The limited depth of field of optical lenses, makes multi-focus image fusion (MFIF) algorithms of vital importance. Lately, Convolutional Neural Networks (CNN) have been widely adopted in MFIF methods, however their predictions mostly lack structure and are limited by the size of the receptive field. Moreover, since images have noise due to various sources, the development of MFIF methods robust to image noise is required. A novel robust to noise Convolutional Neural Network-based Conditional Random Field (mf-CNNCRF) model is introduced. The model takes advantage of the powerful mapping between input and output of CNN networks and the long range interactions of the CRF models in order to reach structured inference. Rich priors for both unary and smoothness terms are learned by training CNN networks. The α -expansion graph-cut algorithm is used to reach structured inference for MFIF. A new dataset, which includes clean and noisy image pairs, is introduced and is used to train the networks of both CRF terms. A low-light MFIF dataset is also developed to demonstrate real-life noise introduced by the camera sensor. Qualitative and quantitative evaluation prove that mf-CNNCRF outperforms state-of-the-art MFIF methods for clean and noisy input images, while being more robust to different noise types without requiring prior knowledge of noise.

Abstract The authors propose a method for breast dynamic contrast enhanced‐magnetic resonance imaging classification by combining radiomic texture analysis with a hybrid adaptive neuro‐fuzzy inference system (ANFIS)‐particle swarm optimization (PSO) classifier. The fast discrete curvelet transform is utilized as a decomposition scheme in multiple scales. The mean and entropy features extracted from the produced scheme are used as texture descriptors. Principal component analysis (PCA) involves reduction of the dimensionality of the initial feature set. The transformed feature vector is subsequently introduced to a hybrid ANFIS‐PSO classifier. The average overall classification power of the proposed hybrid ANFIS‐PSO classifier is comparatively assessed to that obtained using several classifiers (ANFIS, linear discriminant analysis, Naïve Bayes, artificial neural networks, random forest and support vector machine) by using the 70 training‐30 testing data ratio. The comparison performed highlights the superiority of the proposed methodology, thus underlying the potential of ANFIS‐PSO for the breast cancer diagnosis with a classification accuracy of 94%.

In this paper, a novel multi-focus image fusion algorithm based on conditional random field optimization (mf-CRF) is proposed. It is based on an unary term that includes the combined activity estimation of both high and low frequencies of the input images, while a spatially varying smoothness term is introduced, in order to align the graph-cut solution with boundaries of focused and defocused pixels. The proposed model retains the advantages of both spatial-domain methods and multi-spectral methods and by solving an energy minimization problem and finds an optimal solution for the multi-focus image fusion problem. Experimental results demonstrate the effectiveness of the proposed method that outperforms current state-of-the-art multi-focus image fusion algorithms in both qualitative and quantitative comparisons. In this paper, the successful application of the mf-CRF model in multi-modal image fusion (visible-infrared and medical) is also presented.

Mathematical tools that are widely employed for the distribution-free analysis of the finite-sample performance of empirical estimators-like Efron-Stein's inequality, or Azuma's martingale inequality and derivatives-rely on the provision of tight bounds for the Lipschitz constants of the estimators. Lipschitz constants can be easily derived for simple functionals of the empirical measures, such as the Shannon or the Tsallis entropy. However, obtaining tight bounds for more generalized entropy functionals which cannot be decomposed into sums of identical terms is much more involved. The goal in this paper is the derivation of (empirical) Lipschitz constants for the maximum likelihood estimator of the family of Renyi entropies of order λ > 0, λ ≠ 1. Analytic solutions for the optimal constants are obtained for the most important special cases; namely, for 02, λ ≠ 3), an efficient way to compute the optimal constants is derived by reducing the complexity of the underlying optimization problem from exponential Ω(n∥A∥-1) to linear O(n), where n the number of available samples and ∥A∥ ≥ 2 the number of symbols of the source (with the assumption ∥A∥ does not grow with n). The optimal constants are subsequently used for the distribution-free performance analysis of the maximum likelihood estimator of Renyi entropy; this includes variance and concentration bounds, both under the assumption of independence as well as under strong mixing conditions.

