We show that the two countermeasures proposed in a paper on the security of ownership watermarking of digital images based on singular value decomposition by Loukhaoukha and Chouinard do not solve the false-positive detection problem in contrast to designers' claim and therefore should not be used for proof of ownership application.

This paper proposes a new colorization method based on the chrominance blending. The weights for the blending are computed by using the random walker algorithm, which is a soft segmentation technique that provides sharp probability transition on object boundaries. As a result, the proposed method reduces color bleeding and provides improved colorization performances compared to conventional ones.

Reversible watermarking inserts watermarks into digital media in such a way that visual transparency is preserved, which enables the restoration of the original media from the watermarked one without any loss of media quality. It has various applications where high capacity and high visual quality are major requirements for reversible watermarking. This work presents a new reversible watermarking algorithm that embeds message bits by modifying the differential histogram of adjacent pixels. To satisfy both high embedding capacity and visual quality, the proposed technique exploits the fact that the adjacent pixels are highly correlated. Also, we prevent overflow and underflow problems by designing a predicted error compensation scheme. Through experiments using multiple kinds of test images, we prove that the presented algorithm provides 100% reversibility, higher capacity, and higher visual quality than any previous method, while maintaining low induced distortion.

Light source calibration is an important issue in many computer vision fields such as photometric stereo and shape from shading. Spheres, with either diffuse or specular reflections, are frequently deployed as calibration objects to recover the direction and intensity of light source from images. We present a novel method for light source calibration by using a planar mirror with a chessboard pattern and a diffuse region. The light direction can be accurately estimated from one mirror orientation by recovering the normal direction of the mirror plane. The location and intensity of light source can be further estimated if two mirror orientations are used. Experimental results show that the calibration accuracy of the proposed method is much higher than traditional sphere-based techniques and can offer improved three-dimensional reconstruction in photometric stereo.

The k-modes algorithm was recently proposed to cluster mixed-type data. However, in solving clustering problems, the k-modes algorithm and its variants usually ask the user to provide the number of clusters in the data sets. Unfortunately, the number of clusters is generally unknown to the user. Therefore, clustering becomes a tedious task of trial-and-error and the clustering result is often poor, especially when the number of clusters is large and not easy to guess. Also, it is hard for a user to select the weight between categorical and numeric attributes in the k-modes algorithm. In this paper, a genetic algorithm for clustering large data sets with mixed-type data is proposed, and this algorithm can automatically search the number of clusters in the data set. Also, a weight can be automatically selected by the genetic algorithm to prevent favoring either type of attribute. Experimental results illustrate the effectiveness of the genetic algorithm.

Based on the cellular neural network (CNN) paradigm, the bio-inspired (bi-i) cellular vision system is a computing platform consisting of state-of-the-art sensing, cellular sensing-processing and digital signal processing. This paper presents the implementation of a novel CNN-based segmentation algorithm onto the bi-i system. The experimental results, carried out for different benchmark video sequences, highlight the feasibility of the approach, which provides a frame rate of about 26 frame/sec. Comparisons with existing CNN-based methods show that, even though these methods are from two to six times faster than the proposed one, the conceived approach is more accurate and, consequently, represents a satisfying trade-off between real-time requirements and accuracy.

We present an analysis of dark current from a complementary metal-oxide-semiconductor (CMOS) active pixels sensor with global shutter. The presence of two sources of dark current, one within the collection area of the pixel and another within the sense node, present complications to correction of the dark current. The two sources are shown to generate unique and characteristic dark current behavior with respect to varying exposure time, temperature, and/or frame rate. In particular, a pixel with storage time in the sense node will show a dark current dependence on frame rate and the appearance of being a "stuck pixel" with values independent of exposure time. On the other hand, a pixel with an impurity located within the collection area will show no frame rate dependence, but rather a linear dependence on exposure time. A method of computing dark frames based on past dark current behavior of the sensor is presented and shown to intrinsically compensate for the two different and unique sources. In addition, dark frames requiring subtraction of negative values, arising from the option to modify the bias offset, are shown to be appropriate and possible using the computational method.

