Background subtraction is a major part of many motion detection, tracking and surveillance systems. In this paper a new algorithm for the purpose of the background model initialization has been presented. The key idea of the proposed method lies in the identification of the background based on QRDecomposition method in linear algebra. R-values produced with QR-Decomposition can be applied to decompose a given system to indicate the degree of the significance of the decomposed parts. We split the image into small blocks and select the background blocks with the weakest contribution, according to the assigned R-values. The main advantage of the proposed method is that in contrast to many other methods, here, there is no need for an empty scene with no foreground object. Simulation results showed that the proposed method produced better background model with respect to some others.

In this paper, we propose a new method for fast and effective image segmentation using snakes (active contours). This method incorporates the active contour models with the wavelet transform features using directional edges in high and low frequency content sub-images. The proposed method takes advantages of these features to improve the boundary attraction process in active contour models. It also enhances the speed of image segmentation process by using a multi-scale approach. The analytical results show the efficiency of the proposed method in extraction of objects in images while performing the image segmentation at a higher speed compared to previously proposed active contour models.

This paper presents a Fuzzy Hybrid Learning Algorithm (FHLA) for the Radial Basis Function Neural Network (RBFNN) with application in human face recognition. The method determines and initializes the number of hidden neurons and their characteristics in the RBFNN structure by using of cluster validity indices with majority rule and advanced fuzzy clustering respectively. The FHLA combines the gradient method and the linear least squared method for adjusting the RBF parameters and connection weights. The designed RBFNN with the FHLA is used as a classifier in a face recognition system, which its feature vectors, obtained by combining shape information and Principal Component Analysis (PCA). The efficiency of the proposed method is demonstrated on the ORL and Yale face databases.

Determining the capacity of image for data hiding is a complex task. This capacity may be influenced by many factors such as the content of image and also Human Visual System (HVS). A few works on this topic are reported. In this paper we propose a new method for calculating capacity in gray scale images. Our method is based on measuring Smoothness and Connectivity of neighbors of a specific pixel in image bitplanes. This method can produce a signed value for each pixel capacity such that the sign shows the modification in pixel value must be increasing or decreasing. Comparison with the existing methods showed that the proposed method gave better estimation for capacity of data hiding in an image

A hybrid genetic-neural (GA-ANN) model was designed to differentiate malignant from benign in a group of patients with histopathologically proved breast lesions on the base of BI-RADS descriptors and data derived independently from time-intensity curve. We used a database with 117 patients' records each of which consisted of 27 quantitative parameters mostly derived from time-intensity curve, 4 BI-RADS qualitative data which determined by expert radiologist and patient age. These findings were encoded as features for a genetic algorithm (GA) as a preprocessor for feature selection and classified with a three-layered neural network to predict the outcome of biopsy. The network was trained and tested using the jackknife method and its performance was then compared to that of the experienced radiologist in terms of sensitivity, specificity, accuracy and receiver operating characteristic curve (ROC) analysis. The network was able to classify correctly 107 of 117 original cases and yielded a good diagnostic accuracy (91%), sensitivity (95%) and specificity (78%) compared to that of the radiologist (92%), (96%) and (78%).

Unsupervised clustering is a powerful technique that can be used for distinguishing the real target from the false targets such as flares in the images formed by infrared sensors in missiles. The rosette scan image is formed by a single infrared sensor scanning the total field of view by rosette pattern. This image is a greyscale image in which the real target could be differentiated from flares regarding the intensity of radiation sensed by IR sensor or the spatial position of the target. The ISODATA is one of the unsupervised clustering algorithms that could be used in infrared guided missiles detectors, since the algorithm itself computes the number of clusters or in other words the number of flares. Although the only drawback of the ISODATA is the extension in the processing time of the algorithm while the missile approaches the target and the number of detected clusters varies frequently, we can still take advantage of the algorithm by speeding up the most time consuming parts. In our approach to identify and locate the time consuming parts of the algorithm, first a profiling on a software implementation of the ISODATA algorithm has been carried out. The results show that over 60 percent of the complete execution time of the algorithm is consumed in computation of the distance from cluster centres. In this paper we propose a pipelined and parallelized datapath for hardware implementation of the algorithm in order to speed up the distance computation process and overcome the problem

The aim of this paper is presentation of a vision based approach to tracking of human motion for the study of athletic performance. The proposed method is developed to capture markers positions associated with human motion obtained from video data. A datadriven predictor is used for tracking of wearing reflective markers. The uncertainty and occlusion of detected markers increase the noise in captured positions. So, a memory based data-driven system is proposed for solving of occlusion problems. The proposed system contains maximum a posteriori (MAP) estimator is jointed into loop predictor for reducing of noisy positions. The convergence behavior and tracking capability of proposed algorithm are proved. The result of tracking algorithm is tested over 12 sport men in 6 velocities over more than 30000 captured frames. Simulation results validate the analysis and ensuing method.

Pumpkin are traditionally dried with solar dryers or hot air dryers ,which decrease product quality .To improve the quality of dehydrated pumpkins ,the technique of osmotic dehydration followed by hot air drying has been tested in recent years.The objectives of this study were to investigate the effect of osmotic dehydration as a pretreatment before hot air drying. A computer vision system(CVS) apply to study the color changes during drying.Pumpkin cubes were soaked in 50 %w/w sorbitol and sucrose solutions at 50c for up to 6 hours followed by hot air drying at 60 c and air velocity 1 m/s .A CVS was used for color evaluation during which the color image were converted to L/a/b values .Parameters related to shape (area, perimeter, energy) decreased during drying time. Parameters related to the texture of the image and calculated from the color co-ordinates represented well the complexity and non-homogeneity of the visual appearance of samples. The color changes of dried samples were evaluated with total color change (ΔE*),which was fitted to 0 1( ) 2 ΔE* = a + a t a .The “b” and L values were decreased,but the a values were increase in pumpkin samples during hybrid osmotic –air drying respectively.Hot air dried samples pretreated sorbitol had a higher L values compared to the hot air along drying while that pretreated with sucrose exhibited lower L values.

In many steganographic systems LSB filliping is used to hide messages in spatial domain. This method, with ease of implementation as its advantage, is a weak approach against statistical attacks. Methods that can avoid statistical attacks offer a small payload capacity. It has been shown that ±1 embedding is much harder to detect than LSB filliping. The ±1 embedding method does not introduce the same asymmetries into the stego image histogram. In this paper we propose a method that increases the capacity of ±1 embedding while maintains the same security level.

A considerable number of papers have been published in the last decade about biometric recognition using palm-print features. One of the most important stages in these methods is pre-processing which contains some operations such as filtering, Region Of Interest (ROI) extraction, normalization, etc. This paper proposes a precise method for extracting ROI of the palm-print. The basis of this method is geometrical calculations and Euclidean distance. Although preprocessing has been discussed in different papers incidentally, but no definite precise approach has been introduced. We show that our approach will extract the ROI with the accuracy of 99.7%

White matter tractogarphy is a non-invasive method for reconstructing three dimensional fibre pathways of the brain. Several fibre tracking algorithms have been proposed for this purpose. To evaluate and compare these algorithms, it is required to use synthetic datasets for which the simulated pathways are known to the user. This paper describes an algorithm designed in Matlab to simulate a diffusion tensor digital phantom for evaluating white matter fibre tractography algorithms and assessing their ability to detect fibre crossing. This digital phantom allows quantitative assessment of the robustness of fibre tracking algorithms by varying the thickness, the angles between crossing fibres, the Fractional Anisotropy ( FA) value of synthetic paths, and background.