Fakhry M. Khellah

Texture Classification Using Dominant Neighborhood Structure

This paper proposes a new approach to extract global image features for the purpose of texture classification. The proposed texture features are obtained by generating an estimated global map representing the measured intensity similarity between any given image pixel and its surrounding neighbors within a certain window. The intensity similarity map is an average representation of the texture-image dominant neighborhood similarity. The estimated dominant neighborhood similarity is robust to noise and referred to as image dominant neighborhood structure. The global rotation-invariant features are then extracted from the generated image dominant neighborhood structure. Features obtained from the local binary patterns (LBPs) are then extracted in order to supply additional local texture features to the generated features from the dominant neighborhood structure. Both features complement each other. The experimental results on representative texture databases show that the proposed method is robust to noise and can achieve significant improvement in terms of the obtained classification accuracy in comparison to the LBP method. In addition, the method classification accuracy is comparable to the two recent LBP extensions: dominant LBP and completed LBP.

Statistical processing of large image sequences

The dynamic estimation of large-scale stochastic image sequences, as frequently encountered in remote sensing, is important in a variety of scientific applications. However, the size of such images makes conventional dynamic estimation methods, for example, the Kalman and related filters, impractical. We present an approach that emulates the Kalman filter, but with considerably reduced computational and storage requirements. Our approach is illustrated in the context of a 512 /spl times/ 512 image sequence of ocean surface temperature. The static estimation step, the primary contribution here, uses a mixture of stationary models to accurately mimic the effect of a nonstationary prior, simplifying both computational complexity and modeling. Our approach provides an efficient, stable, positive-definite model which is consistent with the given correlation structure. Thus, the methods of this paper may find application in modeling and single-frame estimation.

Detection of crashed aircraft in polarimetric imagery

This manuscript presents studies examining the use of C-band polarimetric Synthetic Aperture Radar (SAR) systems for the detection of crashed aircraft. The ultimate aim is to assist Search and Rescue in Canada in the location of such targets. Detection methodologies based on the Polarimetric Whitening Filter, Cameron Decomposition, and measures of even bounce contributions to the backscatter have been examined. Tests were performed using imagery of serviceable and crashed aircraft and crashed aircraft parts. Although individual methods make it possible to detect the crashed aircraft, best results for target detection with decreased numbers of false alarms occur when these methods are used in combination.

Evolution Impact on Architecture Stability in Open-Source Projects

Software systems usually evolve constantly, which requires constant development and maintenance. Subsequently, the architecture of these systems tends to degrade with time. Therefore, stability is a key measure for evaluating an architecture. Open-source software systems are becoming progressively vital these days. Since open-source software systems are usually developed in a different management style, the quality of their architectures needs to be studied. ISO/IEC SQuaRe quality standard characterized stability as one of the sub-characteristics of maintainability. Unstable software architecture could cause the software to require high maintenance cost and effort. In this work, the authors propose a simple, yet efficient, technique that is based on carefully aggregating the package level stability in order to measure the change in the architecture level stability as the architecture evolution happens. The proposed method can be used to further study the cause behind the positive or negative architecture stability changes.

Large scale dynamic estimation of ocean surface temperature

This paper addresses the dynamic estimation of the ocean surface temperature based on data from the Along-Track Scanning Radiometer (ATSR) for large (512/spl times/512) fields. For such huge problems, the conventional dynamic estimation tool (the Kalman filter) is not directly applicable, instead, we develop a recursive estimation algorithm that emulates the Kalman filter. Our approach uses a multiscale estimation algorithm for the update step, and makes simplifying assumptions about the surface dynamics leading to a computationally efficient prediction step.

Synthetic aperture radar for search and rescue: evaluation of advanced capabilities in preparation for RADARSAT-2

Researchers at the Canada Centre for Remote Sensing of Natural Resources Canada are exploring the use of remotely sensed imagery to assist Search and Rescue in Canada. Studies have been examining the use of Synthetic Aperture Radar for the detection of crashed aircraft. Promising results have been obtained with techniques for detection of dihedrals in interferometric and polarimetric data. With further development in technologies and techniques, and improved coverage of the Canadian landmass by future spaceborne systems such as RADARSAT-2, it is expected that it will be possible to assist in Search and Rescue for land targets.

