This PDF file contains the editorial “Changing of the Guard” for JEI Vol. 19 Issue 04

We propose an efficient technique for temporally aligning video sequences of similar activities. The proposed technique is able to synchronize view-variance videos from different scenes performing similar 3-D activities. Unlike existing techniques that just consider unidirectional alignment, the proposed technique considers symmetric temporal alignment and computes the optimal alignment by eliminating any view-based bias. The advantages of our technique are validated by experiments conducted on synthetic and real video data. The experimental results show that the proposed technique out-performs existing techniques in several test video sequences.

The goal of image segmentation is to partition an input image into a set of regions. In mathematical morphology, the reconstruction of images from markers has proven to be useful in morphological filtering and image segmentation. The utilization of a criterion in the problem of the image reconstruction from an image marker has been partially treated elsewhere. We further investigate this idea and extend it to the problem of image reconstruction from labeled markers by proposing a new method based on the "watershed" transformation as an alternative in image segmentation. The image gradient is considered as a topographic relief that is flooded (similarly as in a normal watershed). However, a criterion is added in this reconstruction process that enables the flexibility to separate structures of interest. Following the flooding analogy on topographic reliefs, this flooding process is limited to certain zones to control the recovering process of structures shapes. Experimental results are provided. A comparison with a viscous watershed is performed to show the differences between them. The technique is applied mainly in the biomedical domain, although the technique can generally be applied to other areas.

Conventional learning-based superresolution algorithms tend to boost noise components existing in input images because the algorithms are usually learned in a noise-free environment. Even though a specific noise reduction algorithm is applied to noisy images prior to superresolution, visual quality degradation is inevitable due to the mismatch between noise-free images and denoised images. Accordingly, we present a noise-robust superresolution algorithm that overcomes this problem. In the learning phase, a dictionary classified according to noise level is constructed, and then a high-resolution image is synthesized using the dictionary in the inference phase. Experimental results show that the proposed algorithm outperforms existing algorithms for various noisy images.

It is difficult to precisely detect all impulsive noise in color images due to the nonstationarity caused by edges and fine details. For many pixels, we can not absolutely classify them as noisy or noise-free, but can only describe them using the likelihood that they are corrupted by impulsive noise. Based on this consideration, we present a new filtering solution to removing impulsive noise from color images. The proposed method first utilizes the unit transforms of quaternions to represent the chromaticity difference of two color pixels, and then divides the image into noise-free and possible noisy pixels. Finally it performs adaptive weighted vector median filtering operations on only the possible noisy pixels to suppress noise. The new weighting mechanism is based on a joint spatial/quaternion-chromaticity criterion, which ensures that pixels with different contamination likelihoods have different contributions to the filter's output. The extensive simulation results indicate that the proposed method significantly outperforms some other well-known multichannel filtering techniques.

A number of image quality metrics are based on psychophysical models of the human visual system. We propose a new framework for image quality assessment, gathering three indexes describing the image quality in terms of visual performance, visual appearance, and visual attention. These indexes are built on three vision models grounded on psychophysical data: we use models from Mantiuk et al. (visual performance), Moroney et al. (visual appearance), and Itti, Koch, and Niebur (visual attention). For accuracy reasons, the sensor and display system characteristics are taken into account in the evaluation process, so that these indexes characterize the image acquisition, processing, and display pipeline. We give evidence that the three image quality indexes, all derived from psychophysical data, are very weakly correlated. This emphasizes the need for a multicomponent description of image quality.

Texture analysis plays an important role in content-based image retrieval and other areas of image processing. It is often desirable for the texture classifier to be rotation and scale invariant. Furthermore, to enable real-time usage, it is desirable to perform the classification efficiently. Toward these goals, we propose several enhancements to the multiresolution Gabor analysis. The first is a new set of kernels called Slit, which can replace Gabor wavelets in applications where high computational speed is desired. Compared to Gabor, feature extraction using Slit requires only 11 to 17% of the numeric operations. The second is to make the features more rotation invariant. We propose a circular sum of the feature elements from the same scale of the feature vector. This has the effect of averaging the feature vector from all orientations. The third is a slide-matching scheme for the final stage of the classifier, which can be applied to different types of distance measures. Distances are calculated at slightly different scales, and the smallest value is used as the actual distance measures. Experimental results using different image databases and distance measures show distinct improvements over existing schemes.

