We present a method based on deep learning for detecting and localizing abnormal/extreme events in sea surface temperature (SST) of the Red Sea images using training samples of normal events only. The method operates in two stages; the first one involves features extraction from each patch of the SST input image using the first two convolutional layers extracted from a pretrained convolutional neural network. In the second stage, two methods are used for training the model from the normal training data. The first method uses one-class support vector machine (1-SVM) classifier that allows a fast and robust abnormal detection in the presence of outliers in the training dataset. In the second method, a Gaussian model is defined on the Mahalanobis distances between all normal training data. Experimental tests are conducted on satellite-derived SST data of the Red Sea spanning for a period of 31 years (1985–2015). Our results suggest that the Gaussian model of Mahalanobis distances outperformed 1-SVM by providing better performance in terms of sensitivity and specificity.
In this communication, we propose an original temporal imaging concept for accurate spatio-temporal localization of scintillation events within a monolithic scintillator and a digital Si-PM matrix. Jointly analyzing the light distribution and the arrival time distribution of the first detected photons, it was possible to better recognize a photoelectric event and to accurately localize it in space, time and energy.
KEYWORDS: High dynamic range imaging, Image fusion, Cameras, Sensors, Video, Image processing, Image quality, High dynamic range image sensors, Imaging systems, Algorithm development
Nowadays, the high dynamic range (HDR) imaging represents the subject of the most researches. The major problem lies in the implementation of the best algorithm to acquire the best video quality. In fact, the major constraint is to conceive an optimal fusion which must meet the rapid movement of video frames. The implemented merging algorithms were not quick enough to reconstitute the HDR video. In this paper, we detail each of the previous existing works before detailing our algorithm and presenting results from the acquired HDR images, tone mapped with various techniques. Our proposed algorithm guarantees a more enhanced and faster solution compared to the existing ones. In fact, it has the ability to calculate the saturation matrix related to the saturation rate of the neighboring pixels. The computed coefficients are affected respectively to each picture from the tested ones. This analysis provides faster and efficient results in terms of quality and brightness. The originality of our work remains on its processing method including the pixels saturation in the totality of the captured pictures and their combination in order to obtain the best pictures illustrating all the possible details. These parameters are computed for each zone depending on the contrast and the luminosity of the current pixel and its neighboring. The final HDR image’s coefficients are calculated dynamically ensuring the best image quality equilibrating the brightness and contrast values and making the perfect final image.
The tone mapping field represents a challenge for all the HDR researchers. Indeed, this field is very important since, it offers better display terms for the end-user. This paper details a design of a recent tone mapping operator used in high dynamic range imaging systems. The proposed operator represents a local method that uses an adaptable factor which combines both the average neighbouring contrast and the brightness difference. Thanks to that, this solution provides good results with better brightness, contrast, and visibility and without producing neither undesired artifacts nor shadow effects.
The problem of human pose estimation in still images is considered. Most previous works predicted the pose directly with either local deformable models or a global mixture representation in the pose space. We argue that this process of pose estimation can be divided into different stages. We propose a new two-stage framework for human pose estimation. In the pre-estimation stage, there are three steps: upper body detection, model category estimation for the upper body, and full model selection for pose estimation. A new method based on pairwise scores of the upper body is proposed for upper body detection. In the estimation stage, we address the problem of a variety of human poses and activities. The upper body-based multiple mixture parts (MMP) model is proposed. This model not only joins different mixture models together, but can also analyze activities with complex kinematic structures. The model is compared with the state-of-the-art. The experimental results demonstrate the effectiveness of the MMP model.
High Dynamic Range (HDR) imaging has been the subject of significant researches over the past years, the goal of acquiring the best cinema-quality HDR images of fast-moving scenes using an efficient merging algorithm has not yet been achieved. In fact, through the years, many efficient algorithms have been implemented and developed. However, they are not yet efficient since they don't treat all the situations and they have not enough speed to ensure fast HDR image reconstitution. In this paper, we will present a full comparative analyze and study of the available fusion algorithms. Also, we will implement our personal algorithm which can be more optimized and faster than the existed ones. We will also present our investigated algorithm that has the advantage to be more optimized than the existing ones. This merging algorithm is related to our hardware solution allowing us to obtain four pictures with different exposures.
Although High Dynamic Range (HDR) imaging has represented, in the recent years, the topic of important researches, it
has not reached yet an excellent level of the HDR scenes acquisition using the available components. Indeed, many solutions
have been proposed ranging from bracketing to the beamsplitter but none of these solutions is really consistent with
the moving scenes representing light’s level difference.
In this paper, we present an optical architecture, which exploits the stereoscopic cameras, ensuring the simultaneous
capture of four different exposures of the same image on four sensors with efficient use of the available light.
We also present a short description of the implemented fusion algorithm implemented.
Region covariance descriptor has been employed in many computer vision applications such as texture classification, object detection and object tracking. It provides a natural way of fusing multiple features based on a set of pixels of a given region. However, the discriminative capacity of covariance descriptor can vary dramatically regarding different combination of feature sets that are fused and thus gives rise to the problem of discriminative feature selection when given a specific application. In this work, we propose a PCA-based feature selection approach in the construction procedure of the covariance descriptor. We show that covariance descriptor computed in a minutely-learned subspace can be adaptive to a specific target and thus results in a more compact and potentially more discriminative descriptor. Comparing experiments on real world video sequences demonstrate superior representational ability of the proposed method with respect to the conventional region covariance descriptor.
The authors propose a fragment-based variational filtering technique for human tracking. Based on human classifiers and histograms of oriented gradients descriptor, more informative local parts of the human body are selected in the reference model and updated during the tracking process. Hyper-parameters of the variational Bayesian filter are adaptively tuned in order to cope with variable scenes and occlusions. To speed up the initialization and reference updating, an efficient motion cue is fused with the human detection. Extensive experimental results on benchmark datasets show that the proposed tracker is effective and robust.
This paper describes about real-time moving pedestrian object detection and tracking. In this work, a new pedestrian object tracking approach is proposed, which uses a new texture model for better representing the pedestrian object. This approach based primarily on a step of object detection for automatic pedestrian object area extraction, features extraction step for pedestrian object representation and finally tracking step based on region matching method. We propose a new texture model for discriminant pedestrian object representation. This model based on a set of Haralick indexes computed from co-occurrence matrix. The presented approach can handle orientation, object scale, illumination changes, and clutter and complex occlusions.
KEYWORDS: Image segmentation, Signal to noise ratio, Matrices, Monte Carlo methods, Sensors, Computer simulations, Data modeling, Image classification, Interference (communication), Expectation maximization algorithms
We consider the problem of the blind separation of noisy instantaneously mixed images. The images are modeled by hidden Markov fields with unknown parameters. Given the observed images, we give a Bayesian formulation and we propose a fast version of the MCMC (Monte Carlo Markov Chain) algorithm based on the Bartlett decomposition for the resulting data augmentation problem. We separate the unknown variables into two categories: 1. The parameters of interest which are the mixing matrix, the noise covariance and the parameters of the sources distributions. 2. The hidden variables which are the unobserved sources and the unobserved pixel segmentation labels. The proposed algorithm provides, in the stationary regime, samples drawn from the posterior distributions of all the variables involved in the problem leading to great flexibility in the cost function choice. Finally, we show the results for both synthetic and real data to illustrate the feasibility of the proposed solution.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.