Abstract A method based on the combination of Local Binary Pattern operator and radial lengths is presented aiming at the identification of Architectural Distortions (ADs) in mammograms. Local Binary Pattern operator, a number of its variants, and radial lengths are combined together producing a high‐dimensional feature space. A process, based on the combination of Principal Component Analysis and ttest, is used to effectively transform feature space and reveal the most descriptive features. The classification step is performed using a Support Vector Machine classifier. Open access databases (Mammographic Image Analysis Society and Digital Database for Screening Mammography) are used through an exhaustive evaluation framework that aims at eliminating both mammogram selection bias and limited subtlety variation, thus enabling a fair and complete comparison procedure. Furthermore, in order to provide a test bed for future comparisons, a dataset is constructed from all the available AD Regions Of Interest in Digital Database for Screening Mammography (163 AD vs 375 Regions Of Interest from specific normal cases) and is used to further evaluate the performance of the proposed method. The method performed flawlessly and classified correctly all cases.

A computer-aided diagnosis (CADx ) framework for the diagnosis of clustered microcalcifications (MCs) has already been developed, which is based on the analysis of MCs' morphologies,the shape of the cluster they form and the texture of the surrounding tissue. In this study, we investigate the diagnostic information that the relative location of the cluster inside the breast may provide. Breast probabilistic maps are generated and adopted in the CADx pipeline, expecting to empower its diagnostic procedure. We propose a flowchart combining alternative classification algorithms and the aforementioned probabilistic maps in order to provide a final risk for malignancy for new considered mammograms. For the evaluation performance, a large dataset of mammograms provided from the Digital Database of Screening Mammography (DDSM) has been used. The obtained results indicate that the proposed modifications lead to the enhancement of the diagnostic process, as the classification results are further improved. Additionally, a straightforward comparison between the CADx pipeline and the radiologists who assessed the same mammograms, reveal that the CADx pipeline performs toward the right direction, as the sensitivity remains at high levels, while improving both the accuracy, from 51.4% to 69%, and the specificity, from 16.6% to 54.7%.

3-D stereo reconstruction, a technique that estimates per-pixel depth in a scene, is still a challenging problem mainly due to some prohibitive factors that limit its performance and computational ability. The aim of this paper is to present a new hardware-efficient disparity map computation, which is based on disparity space image processing using discrete dynamic systems. The hardware architecture of the proposed system was implemented on a high-end field programmable gate array (FPGA) device, offering real-time 3-D reconstruction speeds using a hardware aware architecture based on parallelism and process pipelining. The proposed architecture fulfills the requirements of real-world applications regarding resource usage, frame rates, and disparity resolution, while its implementation on an Altera Stratix IV family FPGA device can extract disparity maps of up to 1280 × 1024 pixels with up to 128 disparity levels under real-time or near real-time conditions at a clock rate of 168 MHz. Qualitative and quantitative results also demonstrate its performance and improvement over previous hardware-related studies, making our approach a suitable candidate for applications in which timing and processing constraints are critical.

Breast microcalcifications are one of the most important mammographic findings related to the existence of the breast cancer. Radiologists usually characterize microcalcifications based on their morphologies, the distribution within the cluster they form, the shape of the cluster and its relative location inside the breast. In this study, we focus on the latter factor and we study its effect on the probability of malignancy. The main purpose of our study is to generate probabilistic breast cancer atlases for clusters of microcalcifications in order to visualize the influence of cluster location on cancer probability. We propose a framework for the generation of such atlases, including segmentation of important breast landmarks and projection of different clusters of microcalcifications on a reference breast shape. The generation of the atlases is implemented using mammograms from the Digital Database of Screening Mammography. The obtained probabilistic atlases reveal specific areas in the breast of higher occurrence of clusters and higher risk of malignancy.

In this work, we study the factors that influence the efficacy of a proposed Computer Aided Diagnosis (CADx) framework for the diagnosis of clustered microcalcifications (MCs) using a large dataset of mammograms containing cases of varying breast density and findings' subtlety. The reported results indicate that the proposed framework performs towards the right direction, as it appears high classification performance (Az=0.909) for specific subsets of cases, while outperforming at the same time the performance of the radiologists who evaluated the same cases. The effect of the initial enhancement of mammograms in the CADxpipeline is then investigated, by applying three different image enhancement techniques on several subsets of mammograms. We observed that for the considered subsets of dense mammograms, a wavelet-based enhancement algorithm outperformed the rest and provided superior classification performance (Az=0.849). We indicate therefore that the density of the breast determines the need of different computational algorithms for the analysis of a mammogram and as a result the a priori knowledge of this factor may be exploited for the optimization of the diagnostic process.

