Timely detection of packages that are left unattended in public spaces is a security concern, and rapid detection is important for prevention of potential threats. Because constant surveillance of such places is challenging and labor intensive, automated abandoned-object-detection systems aiding operators have started to be widely used. In many studies, stationary objects, such as people sitting on a bench, are also detected as suspicious objects due to abandoned items being defined as items newly added to the scene and remained stationary for a predefined time. Therefore, any stationary object results in an alarm causing a high number of false alarms. These false alarms could be prevented by classifying suspicious items as living and nonliving objects. In this study, a system for abandoned object detection that aids operators surveilling indoor environments such as airports, railway or metro stations, is proposed. By analysis of information from a thermal- and visible-band camera, people and the objects left behind can be detected and discriminated as living and nonliving, reducing the false-alarm rate. Experiments demonstrate that using data obtained from a thermal camera in addition to a visible-band camera also increases the true detection rate of abandoned objects.

We propose a new autofocus method for digital cameras based on the separation of color components from the incoming light rays and the measure of their disparity. For separating color components, we place two apertures with the red and blue color filters on them. This enables us to get two monochromatic images that have disparities proportional to the distances of the objects from the camera. We also propose a new measure to find the disparity of these color components because the conventional disparity measures show low accuracy for the pair of different color channel images. The measure is based on the observation that the overlap of images with disparity has many weak gradients, whereas the overlap with no disparity has a small number of strong gradients. One of two images is shifted from left to right, and the measure is computed for each position. Then, the position with the maximum measure is considered as the disparity, and the direction and distance of focus are computed from the estimated disparity and camera parameters. The proposed method is implemented in a commercial compact camera, and it has been demonstrated that the method finds the focus in a wide range of distance and illumination conditions.

Multiple reference frame motion estimation (MRF-ME) is one of the most crucial tools in H.264/AVC to improve coding efficiency. However, it disciplines an encoder by giving extra computational complexity. The required computation proportionally expands when the number of reference frames used for motion estimation increases. Aiming to reduce the computational complexity of the encoder, various motion vector (MV) composition algorithms for MRF-ME have been proposed. However, these algorithms only perform well in a limited range of reference frames. The performance deteriorates when motion vector composition is processed from the current frame to a distant reference frame. In this paper, a reliable tracking mechanism for MV composition is proposed by utilizing only the relevant areas in the target macroblock and taking different paths through a novel selection process from a set of candidate motion vectors. The proposed algorithm is especially suited for temporally remote reference frames in MRF-ME. Experimental results show that compared with the existing MV composition algorithms, the proposed one can deliver a remarkable improvement on the rate-distortion performance with similar computational complexity.

We describe and evaluate a fast implementation of a classical block-matching motion estimation algorithm for multiple graphical processing units (GPUs) using the compute unified device architecture computing engine. The implemented block-matching algorithm uses summed absolute difference error criterion and full grid search (FS) for finding optimal block displacement. In this evaluation, we compared the execution time of a GPU and CPU implementation for images of various sizes, using integer and noninteger search grids. The results show that use of a GPU card can shorten computation time by a factor of 200 times for integer and 1000 times for a noninteger search grid. The additional speedup for a noninteger search grid comes from the fact that GPU has built-in hardware for image interpolation. Further, when using multiple GPU cards, the presented evaluation shows the importance of the data splitting method across multiple cards, but an almost linear speedup with a number of cards is achievable. In addition, we compared the execution time of the proposed FS GPU implementation with two existing, highly optimized nonfull grid search CPU-based motion estimations methods, namely implementation of the Pyramidal Lucas Kanade Optical flow algorithm in OpenCV and simplified unsymmetrical multi-hexagon search in H.264/AVC standard. In these comparisons, FS GPU implementation still showed modest improvement even though the computational complexity of FS GPU implementation is substantially higher than non-FS CPU implementation. We also demonstrated that for an image sequence of 720 × 480 pixels in resolution commonly used in video surveillance, the proposed GPU implementation is sufficiently fast for real-time motion estimation at 30 frames-per-second using two NVIDIA C1060 Tesla GPU cards.

A fast pixel-size-based very large-scale enlargement and reduction of image via an adaptive combination of bilinear interpolation and discrete cosine transform (DCT) is proposed to improve the quality and speed of the image zooming process. The proposed algorithm combined with DCT and bilinear interpolation is proposed to adaptively process different blocks of an image with different characteristics for a very large-scale image enlargement and reduction. Moreover, if quality is the main concern factor, the evolutionary programming is applied to further improve the quality of reduction algorithm of an image by tuning the frequency coefficient matrix of the reduced image. Finally, for dealing with arbitrary pixel-size-based zooming of an image, a new algorithm is also proposed.