The extraction of texture features from images faces two new challenges: large-scale databases with diversified textures, and varying imaging conditions. We propose a novel method termed multiscale blob features (MBF) to overcome these two difficulties. MBF analyzes textures in both resolution scale and gray level. Proposed statistical descriptors effectively extract structural information from the decomposed binary images. Experimental results show that MBF outperforms other methods on combined large-scale databases (VisTex+Brodatz+CUReT+OuTex). Moreover, experimental results on the University of Illinois at Urbana-Champaign database and the entire Brodatz's atlas show that MBF is invariant to gray-level scaling and image rotation, and is robust across a substantial range of spatial scaling.

A framework for clustered-dot color halftone watermarking is considered, wherein watermark patterns are embedded in individual colorant halftones prior to printing and embedded watermarks are detected from scans of the printed images after obtaining estimates of the individual halftone separations. The principal challenge in this methodology arises in the watermark detection phase. Typical three-channel RGB scanner systems do not directly provide good estimates of the four CMYK colorant halftones that are commonly used in color printing systems. To address this challenge, we propose an estimation method that, when used with suitably selected halftone periodicities, jointly exploits the differences in the spatial periodicities and the color (spectra) of the halftone separations to obtain good estimates of the individual halftones from conventional RGB scans. We demonstrate the efficacy of this methodology experimentally using continuous phase modulation for the embedding of independent visual watermark patterns in the individual halftone separations. Watermarks detected from the estimates of halftone separations obtained using the proposed estimation method have a much higher contrast than those detected directly. We also evaluate the accuracy of the estimated halftones through simulations and demonstrate that the proposed estimation method offers high accuracy.

Based on sharing and lattice-embedding techniques, we present an authentication-recovery method for an image. The recovery data are shared among many shadows, then lattice-embedding is utilized to embed each shadow in the discrete cosine transform domain of an 8×8 block. The proposed method can resist certain content-preserving operations such as JPEG compression, Gaussian noise, and brightness adjustment, up to a tolerance level controlled by a quantization parameter value. The method can also resist certain security attacks such as a cut-and-paste attack, a collage attack, and a vector quantization attack. Compared with previous works, the proposed method has following major advantages and novelty: (1) the method does not require prediction of the tampering trace, and the tampered blocks are always recovered as long as the number of valid blocks reaches a threshold and (2) lattice-embedding yields smaller distortion than does parity-check quantization, and the latter was often used in reported works.

We present a novel stereo imaging technique called dual-focus stereo imaging or DFSI. DFSI uses a pair of images captured from different viewpoints and at different foci, but with identical wide aperture size. Each image in an DFSI pair exhibits different defocus blur, and the two images form a defocused stereo pair. To model defocus blur, we introduce a defocus kernel map (DKM) that computes the size of the blur disk at each pixel. We derive a novel disparity defocus constraint for computing the DKM in DFSI, and integrate DKM estimation with disparity map estimation to simultaneously recover both maps. We show that the recovered DKMs provide useful guidance for segmenting the in-focus regions. We demonstrate using DFSI in a variety of imaging applications, including low-light imaging, automatic defocus matting, and multifocus photomontage.

We present a novel method for automatically correcting the radial lens distortion in a zoom lens video camera system. We first define the zoom lens distortion model using an inherent characteristic of the zoom lens. Next, we sample some video frames with different focal lengths and estimate their radial distortion parameters and focal lengths. We then optimize the zoom lens distortion model with preestimated parameter pairs using the least-squares method. For more robust optimization, we divide the sample images into two groups according to distortion types (i.e., barrel and pincushion) and then separately optimize the zoom lens distortion models with respect to divided groups. Our results show that the zoom lens distortion model can accurately represent the radial distortion of a zoom lens.