We propose a method for surface reconstruction of artist paintings. In order to reproduce the appearance of a painting, including color, surface texture, and glossiness, it is essential to acquire the pixel-wise light reflection property and orientation of the surface and render an image under an arbitrary lighting condition. A photometric approach is used to estimate bidirectional reflectance distribution functions (BRDFs) and surface normals from a set of images photographed by a fixed camera with sparsely distributed point light sources. A robust and computationally less expensive nonlinear optimization algorithm is proposed that optimizes the small number of parameters to simultaneously determine all of the specular BRDF, diffuse albedo, and surface normal. The proposed method can be applied to moderately glossy surfaces without separating captured images into diffuse and specular reflections beforehand. Experiments were conducted using oil paintings with different surface glossiness. The effectiveness of the proposed method is validated by comparing captured and rendered images.

To achieve auto exposure in digital cameras, image brightness is widely used because of its direct relationship with exposure value. To use image entropy as an alternative statistic to image brightness, it is required to establish how image entropy changes as exposure value is varied. This paper presents a mathematical proof along with experimental verification results to show that image entropy reaches a maximum value as exposure value is varied by changing shutter speed or aperture size.

Moiré in color printing is an undesirable visible artifact that can arise from overlaying multiple halftone color separations. Halftone geometric configurations designed to avoid moiré in the overlays strictly require that individual halftone color separations must possess a low degree of relative distortion. However, optical and mechanical errors of multiple imaging systems within a printer usually produce differences between the color planes in the trajectory and placement of the exposure spots. We study color halftone moiré due to these optical and mechanical errors for otherwise moiré-free halftone configurations. Distortions due to commonly used imaging systems in xerography (i.e., raster output scanners and image bars) are categorized into two classes that depend on the direction of the displacement errors [i.e., process direction distortions (such as shear, bow, and skew) and cross-process direction distortions (such as scanline magnification, magnification imbalance, and high-order scanline distortions)]. Using frequency vector representation of color halftones, we derive analytical expressions for acceptability bounds on these distortions. We evaluate the analytical expressions for a classical halftone screen configuration and a minimum rosette geometry to enable specification allocations for different imaging components in the design of an imaging system.

This article describes a mechanism that segments an image by optimizing a homogeneity measure in a level-set formulation. The mechanism has uniform treatment toward texture, gray level, and color boundaries. The mechanism commences by quantizing the entire range of intensity values into a finite number of classes, thereby approximating the image as a class map, in which each image position carries a particular class label. For any region, each class label in it generally appears in a number of positions, which together define the spread-size of the class label in the region. The average spread-size of the class labels in a region, as a ratio to the region size, then constitutes a measure of how homogeneous the region is. The segmentation problem can be formulated as the optimization of such a measure of all the segmented regions. This work contributes chiefly by expressing the above optimization functional in such a way that the optimization can be conducted in a variational formulation, and that the solution can be reached by the deformation of an active contour. In addition, this work incorporates an additional geodesic term into the optimization functional, which maintains the contour's mobility at certain adverse conditions.

We present a system for automatic, on-line visual inspection and print defect detection for variable data printing (VDP). This system can be used to automatically stop the printing process and alert the operator to problems. We lay out the components required for constructing a vision-based inspection system and show that our approach is novel for the high-speed detection of defects on variable data. When implemented in a high-speed digital printing press, the system allows a single skilled operator to monitor and maintain several presses, reducing the number of operators required to run a shop floor of presses as well as reduce wasted consumables when a defect goes undetected.