Residual bulk image (RBI) is significant in the KAF09000 CCD. Residual images are observed 2 h after illumination at −20C. Trap leakage and capacity are studied as a function of temperature. A flood-flush protocol is evaluated for eliminating RBI artifacts. A substrate trap fixed pattern noise is observed and is removed by dark-subtraction. An increase of dark current shot noise due to trap leakage will occur but can be minimized by deep cooling. Operating temperature targets are set as a function of target noise levels. An operating temperature of -87C for a 30 min exposure is projected to support a read noise constraint of 5 e-.

We propose novel 5/3 and 9/7 wavelet-based image interpolations in lifting structures with/without involving the evolutionary programming (EP) for image resolution enhancement, respectively. On the basis of the low-band wavelet coefficients, the proposed prediction algorithms estimate high-frequency wavelet coefficients of the original image, so that the original image can be reconstructed from the low-band wavelet coefficients by using the proposed prediction algorithms. Because of the characteristics of the odd-length wavelet filter, which is dominated by the intermediate coefficients, taking even or odd downsampling of the filtered coefficients significantly dominates the image reconstruction performance. Therefore, constructing the low-band wavelet coefficients through the odd-length wavelet filter in lifting structures based on even and odd pixels of the original image are respectively performed. Then, the better one, which yields a high-performance image resolution enhancement, is selected. To further improve the reconstruction performance and show the optimality of the above-mentioned proposed algorithms, we also present the EP-based image interpolations. The proposed prediction algorithms with/without EP outperform other standard methods.

In this paper, we evaluate the degradation of an image due to the implementation of a watermark in the frequency domain of the image. As a result, a watermarking method, which minimizes the impact of the watermark implementation on the overall quality of an image, is developed. The watermark is embedded in magnitudes of the Fourier transform. A peak signal-to-noise ratio is used to evaluate quality degradation. The obtained results were used to develop a watermarking strategy that chooses the optimal radius of the implementation to minimize quality degradation. The robustness of the proposed method was evaluated on the dataset of 1000 images. Detection rates and receiver operating characteristic performance showed considerable robustness against the print-scan process, print-cam process, amplitude modulated, halftoning, and attacks from the StirMark benchmark software.

All things considered, electronic imaging systems do not rival the human visual system despite notable progress over 40 years since the invention of the CCD. This work presents a method that allows design engineers to evaluate the performance gap between a digital camera and the human eye. The method identifies limiting factors of the electronic systems by benchmarking against the human system. It considers power consumption, visual field, spatial resolution, temporal resolution, and properties related to signal and noise power. A figure of merit is defined as the performance gap of the weakest parameter. Experimental work done with observers and cadavers is reviewed to assess the parameters of the human eye, and assessment techniques are also covered for digital cameras. The method is applied to 24 modern image sensors of various types, where an ideal lens is assumed to complete a digital camera. Results indicate that dynamic range and dark limit are the most limiting factors. The substantial functional gap, from 1.6 to 4.5 orders of magnitude, between the human eye and digital cameras may arise from architectural differences between the human retina, arranged in a multiple-layer structure, and image sensors, mostly fabricated in planar technologies. Functionality of image sensors may be significantly improved by exploiting technologies that allow vertical stacking of active tiers.

An interpolation method for a line-pruned image is presented. The goal is to alleviate the time complexity of the previous NEDI-8 (new edge directed interpolation) method. The idea of our solution is to use NEDI-4 selectively, applying the NEDI-4 for edge pixels while using a linear filter for others. For the linear filter, the interpolation coefficients for each pruned line are adaptively determined.

Visual fatigue is a common problem when viewing stereoscopic images. We propose a depth adjustment method that controls the amount of disparity in stereoscopic images using visual fatigue prediction and conducted subjective preference evaluation based on individual fusional response characteristics. Visual fatigue level is predicted by examining the characteristics of horizontal disparity. Viewers are classified into two groups (those with normal susceptibility to visual fatigue and those with high susceptibility to visual fatigue) according to individual fusional limit and speed of fusion, which are determined using a random dot stereogram test. Subjective preferences for the amount of depth adjustment are investigated based on the degree of fusion ability. Our experimental results support the suitability of the proposed depth adjustment method for reducing visual fatigue.