A local text alignment algorithm is introduced in this work for synchronizing transcripts. The proposed algorithm can be used for any transcript alignment process where high computational complexity is a concern. Dynamic programming is typically used to align a set of transcripts: however, the computational complexity of dynamic programming is high. To reduce the computational complexity, a local dynamic programming algorithm is introduced that aligns subsections of the transcripts. Aligning subsections of the transcripts greatly reduces the information needed for accurate synchronization. The information is reduced because it is not necessary to compare all words between the two transcripts. For example, words at the beginning of one transcript would not be compared to the words at the end of the other transcript. The subsection size is dependent on the total number of alignment errors between the transcripts. It is shown that the computational complexity of the proposed local dynamic programming algorithm is greatly reduced while preserving alignment accuracy.

Decision trees and their variants recently have been proposed. All trees used are fixed M-ary tree structured, such that the training samples in each node must be artificially divided into a fixed number of branches. This work proposes a fuzzy variable-branch decision tree (FVBDT) based on the fuzzy genetic algorithm (FGA). The FGA automatically searches for the proper number of branches of each node according to the classification error rate and the classification time of FVBDT. Therefore, FGA reduces both the classification error rate and classification time, and then optimizes the FVBDT. In our experiments, FVBDT outperforms the traditional C-fuzzy decision tree (CFDT) based on the fuzzy C-means (FCM) algorithm.

Much research has shown that perceived image contrast increases as the surround luminance increases, but a number of recent studies reported opposite trends under higher surround luminance levels. We measured the change in perceived image contrast under a wide range of surround luminance levels covering from dark up to 2087 cd/m². A large-area illuminator was used to illuminate the surround. It consists of 23 dimmable fluorescent lamps and a diffuser. Its maximum luminance is 2087 cd/m² and could be adjusted to six lower levels. A set of paired comparison experiments was conducted to compare the perception of image contrast under seven different surround luminance levels. The results showed that the perceived image contrast varies with surround luminance and the maximum perceived image contrast is found near a surround ratio (S_R) of 1. As S_R increases from 0 to 1, the z score is increased, which can be fully expected by the Bartleson and Breneman effect. However, it is drastically decreased in the region of S_R > 1; thus, the perceived image contrast is eventually decreased.

We examine the relationship between controlled and uncontrolled visual preference tests. We compare the results for the preference of printed images in various viewing environments. The data are examined with regard to different numbers of observer subsets and we can derive an experimental guideline for an equivalence of the controlled and uncontrolled preference experiments based on the certainty of the expected result.

We examine one way to extend recently proposed wavelet-based maximum a posteriori estimation rules for image denoising to video. The proposed approach takes into account both spatial and temporal dependencies between wavelet coefficients, and is general enough to incorporate different spherically contoured prior distributions on noiseless coefficients, as well as different spatiotemporal coefficient neighborhoods. Presented extensions of the algorithm have reasonable complexity and are suited to vectorized, convolution-based implementations.

We apply the dynamic differential evolution (DDE) algorithm to solve the inverse scattering problem for which a two-dimensional perfectly conducting cylinder with unknown cross section is buried in a dielectric slab medium. The finite-difference time domain method is used to solve the scattering electromagnetic wave of a perfectly conducting cylinder. The inverse problem is resolved by an optimization approach, and the global searching scheme DDE is then employed to search the parameter space. By properly processing the scattered field, some electromagnetic properties can be reconstructed. One is the location of the conducting cylinder, the others is the shape of the perfectly conducting cylinder. This method is tested by several numerical examples, and it is found that the performance of the DDE is robust for reconstructing the perfectly conducting cylinder. Numerical simulations show that even when the measured scattered fields are contaminated with Gaussian noise, the quality of the reconstructed results obtained by the DDE algorithm is very good.