In this paper, a multiresolution approach is suggested for texture classification of atherosclerotic tissue from B-mode ultrasound. Four decomposition schemes, namely, the discrete wavelet transform, the stationary wavelet transform, wavelet packets (WP), and Gabor transform (GT), as well as several basis functions, were investigated in terms of their ability to discriminate between symptomatic and asymptomatic cases. The mean and standard deviation of the detail subimages produced for each decomposition scheme were used as texture features. Feature selection included 1) ranking the features in terms of their divergence values and 2) appropriately thresholding by a nonlinear correlation coefficient. The selected features were subsequently input into two classifiers using support vector machines (SVM) and probabilistic neural networks. WP analysis and the coiflet 1 produced the highest overall classification performance (90% for diastole and 75% for systole) using SVM. This might reflect WP's ability to reveal differences in different frequency bands, and therefore, characterize efficiently the atheromatous tissue. An interesting finding was that the dominant texture features exhibited horizontal directionality, suggesting that texture analysis may be affected by biomechanical factors (plaque strains).

In this paper, a hardware implementation of a fuzzy-modified ant colony processor that is suitable for image retrieval is presented for the first time. The proposed method utilizes three different descriptors in a two-stage fuzzy ant algorithm where the query image represents the nest and the database images represent the food. From the hardware point of view, only a small number of algorithms for hardware implementation have been reported in the image retrieval literature, since research focuses mainly on possible software solutions and the acceleration of existing algorithms. The proposed digital hardware structure is based on a sequence of pipeline stages, while parallel processing is also used in order to minimize computational times. It is capable of performing the extraction and comparison of features from (64$\,\times\,$64)-pixel-size color images, although through a simple transformation it can be easily expanded to accommodate images of larger sizes. The architecture of the processor is generic; the units that perform the fuzzy inference can be used with different descriptors than the ones proposed here and can be utilized for other fuzzy applications. It was designed, compiled, and simulated using the Quartus Programmable Logic Development System by the Altera Corporation. The fuzzy processor exhibits a level of inference performance of 800 K fuzzy logic inferences per second with 24 rules, and can be used for real-time applications where the need for short processing times is of the utmost importance.

This paper presents a new method suitable for the contrast enhancement of digital images. The proposed algorithm employs the center-surround architecture of the human visual system (HVS) for the comparison of the central pixel with the mean intensity of its surround. These two values are inputs to a mapping function, which calculates the new intensity value of the central pixel, in a way that increases small local intensity differences while preserving the large ones. The proposed method is applied to three spatial scales by employing different surround sizes. As a result, contrast enhancement is performed in different spatial frequencies of the original image. The proposed method is tested against other similar methods and is found to exhibit better results in the enhancement of low-contrast images, such as medical, aerial or foggy scenes, revealing visual information that otherwise would not be available to the human observer.

Image moments are used extensively in various applications. Depending on the task at hand, the type as well as the order of moments to be implemented varies greatly. Simple geometric characteristics require the first few geometrical moments, while robust image watermarking employs high-order orthogonal moments. In this paper, we present an architecture which can be easily modified to implement various types of moments, including geometric, continuous as well as discrete orthogonal, in real-time. Although this architecture is optimized for hardware, software implementation is feasible and advantageous in some cases.

The angular radial transform (ART) is the region-based shape descriptor for MPEG-7. It is a compact descriptor that presents invariance to simple transformations, while presenting very low information redundancy. However, limited effort has been spent on the fast computation of ART, which is of the essence, especially in systems with sparse computational resources. In this paper, we present an efficient implementation of the descriptor, relating it to accumulation moments. The approximations used are shown not to affect the final 140-bit descriptor, and the architecture is capable of processing images up to one order of magnitude faster than the conventional method.

