Defining criteria to determine precisely the boundaries of masses in mammograms is a difficult task. The problem is compounded by the fact that most malignant tumors possess fuzzy boundaries with a slow and extended transition from a dense core region to the surrounding less-dense tissues. We propose two segmentation methods that incorporate fuzzy concepts. The first method determines the boundary of a mass or tumor by region growing after a preprocessing step based on fuzzy sets to enhance the region of interest (ROI). Contours provided by the method have demonstrated a good match with the contours drawn by a radiologist, as indicated by good agreement between the two sets of contours for 47 mammograms. The second segmentation method is a fuzzy region-growing method that takes into account the uncertainty present around the boundaries of tumors. The difficult step of deciding on a crisp boundary is obviated in the proposed method. Mea-sures of inhomogeneity computed from the pixels present in a suitably defined fuzzy ribbon have indicated potential use in classifying the masses and tumors as benign or malignant, with a sensitivity of 0.8 and a specificity of 0.9.

The detection of masses and tumors in a mammogram is a difficult problem that could benefit from the use of multiple approaches. We propose an abstract concept of information fusion based on a finite automaton and fuzzy sets to integrate and evaluate results of multiple image segmentation procedures. We give examples on how the method can be applied to the problem of mammographic image segmentation, combining results of region growing and closed-contour detection techniques. We also propose a measure of fuzzyness to assess the agreement between a segmented region and a reference contour. Application of the fusion technique to breast tumor detection in mammograms indicates that the fusion results agree with the reference contours provided by a radiologist to a higher extent than the results of the individual methods.

We propose a direct image segmentation algorithm in the JPEG compressed domain. The algorithm features extracting statistical parameters from direct cosine transform (DCT) coefficients without its inverse transform, and growing regions in line with JPEG compression seamlessly in blocks of 8×8 pixels. In comparison with the latest research efforts in region-based image segmentation, our proposed algorithm achieves significant advantages, including (1) no iteration is involved, (2) no full decompression is required, and (3) segmentation performance is competitive.

The application of pattern recognition techniques to the data from the Faint Images of the Radio Sky at Twenty Centimeters Survey (FIRST) is explored. The methodology developed for the automatic classification of a particular type of radio-source morphology, the so-called "bent doubles" is described. While the procedures and techniques for doing supervised pattern recognition using accurately classified training sets, high-resolution images, and determinate shapes are well established, there exists little guidance for dealing with more ambiguous populations. The implications of low resolution for pattern recognition are discussed and methods for dealing with more ambiguous populations are presented. Bent doubles with three components are considered. With approximately 6% of the catalog sources being a component of a triple, visual classification suggests that approximately 6% of triples are bent. In addition to the problems associated with low resolution, the classification task is complicated by the lack of scale and orientation information, the chance superposition of sources, the very ambiguous nature of the visual classifications, and the considerable variation in bent morphology. For a bent or ambiguous or nonbent visual classification, classifiers are presented that produce results having up to an 80% recognition rate when looking at the top-ranked 16% of sources. The associated nonambiguous false positive rate varies between 14 and 18%. Given that recognition rates cannot be expected to be better than the accuracy of the training set, this is believed to be within a few percent of the maximum recognition rates attainable without an observationally verified training set. Decision tree and neural network results are compared and show, within associated errors, substantially equivalent results. Ranking of sources is demonstrated and the highest-ranked sources are in good agreement with visual classifications. In addition to providing determinate expressions for characterizing the target class, the resulting distributions suggest that the automated classifications constrain the target class in a better manner than the visual classifications.

Text detection in video images is a challenging research problem because of the poor spatial resolution and complex background, which may contain a variety of colors. An automated system for text detection in video images is presented. It makes use of four modules to implement a series of processes to extract text regions from video images. The first module, called the multistage pulse code modulation (MPCM) module, is used to locate potential text regions in color video images. It converts a video image to a coded image, with each pixel encoded by a priority code ranging from 7 down to 0 in accordance with its priority, and further produces a binary thresholded image, which segments potential text regions from the background. The second module, called the text region detection module, applies a sequence of spatial filters to remove noisy regions and eliminate regions that are unlikely to contain text. The third module, called the text box finding module, merges text regions and produces boxes that are likely to contain text. Finally, the fourth module, called the optical character recognition (OCR) module, eliminates the text boxes that produce no OCR output. An extensive set of experiments is conducted and demonstrates that the proposed system is effective in detecting text in a wide variety of video images.

We propose a skin-color-based face location algorithm that can locate the human face in a complex background including nonface skin color regions, such as the hands. The algorithm is divided into three steps. The first step involves the use of color components to segment the skin colors from the complex background. The second step detects a curve called the division curve that delineates the boundary between the hair region and the face region. In the last step, the division curve is used to estimate the face region. The proposed algorithm has wide application to problems where the detection of both hands and face is requisite and to problems in which the distribution of skin/hair color can be known a priori.