Main subject detection (MSD) refers to the task of determining which spatial regions in an image correspond to the most visually relevant or scene-defining object(s) for general viewing purposes. This task, while trivial for a human, remains extremely challenging for a computer. Here, we present an algorithm for MSD which operates by adaptively refining low-level features. The algorithm computes, in a block-based fashion, five feature maps corresponding to lightness distance, color distance, contrast, local sharpness, and edge strength. These feature maps are adaptively combined and gradually refined via three stages. The final combination of the refined feature maps produces an estimate of the main subject's location. We tested the proposed algorithm on two extensive image databases. Our results show that relatively simple, low-level features, when used in an adaptive and iterative fashion, can be very effective at MSD.

In this paper, an enhanced algorithm is proposed to detect foot inflammation and, hence, predict ulcers before they can develop. This algorithm is based on an asymmetry analysis combined with a segmentation technique with a genetic algorithm to achieve higher efficiency in the detection of inflammation. The analysis involves several steps: segmentation, geometry transformation, overlapping, and abnormality identification. To enhance the results of this analysis, an additional step, features extraction, is performed. In this step, low and high order statistics are computed for each foot. Preliminary results show that the proposed algorithm combined with features extraction can be reliable and efficient to predict potential ulceration.

The iris is currently believed to be one of the most accurate biometrics for human identification. The majority of fielded iris identification systems use fractional Hamming distance to compare a new feature template to a stored database. Fractional Hamming distance is extremely fast, but mathematically weights all regions of the iris equally. Research has shown that different regions of the iris contain varying levels of discriminatory information when using circular boundary assumptions. This research evaluates four statistical metrics for accuracy improvements on low resolution and poor quality images. Each metric statistically weights iris regions in an attempt to use the iris information in a more intelligent manner. A similarity metric extracted from the output stage of an artificial neural network demonstrated the most promise. Experiments were performed using occluded, subsampled, and motion blurred images from the CASIA, University of Bath, and ICE 2005 databases. The neural network-based metric improved accuracy at nearly every operating point.

Fast variable block-size motion estimation is a key issue for real-time applications of the H.264, whereas the subpixel refinement takes up much computational time as compared to integer-pixel motion estimation. We propose a new fast subpixel precision variable block-size motion-estimation scheme. This algorithm uses the statistical information, which comes from the motion activities of the macroblocks (MBs) in the previous frame, to predict the characteristics of MBs in the current frame. Additionally, the distortion values and motion vectors of MBs in the previous frame are also considered as prior knowledge, based on which we can make decisions on early mode selection and early termination, and on whether or not to skip some candidate modes and candidate checking points. The intermediate results of subpixel motion estimation are used together with the prior knowledge to reduce subpixel search time when searching for stationary blocks. Our new directional information strategy is used in both integer-pixel motion estimation and subpixel motion estimation to accelerate the search procedure. Moreover, our algorithm can eliminate the subpixel motion estimation of all the unselected subpartition modes. The computational resources can then be spent on the modes and locations that deserve to be searched more than others. Extensive experimental is been done, the results of which show that the speed of our approach is nearly five times that of the fast algorithms in H.264 JM, with a better peak signal-to-noise ratio and better bit performance.

General purpose computer designers have recently begun adding cores to their processors in order to increase performance. For example, Intel has adopted a homogeneous quad-core processor as a base for general purpose computing. PlayStation3 (PS3) game consoles contain a multicore heterogeneous processor known as the Cell, which is designed to perform complex image processing algorithms at a high level. Can modern image-processing algorithms utilize these additional cores? On the other hand, modern advancements in configurable hardware, most notably field-programmable gate arrays (FPGAs) have created an interesting question for general purpose computer designers. Is there a reason to combine FPGAs with multicore processors to create an FPGA multicore hybrid general purpose computer? Iris matching, a repeatedly executed portion of a modern iris-recognition algorithm, is parallelized on an Intel-based homogeneous multicore Xeon system, a heterogeneous multicore Cell system, and an FPGA multicore hybrid system. Surprisingly, the cheaper PS3 slightly outperforms the Intel-based multicore on a core-for-core basis. However, both multicore systems are beaten by the FPGA multicore hybrid system by >50%.