In this paper, a two-dimensional Cellular Automaton (CA) model simulates the evacuation process of a crowd responding to fire spread. The crowd consists of individuals and its behaviour is modelled by the response of each individual to a rule that directs him/her to the nearest exit. Furthermore, fire spreading and movements of the crowd members while approaching the fire are successfully simulated. Empirical studies and socio-psychological concepts that attempt to explain how individuals act under fire threat have been considered. Characteristic features of crowd dynamics, such as incoherent pedestrian motion, blockings in front of exits and mass behaviour are successfully simulated. An efficient user-friendly interface has been equipped with parameters defining the arrangement of the area, crowd formation and fire features. Finally, the model is executed fast on typical PCs and can be used for planning evacuation strategies under fire threat or as part of a real-time decision support system.

Image moments have been extensively used as feature descriptors. However, the quantization error introduced in discrete signals presents problems, especially when dealing with small-size images. This results in the fact that the invariant properties of moments are compromised. In this paper, we present a technique suitable for the calculation of moments from a continuous signal, derived by piecewise polynomial interpolation of the corresponding discrete one. The computed moments exhibit significantly increased accuracy while requiring trivial computational effort. Zernike moments are then computed using the proposed scheme and are shown to display increased stability to geometrical transformations.

In this paper, a one dimensional (1-d) cellular automaton (CA) for pseudorandom number generation (PRNG) and its reconfigurable hardware implementation are presented. The proposed 1-d CA based on the real time clock sequence (analytical time description) can generate high-quality random numbers which can pass all of the statistical tests of DIEHARD and NIST which seem to be the most powerfully complete general test suites for randomness. After describing our implementation in field-programmable gate array (FPGA), through experiments, we have identified the efficiency of the presented CA that performs exceptionally well compared to most known CA PRNGs reported in literature. More specifically, our CA implementation outperforms all the previous CA and LFSR PRNGs both in hardware implementation and timing characteristics. Such a CA can be efficiently implemented for PRNG reasons in every real time clock application with finally no silicon overhead.

Chebyshev moments were recently proposed to address problems of moments of continuous orthogonal basis. They are moments of a discrete basis and are orthogonal in the domain of the image coordinate space, a feature which presents a number of advantages over moments of continuous orthogonal basis. No effort has been devoted yet to the efficiency of their calculation. In this paper, we present a hardware architecture which enables real-time processing of binary and grayscale images of up to 4 Mpixels. This architecture is realized in a field-programmable gate array of typical clock frequency of 100 MHz

Zernike moments have been proven to be very powerful image descriptors. However, their computational complexity makes them unsuitable for real-time applications. In this paper, a mathematical relationship between geometric and Zernike moments is extracted. In this way, the computation of geometric moments is adequate to derive Zernike moments. Since geometric moments can be efficiently implemented in hardware and their calculation can be performed in real-time, we propose here a new real-time hardware architecture for the computation of Zernike moments. This method outperforms existing software approaches, especially for large images, allowing real-time processing of images up to 4 Mpixels.

A hardware implementation of a fuzzy processor suitable for morphological color image processing applications is presented for the first time. From the hardware point of view, only a small number of algorithms for hardware implementation of soft gray-scale morphological filters have been reported in the literature, since research in mathematical morphology focuses mainly on possible extensions of the standard definitions (e.g., color and fuzzy mathematical morphology). The proposed digital hardware structure is based on a sequence of pipeline stages, and parallel processing is used in order to minimize computational times. It is capable of performing the basic morphological operations of standard and soft erosion/dilation for color images of 24-bit resolution. For the computation of morphological operations, a 3 /spl times/ 3-pixel image neighborhood and the corresponding structuring element are used. However, the system can be easily expanded to accommodate windows of larger sizes. The architecture of the processor is generic; the units that perform the fuzzy inference can be utilized for other fuzzy applications. It was designed, compiled, and simulated using the MAX+PLUS II Programmable Logic Development System by Altera Corporation. The fuzzy processor exhibits a level of inference performance of 601 KFLIPS with 54 rules, and can be used for real-time applications where the need for short processing times is of the utmost importance. The selection of a latest technology computer system (Pentium 4/3 GHz with SSE-2) can speed up image processing applications, but the time required still cannot be compared to the corresponding time using this hardware structure.

A new algorithm suitable for the computation of weighted order statistics operations is presented. The algorithm is based on the local histogram and a successive approximation technique. The successive approximation technique ensures that the result of the weighted order statistics operation is computed in a fixed number of steps, equal to the pixel value resolution. Comparative experimental results are also included. It is shown that the proposed algorithm is faster than the quick sort algorithm for weighted order statistics operations.