test and some experimental results are shown for the verification of this trial.

Previous sensor/observer performance prediction models have not explicitly treated false alarm effects, especially in relation to clutter. A method for predicting probabilities of detection as a function of observer false alarm probability and clutter is presented. The method is applied to previously collected observer data to determine the effect of observer false alarm probability on resolution criteria associated with specified clutter levels. The interdependent effects of clutter and observer false alarm probability are illustrated by deriving an expression for the probability of target acquisition in the case where the observer has only enough time to select one target candidate. It is concluded that false alarm probability has a significant impact on target resolution criteria in moderate and high background clutter. Example resolution criteria are given for several false alarm probabilities.

Interferometric reconstruction of three-dimensional flow fields, that is, interferometric tomography, can be a very useful diagnostic tool. It is noninvasive and can capture gross fields; however, it frequently confronts a challenging problem of reconstructing fields from insufficient data. In most cases, flow-field interferometric data are sparse, nonuniform, noisy, and incomplete in projection and scanning because of opaque objects present either inside or outside the field. Recently, a new method has been developed in an effort to improve reconstruction under these ill-posed conditions. The method is appropriate for reconstructing flow fields with the distinct data characteristics, being based on natural pixel decomposition of the field. It employs rectangular grid elements of different sizes and aspect ratios. It thus reflects intrinsic spatial resolution information contained in the data and allows better resolution and accuracy in the region with more probing rays. Computer simulation of experiments has demonstrated superiority of the method to the conventional one. In simulation, the temperature field of a gravity-driven flow of two interacting heat sources, produced by a numerical code, are tested. Both the maximum and average reconstruction errors are reduced appreciably. Especially, the reconstruction demonstrates substantial improvement in the region with dense scanning by probing rays.

The PDF contains the Communications on "Pascal program to perform Mie calculations."

This PDF file contains the editorial “Book Rvw: Photelectronic Properties of Semiconductors," by Richard H. Bube for OE Vol. 32 Issue 07

This PDF file contains the editorial “Book Rvw: Fiber-Optic Communication Systems," by Govind P. Agrawal for OE Vol. 32 Issue 07