Browsing by Author "Daamouche, A.(Supervisor)"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Classification of multispectral image using SVM and gabor filter
    (Université M’Hamed BOUGARA de Boumerdes : Institut de génie electrique et electronique (IGEE), 2018) Amara, Kheire Eddine; Blil, Mohamed Amine; Daamouche, A.(Supervisor)
    The present work deals with image classification in the field of remote sensing and retinal blood vessels which is a joint venture between image processing and classification techniques. The advancement in the imagery field in terms of resolution allowed the availability of very high resolution images. Therefore, new techniques and algorithms become necessary to cope with technology. In this regard, Gabor filters have been used in a variety of image processing applications. It had a distinctive effect in improving the image and increasing its clarity and quality for further processing. In this work, the effectiveness of the Gabor filter is explored. In particular, we used Gabor features in conjunction with the SVM to classify remotely sensed data and retinal blood vessels images. The simulation results on different datasets showed that our approach is promising in the field of image classification.
  • No Thumbnail Available
    Item
    Spectral-spatial features for hyperspectral image classification
    (2018) Mounir, Zakaria; Merouani, Mawloud; Daamouche, A.(Supervisor)
    Image classification is one among important branches of artificial intelligence field. Generally, it translates the information contained in images into thematic categories which are suitable for use in many applications using low-level visual features. Nowadays, there exists a large number of machine learning algorithms used for image classification. The main objective of this work is to perform a classification of hyperspectral data by means of spectral-spatial features. The principle component analysis was exploited as a tool to decorrelate and reduce the dimension of the original hyperspectral data. The mathematical morphology is used to extract the spatial features; its parameters were generated empirically. The combination of the morphological features and the spectral features were fed to the state-of-the-art classifier which is the Support Vector Machines (SVM). The obtained results over two benchmark datasets show that the achieved performance using the developed method is promising.