5th International Conference on Signal and Image Processing (SIPRO 2019)

July 13~14, 2019, Toronto, Canada

Accepted Papers


    Reconstruction With High Resolution Sar Tomography VIA Compressive Sensing
    Ishak Daoud1, Assia Kourgli2, Aichouche Belhadj Aissa2.
    1Telecommunications and information processing, Faculty of Electronics and Computer Science, Laboratory of Image processing and radiation, University of Science and Technology Houari Boumediene, Algiers.
    2Laboratory of Image processing and radiation, University of Science and Technology Houari Boumediene, Algiers, Algeria
    ABSTRACT
    The SAR tomography, is anapproach that uses multi-pass SAR images to decompose the target basing on its backscatter mechanisms, this decomposition helps to generate the reflectivity profile on the elevation axis. However, the common Rayleigh resolution related to the Nyquist condition, can cause quality problems in elevation, due to the low number of acquisitions, the non-regular distribution of the baseline and its small aperture in orthogonal baseline. The work we have presented in this paper, concerns the reconstruction of the reflectivity profile of TerraSAR-x radar target images by exploiting the concept of CS ‘Compression Sensing’, assuming that the target has generally a spars representation along the elevation direction. We have presented also, a reconstruction for some simulated profiles based on the radar characteristics of TERRASR-X, using some algorithms asBasis Pursuit‘BP’ and Basis Pursuit Denoised ‘BPDN’, to well Simulated a real reconstruction when the measurements are noised. Based on the results obtained by the convex reconstruction algorithmsimplemented on MATLAB, we have shown how the number of measures necessary for reconstruction can be reduced and how reconstructed samples can be increased, which can bring Better resolution in elevation for anoisy measurement, based on a lemma that binds the sparsity, the total reconstruct samples and the number of measures.
    KEYWORDS

    SAR Tomography Compressive sensing, sparsity, L1&L0 norm-minimization, SAR tomography, Restricted Isometry Property.

      Particle Visualization Systemn Based on the Scattering Of Light Produced By a Slide Beam Laser In a Clean Room Using Image Processing
      O. Juina 1, SC Hu2 and T.Lin3
      1Department of Mechanical and Automation Engineering, National Taipei University of Technology, Taipei, 10608 Taiwan,R.O.C
      2,3Department of Energy and Refrigerating Air-Conditioning Engineering, National Taipei University of Technology, Taipei, 10608 Taiwan, R.O.C
      ABSTRACT
      In the field of clean room systems, the need to have high standards of cleaning and environmental control has generated the creation of new equipment which can solve the different problems of monitoring in particle filtration. The system proposed below has been developed based on new technologies like the evolution of camera sensors, the use of a beam laser to visualize particles, and the link between programming algorithms with free platforms. The first system used consisted of a Canon 650D camera with a 17-55mm lens, a Ld Pumped All-SolidState Green Laser. Tests were performed inside a controlled environment where the external light insulation was removed, a Transparent FOUP (Front Opening Unified Pod) where the sample of white marble dust was introduced to see its dispersion among the particles. At that moment, the photos were taken at different angles of incidence such as against the laser and 60 degrees with reference to the line of action of the laser. Furthermore, other tests were carried out in an external environment, where photographs of particles from the human body were taken. In order to implement a comparison of results, we used a Second System composed by a Camera HighResolution CMOS Sensor with Global Shutter lt225. The Ld Pumped All-Solid-State Green Laser, using the same parameters as the first system. Cases were repeated with the transparent FOUP and large particles from the body. Simulating the normal movements inside a clean room. The third stage, is the image processing using OpenCV libraries, in this case, EmguCV which processes images, the fundamental principle of the image processing is the reading of each pixel, the intensities of each pixel and in the case of the processing of black and white images, each pixel receives values from 0 to 255, with 0 being the value for black and 255 for white. The program algorithm responds to these values and will separate the high-intensity values from the low-intensity values. In this case, the green color will become an important value, which by means of mathematical filters, will generate a clearer image of where the particles are. The main features of the images taken by the camera Canon 650D are its resolution, which at short distances allows displaying small particles in the order of 12um but, its limitation is its speed of capture. On the other hand, the processed results of the photographs taken by the CMOS camera showed greater accuracy in frame per second, but a large amount of storage memory is required due to the capture speed.
      KEYWORDS

      Particle Visualization System, FOUP (Front Opening Unified Pod), CMOS Sensor, OpenCV.

        Blind Image Quality Assessment Using Singular Value Decomposition Based Dominant Eigenvectors For Feature Selection
        Besma Sadou, Atidel Lahoulou and Toufik Bouden, University of Jijel, Algeria
        ABSTRACT
        In this paper, a new no-reference image quality assessment (NR-IQA) metric for grey images is proposed using LIVE II image database. The features used are extracted from three well-known NR-IQA objective metrics based on natural scene statistical attributes from three different domains. These metrics may contain redundant, noisy or less informative features which affect the quality score prediction. In order to overcome this drawback, the first step of our work consists in selecting the most relevant image quality features by using Singular Value Decomposition (SVD) based dominant eigenvectors. The second step is performed by employing Relevance Vector Machine (RVM) to learn the mapping between the previously selected features and human opinion scores. Simulations demonstrate that the proposed metric performs very well in terms of correlation and monotonicity.
        KEYWORDS

        Natural Scene Statistics (NSS), Singular Value Decomposition (SVD), dominant eigenvectors, Relevance Vector Machine (RVM).