Infrared imaging is based on the measurement of radiation of an object and its conversion to temperature. A vital parameter of the conversion procedure is emissivity, which defines the capability of a material to absorb and radiate energy. For most applications, emissivity is assumed to be constant. In applications measuring the temperature of objects with high emissivity, this is not problematic, as slight variations in the chosen emissivity value cause only minor changes in the resulting surface temperatures. However, when emissivities are low, as in steel strips, considering emissivity as a constant can lead to significant errors in temperature measurement. To overcome problems generated by variations in emissivity, one solution is to measure temperature where the steel strip forms a wedge, acting as a cavity. In the deepest part of the wedge, emissivity is sufficiently close to one. This work presents a real time image processing system to acquire infrared line scans for steel strips using the wedge method. The proposed system confronts two challenges: extracting infrared line scans in real time from the deepest part of the wedge in rectangular infrared images, and translating pixels from the line scan into real-world units.

Color demosaicking is critical to the image quality of single-sensor-based imaging devices. Caused by the sampling pattern of color filter array (CFA), the demosaicked images typically suffer from visual color artifacts in regions of high frequency and sharp edge structures, degrading the quality of camera output. We present a new high-quality demosaicking algorithm by taking advantage of deinterlacing and median-based filtering techniques. We treat the sampled green data of Bayer CFA as a form of diagonal interlaced green planes and make use of some key concepts about spatial deinterlacing to help the edge estimation in terms of both various directions and accuracy. In addition, a specific edge feature, sharp line edge of width 1 pixel, can also be handed well by the proposed method. The median-based filtering techniques are developed for suppressing most visual demosaicking artifacts, such as zipper effect, false color artifact, and interpolation artifact. Experimental results show that our algorithm is effective in suppressing visual artifacts, preserving the edges of image with sharpness and satisfying visual inspection, while keeping computational efficiency.

The design and testing of an active 190-GHz imaging system is presented. The system features two beam-scanning antennas, one of which transmits a vertical fan beam, and the other which receives a horizontal fan beam. By correlating the transmitted and received signals, an output is obtained that is proportional to the millimeter-wave reflectivity at the intersection of the two fan beams. Beam scanning is obtained by rotating a small subreflector within each antenna, allowing rapid scanning. The system has an angular resolution of 0.3 deg, a field of view of 14×14 deg, and operates at a standoff distance of 5 m.

We propose a joint-source-channel coding (JSCC) scheme that can provide and sustain high-quality video service in spite of deteriorated transmission channel conditions of the second generation of the digital video broadcasting (DVB-S2) satellite broadcasting service. Especially by combining the layered characteristics of the SVC (scalable video coding) video and the robust channel coding capability of LDPC (low-density parity check) employed for DVB-S2, a new concept of JSCC for digital satellite broadcasting service is developed. Rain attenuation in high-frequency bands such as the Ka band is a major factor for lowering the link capacity in satellite broadcasting service. Therefore, it is necessary to devise a new technology to dynamically manage the rain attenuation by adopting a JSCC scheme that can apply variable code rates for both source and channel coding. For this purpose, we develop a JSCC scheme by combining SVC and LDPC, and prove the performance of the proposed JSCC scheme by extensive simulations where SVC coded video is transmitted over various error-prone channels with AWGN (additive white Gaussian noise) patterns in DVB-S2 broadcasting service.

We investigate the use of a new binary coherent vector approach, integrated in a proposed content-based medical retrieval (CBMIR) system, to retrieve computed tomography (CT) brain images. Five types of hemorrhages consisting of 150 plain axial CT brain images are queried from a database of 2500 normal and abnormal CT brain images. Possible combinations of shape features are portrayed as feature vectors and are evaluated based on precision-recall plots. Solidity, form factor, equivalent circular diameter (ECD), and Hu moment are proposed as identifying features of intracranial hemorrhages in CT brain images. In addition to identifying hemorrhages, the proposed approach significantly improves the CBMIR system performance. This retrieval system can be widely useful due to rapid development in computer vision and computer database management, both of which motivated this application of CBMIR.