Although current blind image steganalysis systems utilize a wide variety of features and classifiers, a common shortcoming in all of them is that they almost have similar processes for all images and they do not take advantage of the content diversity of different images. In this paper, a new framework is proposed that enables us to employ the content of images in these systems. All blind image steganalysis methods can be adapted to the proposed framework. In the training phase of our framework, the input images are first divided into classes according to an image content evaluation criterion and then the training process is specialized for each class. In the testing phase, a fuzzy approach is used to include different classes in the decision making process. Experimental results demonstrate that the proposed framework significantly enhance the detection accuracy of these systems.

This paper proposes a new regularization scheme to address blind color image deconvolution. Color images generally have a significant correlation among the red, green, and blue channels. Conventional blind monochromatic deconvolution algorithms handle each color image channels independently, thereby ignoring the interchannel correlation present in the color images. In view of this, a unified regularization scheme for image is developed to recover edges of color images and reduce color artifacts. In addition, by using the color image properties, a spectral-based regularization operator is adopted to impose constraints on the blurs. Further, this paper proposes a reinforcement regularization framework that integrates a soft parametric learning term in addressing blind color image deconvolution. A blur modeling scheme is developed to evaluate the relevance of manifold parametric blur structures, and the information is integrated into the deconvolution scheme. An optimization procedure called alternating minimization is then employed to iteratively minimize the image- and blur-domain cost functions. Experimental results show that the method is able to achieve satisfactory restored color images under different blurring conditions.

In this paper, we describe a block-based threshold modulation error diffusion algorithm to parallelize the halftone process without generating both block-boundary and diagonal artifacts. A novel scan path is used to pass the quantization error effectively between blocks and the input-dependent threshold modulation is applied during the inner-block processing. To obtain a suitable parameter set for the error weights and threshold modulation strength, a cost function is designed. Experimental results show that the proposed algorithm generates high quality halftone images which are visually similar to those generated by serial error diffusion algorithms. Our algorithm achieves better performance both in quality and parallelism compared to other halftoning approaches.

We propose an optical image encryption algorithm based on Arnold transform and gyrator transform. The amplitude and phase, which are the outputs of gyrator transform, are separated into several sub-images. Arnold transform is introduced for scrambling the data of the sub-images. The random spectrum composed of the scrambled sub-images is transformed by gyrator transform. An iterative structure of the algorithm is designed for enhancing the security of the encryption algorithm. The parameters of gyrator transforms and separating scheme serve as the key of the encryption method. The encryption process can be implemented by an electro-optical setup. Some numerical simulations have been given to demonstrate the security and validity of this algorithm.

This paper proposes a visual cryptography (VC) by a random grid (RG) scheme with identifiable shares. The method encodes an image O in two shares that exhibits the following features: 1. each generated share is a RG with the same scale as O; 2. any share singly reveals no secret information on O; 3. the secrets can be revealed by superimposing the two generated shares; 4. folding up a share will display the identification patterns associated with it; and 5. both of the secret information and the identification patterns are recognized by the naked eye without any computation. The property to show identification patterns by simply folding it up is novel to visual cryptographic schemes in the literature. It provides useful information to identify the pair of shares belonging to a secret image, and establishes a simple and friendly interface for users to distinguish among and manage the numerous shares created by VC schemes.

The development of the information technology and computer networks facilitates easy duplication, manipulation, and distribution of digital data. Digital watermarking is one of the proposed solutions for effectively safeguarding the rightful ownership of digital images and video. We propose a public digital watermarking technique for video copyright protection in the discrete wavelet transform domain. The scheme uses binary images as watermarks. These are embedded in the detail wavelet coefficients of the middle wavelet subbands. The method is a combination of spread spectrum and quantization-based watermarking. Every bit of the watermark is spread over a number of wavelet coefficients with the use of a secret key by means of quantization. The selected wavelet detail coefficients from different subbands are quantized using an optimal quantization model, based on the characteristics of the human visual system (HVS). Our HVS-based scheme is compared to a non-HVS approach. The resilience of the watermarking algorithm is tested against a series of different spatial, temporal, and compression attacks. To improve the robustness of the algorithm, we use error correction codes and embed the watermark with spatial and temporal redundancy. The proposed method achieves a good perceptual quality and high resistance to a large spectrum of attacks.