An experimental analysis of chromatic equalization based on a new implementation of the Retinex algorithm is presented. The experiments are carried out on a colored Mondrian patchwork illuminated with different commercial light sources and on synthetic images generated with a photometric ray tracer using different illuminants. Regarding the Mondrian patchwork, the spectral characteristics of the bulbs and the reflected light from each patch are measured using a commercial spectrometer. From the measured data, synthetic images of the patchwork with different illuminants are created and processed by the Retinex algorithm. The chromatic correction capabilities of the Retinex implementation have been measured and compared with unfiltered values and with the results of another Retinex implementation and classic color equalization algorithms. Results show that Retinex performs an unsupervised color correction without requiring any information about the spectral composition of the illuminant.

We propose a heuristic approach to color quantize images with contextual information taken into consideration. The idea is to locate the regions of an image having the greatest need for colors, and allocate more quantization levels to them. We achieve this by scanning the elements of the input image in a way determined by their local intensity and selecting the color representatives that comprise the color map according to their local popularity. The overall performance of the color quantization algorithm is evaluated on a representative set of artificial and real images. The experimental results indicate good performance of our proposed algorithm with the capability to focus on the regions of an image having important color information.

A key problem in multimedia systems is the faithful reproduction of color. One of the main reasons why this is a complicated issue is the different color reproduction technologies used by the various devices; displays use easily modeled additive color mixing, while printers use a subtractive process, the characterization of which is much more complex than that of self-luminous displays. To resolve these problems several processing steps are necessary, one of which is accurate device characterization. We examine different learning algorithms for one particular neural network technique that already has been found to be useful in related contexts—namely, radial basis function network models—and propose a modified learning algorithm that improves the colorimetric characterization process of printers. In particular our results show that it is possible to obtain good performance using a learning algorithm that is trained on only small sets of color samples, and use it to generate a larger look-up table (LUT) through the use of multiple polynomial regression or an interpolation algorithm. We deem our findings to be a good starting point for further studies on learning algorithms used in conjunction with this problem.

We present a collection of methods and algorithms able to deal with high dynamic ranges of real pictures acquired by digital engines e.g., charge-coupled device (CCD/CMOS) cameras. An accurate image acquisition can be challenging under difficult light conditions. A few techniques that overcome dynamic range limitations problems are reported. The presented methods allow the recovery of the original radiance values of the final 8-bit-depth image starting from differently exposed pictures. This allows the capture of both low- and high-light details by merging the various pictures into a single map, thus providing a more faithful description of what the real world scene was. However, in order to be viewed on a common computer monitor, the map needs to be compressed and requantized while preserving the visibility of details. The main problem comes from the fact that the contrast of the radiance values is usually far greater than that of the display device. Various related techniques are reviewed and discussed.

A normal estimation algorithm for visualization is presented that approximates the density function in a local neighborhood with a second-degree polynomial function. The coefficients of the polynomial function can be solved by minimizing the error of the approximation. This method is tested in several volume data sets and comparisons with other methods are presented. It is demonstrated that this method is a fairly robust technique for noise-contained data and is preferable for most applications.

We present two linear algorithms for 3-D reconstruction: one is for finite motion; the other is for optical flow. We compare their performance by simulation and real-image experiments, using the same data. The two algorithms are both theoretically optimal, extracting maximum information from the input. We observe that the 3-D reconstruction by the finite motion algorithm is generally more accurate than by the optical flow algorithm if point correspondences are identified with optical flow.

The aim of the vector Doppler (VD) technique is the quantitative reconstruction of a velocity field independent of the ultrasonic probe axis to flow angle. In particular, vector Doppler is useful for studying vascular pathologies related to complex blood flow conditions. Clinical applications require a real-time operating mode and the capability to perform Doppler measurements over a defined volume. The combination of these two characteristics produces a real-time vector velocity map. In previous works we investigated the theory of pulsed-wave (PW) vector Doppler and developed an experimental system capable of producing off-line 3-D vector velocity maps. Afterward, for producing dynamic velocity vector maps, we realized a new 2-D vector Doppler system based on a modified commercial echograph. The measurement and presentation of a vector velocity field requires correct spatial sampling that must satisfy the Shannon criterion. In this work we tackle this problem, establishing a relationship between sampling steps and scanning system characteristics. Another problem posed by the vector Doppler technique is the data representation in realtime that should be easy to interpret for the physician. With this in mind we attempt a multimedia solution that uses both interpolated images and sound to represent the information of the measured vector velocity map. These presentation techniques are tested for real-time scanning on flow phantoms and preliminary measurements in vivo on a human carotid artery.

Puri and Aravind's method of macroblock bit count estimation for video rate control is based on the classification of the macroblock data into discrete classes and assigning a unique nonlinear estimate for each class and quantization parameter pair. This method stands apart from other methods in the literature, since the model of the bit count versus the quantization parameter relation, parameterized by macroblock variance, is a discrete model generated solely from measurements. We extend their technique for low-delay video rate control (tight buffer regulation) in two ways. We propose a strategy of near-uniform quantization parameter assignments to the macroblocks of a frame that can come close to maximizing an objective spatial quality function, such as PSNR, over the entire frame. We also adaptively update the quantization parameter assignments for the yet to be coded macroblocks, after the encoding of each macroblock, to compensate for any errors in the bit count estimation of the encoded macroblock. Our experiments demonstrate that the proposed rate control method can more accurately control the number of bits expended for a frame, as well as yield a higher objective spatial quality than the method adopted by TMN8.