Architectural Stability Evolution in Open-Source Systems

Open-source software systems are becoming progressively vital these days. Since open-source softwares are usually developed in a different management style, the quality of their architectures needs to be studied. ISO/IEC SQuaRe quality standard characterized stability as one of the sub-characteristics of maintainability. Unstable software architecture could cause the software to require high maintenance cost and effort. Almost all stability related studies target the package level. To our knowledge, there has been no proposed work in literature that addresses the stability at the system architecture level. In this work, we propose a simple, yet efficient, technique that is based on carefully aggregating the package level stability in order to measure the change in the architecture level stability as the architecture evolution happens. The proposed method can be used to further study the cause behind the positive or negative architecture stability changes.

Application of Local Binary Pattern to Windowed Nonlocal Means Image Denoising

This paper presents a new technique to image denoising that mainly addresses the incurred high blurring when the windowed nonlocal means is applied to images corrupted by high noise levels. The proposed method is based on an enhanced weighting function that computes patches similarity based on both their intensities and structural features. The structural features are encoded using Local Binary Pattern (LBP) a well known texture descriptors. A new LBP based weighting function is proposed that has properties complementing the intensity based weighting function. The LBP based weighting function is used to modulate the intensity based weighting function. The modulated weights are noise independent and reflect the actual patch similarity. The method is found to be quantitatively and qualitatively effective in denoising images when corrupted by high noise levels. It suppresses image noise while preserving significant image characteristics.

Motion estimation of sparse, remotely-sensed fields

Sea surface temperature (SST) can be estimated from remotely-sensed images. Because of the sparsity of the available observation it is ideal to do estimation using dynamic methods (such as Kalman filtering). To model dynamics of SST accurately we need to know the motion of sea current. The traditional video motion estimation problem is straightforward, in some ways, because there are so few constraints. That is, the motion vectors are pretty much arbitrary, and successive image frames are densely pixellated, have the same number of pixels, with similar noise statistics. However there are many motion estimation problems, particularly in the area of remote sensing, which do not share these properties. In this paper we investigate the problem of determining the motion field of the sea surface, based on infrared measurements of surface temperature. This problem is challenging in that only a subset of the whole domain is measured at each point in time; specifically, only a few stripes are imaged. In addition, because of clouds, the measured subset varies from time to time. The quality (level of noise) can also vary from pixel to pixel. Our research is based on the following assumptions and observations: the motion field should be smooth and ideally divergence-free, i.e. the motion field is close to time-stationary. Based on these assumptions we choose to use optical flow method for this motion problem. We handle difficulty of data sparcity by pre-estimation to get a dense field. Pre-estimation can be refined by integrating this motion estimation result.

Data fusion of sea-surface temperature data

The problem of data fusion, the merging of data taken by different sensors, becomes ever more relevant with the launch of each new remote-sensing platform. One example of considerable current interest in the climate change community is the production of an improved sea surface temperature (SST) map. In particular, two state-of-the-art instruments - the Along Track Scanning Radiometer (ATSR) and AVHRR - share complementary features: ATSR allows excellent cloud discrimination and atmospheric correction based on its dual-view scanning geometry, but observes only narrow swaths of ocean; AVHRR suffers from low-wavenumber atmospheric distortions and cloud contamination, but has extensive global coverage. The authors propose a methodology for combining ocean skin temperatures from the ATSR and AVHRR instruments to produce a continuous analysis at one-sixth degree spatial and three-day temporal resolutions, together with reliable error estimates, from the fusion of multiple datasets with arbitrary sampling characteristics, resulting in estimated temperatures which unite the precision of ATSR with the superior coverage afforded by AVHRR.