This paper describes a computational method using tensor math for higher order singular value decomposition (HOSVD) of registered images. Tensor decomposition is a rigorous way to expose structure embedded in multidimensional datasets. Given a dataset of registered 2-D images, the dataset is represented in tensor format and HOSVD of the tensor is computed to obtain a set of 2-D basis images. The basis images constitute a linear decomposition of the original dataset. HOSVD is data-driven and does not require the user to select parameters or assign thresholds. A specific application uses the basis images for pixel-level fusion of registered images into a single image for visualization. The fusion is optimized with respect to a measure of mean squared error. HOSVD and image fusion are illustrated empirically with four real datasets: (1) visible and infrared data of a natural scene, (2) MRI and x ray CT brain images, and in nondestructive testing (3) x ray, ultrasound, and eddy current images, and (4) x ray, ultrasound, and shearography images.

In this paper a new filter for image sequences corrupted with random impulse noise is presented. In order to preserve the image details as much as possible, the noise is removed in different successive filtering steps. In each step, only the pixels that are detected as being noisy are filtered, while the noise-free pixels remain unchanged. The noise detection is based on fuzzy set theory and fuzzy rules, which are very useful for the processing of human knowledge and linguistic values. To exploit the temporal information in image sequences as much as possible, detected pixels are finally filtered in a motion compensated way. From the experimental results it can be seen that the proposed method outperforms other state-of-the-art filters both in terms of the peak-signal-to-noise ratio, the mean absolute error, and visually.

A real-time video watermarking scheme against geometric distortions is proposed for DCT-encoded compressed video data. The full DCT coefficients have proven to be invariant to scaling and local geometric attacks. Therefore, watermarks are embedded into watermark minimal sequences (WMS) by modulating the low-frequency full DCT coefficients. To meet the requirement of real-time performance, a fast intertransformation is employed to construct full DCT directly from block DCTs. The modified differences of full DCT coefficients are inversely transformed to the differences of block DCT coefficients that are subsequently added to the original block DCTs to generate the WMS. To resist various video format conversions, the watermark detection is performed in spatial domain. This does not affect the real-time performance because the watermark detection need not re-encode the frames into compressed data, and compressed video can be easily decoded fast by employing some codec or decoding chips. Furthermore, the proposed scheme can resist rotation attacks by employing a rotation compensation strategy. Experimental results demonstrate that the proposed scheme is real-time, transparent, and robust to signal distortions, geometric distortions including rotation, scaling, aspect ratio change, linear geometric transforms, cropping and combination attacks, frame dropping and swapping, file format conversion, as well as camcorder recording.

We present a new algorithm for reconstructing a three-dimensional (3-D) ultrasound image from a series of two-dimensional B-scan ultrasound slices acquired in the mechanical linear scanning framework. Unlike most existing 3-D ultrasound reconstruction algorithms, which have been developed and evaluated in the freehand scanning framework, the new algorithm has been designed to capitalize the regularity pattern of the mechanical linear scanning, where all the B-scan slices are precisely parallel and evenly spaced. The new reconstruction algorithm, referred to as the cyclic Savitzky-Golay (CSG) reconstruction filter, is an improvement on the original Savitzky-Golay filter in two respects: First, it is extended to accept a 3-D array of data as the filter input instead of a one-dimensional data sequence. Second, it incorporates the cyclic indicator function in its least-squares objective function so that the CSG algorithm can simultaneously perform both smoothing and interpolating tasks. The performance of the CSG reconstruction filter compared to that of most existing reconstruction algorithms in generating a 3-D synthetic test image and a clinical 3-D carotid artery bifurcation in the mechanical linear scanning framework are also reported.