The main objective of this work is to determine a numerical measure that can judge block-based compressed images, namely vector quantization (VQ), JPEG, and JPEG2000 encoded images. Various objective measures are investigated and new measures are developed, then all measures are classified into groups. The analysis of normalized versions of the measures within each group has led to the design of a new measure, denoted as the normalized perceptual image quality measure (NPIQM), that better matches the psychophysical evaluation of the image quality than using the peak signal-to-noise ratio (PSNR) criterion. NPIQM can also be plotted graphically in a radar-fashion form, which gives the ability to compare different images at different bit rates, and give a detailed judgment based on the proposed graphical interpretation. A perceptual examination is carried out and proves the superiority of the NPIQM measure over PSNR.

∞A new wavelet-based L∞-constrained fixed and embedded coding technique is proposed. The embedded bit stream can be truncated for any desired distortion bound at a corresponding bit rate, so that the target upper bound on the elements of the reconstruction error signal is guaranteed. The original image can also be coded up to a fixed a priori user-defined distortion bound, ranging up to lossless coding. A lifting-based wavelet decorrelating transform is employed on the original image, and exact relations are established between spatial and wavelet domain distortions. The wavelet coefficients are quantized by symmetric uniform quantizers for fixed-distortion coding and by families of embedded uniform deadzone scalar quantizers for embedded coding. The quantized coefficients are finally losslessly encoded using a quadtree-based coding algorithm. Any floating-point lifting-based wavelet transform can be used, and a few of the popular wavelet transforms included in the JPEG2000 verification model are worked out as examples. We compare other L∞-constrained coding schemes and show that our proposed coder offers in addition a fully embedded L∞-oriented bit stream. We illustrate also that the proposed coder retains the same capabilities as the state-of-the-art embedded wavelet-based codecs, while providing superior compression results and embeddedness with respect to the L distortion measure.

In general, there is a tradeoff between an orthogonal transform performance and its computational complexity. The Haar transform is by far the fastest of all unitary transforms. The objective of this work is to develop a class of unitary transforms that exhibits computational efficiency as well as good performance utilizing the Haar and slant transforms. This class is called the parametric slant-Haar transforms that includes as special cases the commonly used Haar and slant-Haar transforms. Examples of parametric slant-Haar transforms are given. An efficient fast parametric slant-Haar transform is introduced and its computational complexity is discussed. Relative performance of parametric slant-Haar transforms to that of the Karhunen-Loeve transform (KLT) is discussed. A partially signal dependent parametric slant-Haar transform is also introduced.

"Holo-Chidi" is a holographic video processing system designed at the Massachusetts Institute of Technology (MIT) Media Laboratory for real-time computation of computer-generated holograms and the subsequent display of the holograms at video frame rates. The Holo-Chidi system is composed of two sets of cards—the set of processor cards and the set of video concentrator cards (VCCs). The processor cards are used for hologram computation, data archival/retrieval from a host system, and for higher level control of the VCCs. The VCC formats computed holographic data from multiple hologram-computing processor cards, converting the digital data to analog form to feed the acousto-optic modulators of the Media Lab's "MarkII" holographic display system. Both the processor card and VCC are fabricated and tested, integrating both sets of cards with a host system is currently in progress. The design and implementation of the Holo-Chidi system is discussed.

The generation of computer-generated holographic fringes for real-time holographic video (holovideo) display is very computation-intensive, requiring the development of such special systems as the Massachusetts Institute of Technology (MIT) Media Lab's "Holo-Chidi" system, which can generate and display holovideo at video rates. The Holo-Chidi system is made of two sets of cards—the set of processor cards and the set of video concentrator cards (VCCs). The processor cards are used for hologram computation, data archival/retrieval from a host system, and higher level control of the VCCs. The VCC formats compute holographic data from multiple hologram computing processor cards, converting the digital data to analog form to feed the acousto-optic modulators of the Media Lab's "MarkII" holographic display system. The generation of the holographic fringes from the 3-D numerical description of a scene takes place inside field-programmable gate arrays (FPGAs) resident in the processor card. These large FPGAs employ several superposition processing pipelines, all working in parallel to generate the fringes of the hologram frame. With nine processor boards, there are the equivalent of about 288 superposition "processors" generating the fringes simultaneously. A Holo-Chidi system with three VCCs has enough frame buffering capacity to hold up to 32 36-Mbyte hologram frames at a time. Precomputed holograms can also be loaded into the VCC from a host computer through the low-speed universal serial bus (USB) port.

We describe a compact automatic device designed to characterize scientific cameras based on photon-limited CCD sensors, by measuring the gain, the linearity, and the quantum efficiency as a function of the wavelength in the visible region and in the near IR. The device is completely automated and requires only a few simple survey controls because it is designed to be operated by unqualified users.

This PDF file contains the editorial “Book Rvw, The Image Processing Handbook, 4th Edition," by John Russ for JEI Vol. 12 Issue 03