Wavelet transforms that have an adaptive low-pass filter are useful in applications that require the signal singularities, sharp transitions, and image edges to be left intact in the low-pass signal. In scalable image coding, the spatial resolution scalability is achieved by reconstructing the low-pass signal subband, which corresponds to the desired resolution level, and discarding other high-frequency wavelet subbands. In such applications, it is vital to have low-pass subbands that are not affected by smoothing artifacts associated with low-pass filtering. We present the mathematical framework for achieving 1-D wavelet transforms that have a spatially adaptive low-pass filter (SALP) using the prediction-first lifting scheme. The adaptivity decisions are computed using the wavelet coefficients, and no bookkeeping is required for the perfect reconstruction. Then, 2-D wavelet transforms that have a spatially adaptive low-pass filter are designed by extending the 1-D SALP framework. Because the 2-D polyphase decompositions are used in this case, the 2-D adaptivity decisions are made nonseparable as opposed to the separable 2-D realization using 1-D transforms. We present examples using the 2-D 5/3 wavelet transform and their lossless image coding and scalable decoding performances in terms of quality and resolution scalability. The proposed 2-D-SALP scheme results in better performance compared to the existing adaptive update lifting schemes.

We present an adaptive image-acquisition methodology by replacing the traditional birefringent filter with slight out-of-focus blur generated by the camera lens. Because many cameras already have adjustable lenses and autofocus systems, our method can exploit existing hardware by simply changing the focusing strategy. During the image acquisition, the optimal defocus setting is automatically adapted to the power spectrum of the scene, which is evaluated by a generic autocorrelation model. We develop a criterion to estimate reconstruction errors without the baseband knowledge of the scene. This metric helps the camera to choose the optimal defocus settings. An optimal Wiener filter then recovers the captured scene and yields sharper images with reduced aliasing. The numerical and visual results show that our method is superior to current acquisition methods used by most digital cameras.

Since image watermarking has become an important tool for intellectual property protection and authentication, a wavelet-based watermarking scheme for color images is proposed. The watermarking scheme is based on implementing the generalized squared interpoint distance (GSID) in designing a weighting function to be used in a spread-spectrum fashion watermarking (SSW) and is denoted as SSW-GSID. Watermark embedding process is carried out by transforming the host image into the wavelet domain. For highly robust and imperceptible embedding, watermark bits are added to the discrete wavelet transform (DWT) coefficients of all subbands after being weighted using the product of the calculated GSID values (of the host image) and the basis function amplitudes of the DWT decomposition. Experimental results have shown that SSW-GSID exhibits highly reliable extraction of the watermark from attacked images. Based on the provided experimental results, it can be observed that the proposed SSW-GSID method is robust against a wide variety of attacks. Comparison with other existing methods shows the superiority of the proposed SSW-GSID method.

A fuzzy similarity measure-based hybrid image filter (FHF) is proposed for color image restoration in this paper. Operation is carried out in three steps: parameter optimization, hybrid image filter setup, and image restoration. For parameter optimization, a multimethodology evolutionary computation (MMEC) is presented for real-parameter optimization problems. Then, FHF with a fuzzy-based similarity measure is introduced for noise reduction. Finally, a color image is restored with an experience-based construction of FHF which has been optimized via MMEC. Experimental results show the proposed FHF achieves a high peak signal-to-noise ratio and mean structural similarity by effectively reducing Gaussian, impulse, and mixed-noise.

In this paper, we propose a context adaptive nonlinear diffusion method for image denoising in wavelet domain which we call context based diffusion in stationary wavelet domain (SWCD). In diffusing detail coefficients, the method adapts to the local context such that strong edges are preserved and smooth regions are diffused in a greater extent. The local context which is derived directly from the transform energies at scales 1 and 2 of two-level stationary wavelet transform (SWT) controls the diffusion. The shift invariance of SWT contributes to the performance of the method. The experiment is conducted on a number of benchmark images and compared to recently developed denoising methods which explore the adaptation concept for wavelet shrinkage and diffusion. A comparison is performed also to a method of diffusing both approximation and detail coefficients. The proposed SWCD method outperforms recently proposed adaptive shrinkage and adaptive diffusion, particularly at high noise levels. The method is computationally efficient due to the Haar wavelet and fast convergence attained due to exploiting the context information.

A book review of a recently released title.