Acousto-optic tunable filters (AOTFs) can be used as spectral filters for the implementation of multispectral imaging systems. However, obtaining quality images is challenging. In this work, we propose several improvements that enable the use of these systems in quantitative spectroscopic imaging applications. The improvements are based on three pillars: 1. a finer spectral bandpass shaping by dynamically optimizing the radio frequency (rf) driving signal, 2. an extensive calibration process, and 3. careful image preprocessing that uses calibration data to correct some well known AOTF issues in imaging applications. A novel multispectral imaging instrument is built using commercial off-the-shelf components. The instrument includes an Isomet (Springfield, New Jersey) AOTF working in the visible and near-infrared range, and a new concept of rf generator based on a high-speed digital-to-analog converter that allows the generation of multiband signals. The ancillary control software performs the main part of the image optimization process: an initial calibration, a dynamic adjustment of the rf driving signal power and exposure time, and finally the radiometric preprocessing of the acquired multispectral images. Finally, some results of the instrument performance are presented that show the achieved spectral and spatial resolution on different imaging scenarios.

A novel technique for searching for complex patterns in large multimedia databases is presented, based on rotation independent template matching. To handle objects of arbitrary shape while reducing the computational workload, the pattern to be localized is partitioned into small square blocks of sizes adapted to the local image content using quadtree decomposition. The use of Zernike polynomials for representing each block allows the design of a fast and effective maximum likelihood matching procedure to sequentially verify whether the target image contains each block of the quadtree. State of the art methods usually represent the whole pattern by using an orthogonal basis and extracting an invariant feature vector from the representation coefficients. In the proposed scheme, the use of the quadtree decomposition allows us to bind the number of terms of the truncated expansions, still guaranteeing a precise image representation.

An engineering design framework for an autonomous ground vehicle vision system is discussed. We present both the conceptual and physical design by following the design process, development and testing of an intelligent ground vehicle vision system constructed for the 2008 Intelligent Ground Vehicle Competition. During conceptual design, the requirements for the vision system are explored via functional and process analysis considering the flows into the vehicle and the transformations of those flows. The conceptual design phase concludes with a vision system design that is modular in both hardware and software and is based on a laser range finder and camera for visual perception. During physical design, prototypes are developed and tested independently, following the modular interfaces identified during conceptual design. Prototype models, once functional, are implemented into the final design. The final vision system design uses a ray-casting algorithm to process camera and laser range finder data and identify potential paths. The ray-casting algorithm is a single thread of the robot's multithreaded application. Other threads control motion, provide feedback, and process sensory data. Once integrated, both hardware and software testing are performed on the robot. We discuss the robot's performance and the lessons learned.

suboptimal traffic management, with unnecessary delays, higher fuel consumption, and higher emissions. Traffic management can be improved by installing inductive loops; however, installation involves temporary road closures and high maintenance costs, especially if there is normally a lot of heavy traffic on the road. We present a vehicle detection and counting system based on digital image-processing techniques. These images can be taken by digital cameras installed at the top of existing traffic lights. By using the proposed approach, it is possible to detect the number of vehicles waiting on each side of the intersection, hence, providing the necessary information for optimal traffic management. Results achieved after testing this methodology on three real intersections are promising, attaining high accuracy during the day (98.8%) and the night (91.3%) while counting several vehicles at the same time. Hence, the system is equivalent to installing multiple inductive loops in all the streets of the intersection, but with lower installation and maintenance costs. After integrating the proposed algorithms into a traffic-management system, it was possible to reduce fuel and CO₂ emissions by half compared to the standard fixed-time scheduler.

This PDF file contains the editorial “Sparse Image and Signal Processing: Wavelets, Curvelets, Morphological Diversity," by Jean-Luc Starck, Fionn Murtagh, and Jalal M. Fadili for JEI Vol. 19 Issue 04