New Digital Images Magnification Algorithm Based on Integration of Mapping and Synthesis Concept

Abu Amro, Rawia M.A. (2016) New Digital Images Magnification Algorithm Based on Integration of Mapping and Synthesis Concept. Masters thesis, Universiti Sains Malaysia.

[img] PDF - Submitted Version
Download (315Kb)

    Abstract

    Pembesaran imej adalah proses pembinaan semula imej resolusi tinggi (HR) dari versi resolusi rendah (LR). Proses pembesaran imej adalah salah satu proses penting yang digunakan untuk memenuhi keperluan manusia. Proses ini digunakan dalam beberapa aplikasi seperti dalam pengimejan perubatan, penderiaan jauh, mempertingkatkan butiran imej dan percetakan. Pada umumnya, algoritma pembesaran yang biasa menggunakan konsep penentudalaman. Walau bagaimanapun, algoritma pembesaran berasaskan penentudalaman ini mengalami masalah seperti kehadiran artifak-artifak yang tidak diingini dalam imej yang diperbesarkan seperti pinggir terhalang dan pinggir kabur. Artifak-artifak ini kebanyakannya muncul pada pinggir yang jelas. Oleh itu, selain menggunakan konsep penentudalaman, kajian ini memberi fokus kepada memperkenalkan algoritmapembesaran yang baharuberasaskan konsep sintesis. Disebabkan oleh konsep sintesis telah digunakan dalam algoritma sintesis tekstur berasaskan tampalan, pengubahsuaian kepada algoritma sintesis tekstur barasaskan tampalan perlu dilakukan agar dapatdigunakanuntuktujuan pembesaran imej. Pengubahsuaian yang dicadangkan menghasilkan algoritma pembesaran baharu yang dipanggil Algoritma Pembesaran Barasaskan Pemetaan (MBMA). Algoritma MBMA menggantikan setiap piksel imej LR dengan blok HRdua dimensi untuk membina imej HR. Algoritma yang dicadangkan pada asasnya direka untuk memelihara pinggir yangjelas. Dua variasi cadangan MBMA diperkenalkan, iaitu MBMA_Average dan MBMA_Direct. Variasi MBMA yang dicadangkan telah dibandingkan dengan teknologiterkini pembesaran algoritma lain menggunakan 100 imej piawai dan 200 imej plat kereta lesen Malaysia (MLCP). MBMA_Average menghasilkan imej pembesaran yang lebih baik dengan pengurangan artifak yang tidak diingini (iaitu pengurangan pinggir kabur dan pinggir terhalang) berbanding dengan teknologi algoritma yang lain. Seterusnya, analisis kuantitatif menunjukkan bahawa MBMA_Average yang dicadangkan juga menghasilkan nilai yang terbaik dalam pengukuran PSNR, SSIM, MSE dan FSIM berbanding algoritma-algoritma tersebut. ________________________________________________________________________________________________________________________ Image magnification is the process of reconstructing High Resolution (HR) image from its Low Resolution (LR) version. Image magnification process is one of the most important processes that is used to fulfill human needs. This process is used in several applications such as in medical imaging, remote sensing, enhancing image details and printing. In general, the common magnification algorithms employ interpolation concept. However, these interpolation-based magnification algorithms suffer from the appearance of undesirable artifacts in magnified images such as edge blocking and edge blurring. These artifacts mostly appear around the strong edges.Therefore, instead of employing interpolation concept, this study focuses in introducing new magnification algorithm based on synthesis concept. As the synthesis concept has been used in patch based texture synthesis algorithms, a modification to the patch based texture synthesis algorithms has to be carried out in order to use it for the image magnification purpose. The proposed modification produces a new magnification algorithm called the Mapping Based Magnification Algorithm (MBMA). The proposed MBMA replaces each pixel in the LR image with a two dimensional HR block to reconstruct the HR image. The proposed algorithm is basically designed to preserve the strong edges. Two variants of the proposed MBMA are introduced, namely MBMA_Average and MBMA_Direct.The proposed MBMA variants have been compared with other state-of-the-art magnification algorithms by using 100 standard images and 200 Malaysian License Car Plate (MLCP) images. The proposed MBMA_Average produces the best magnified images with less undesirable artifacts (i.e. less of edge blurring and edge blocking) compared with the other state-of-the-art algorithms. Furthermore, the quantitative analyses show that the proposed MBMA_Average also produces the best value of the PSNR, MSE, SSIM and FSIM measurements compared to those algorithms.

    Item Type: Thesis (Masters)
    Additional Information: full text is available at rplus.eng.usm.my:8080/ir_plus/institutionalPublicationPublicView.action?institutionalItemId=2073
    Subjects: T Technology
    T Technology > TK Electrical Engineering. Electronics. Nuclear Engineering > TK7800-8360 Electronics
    Divisions: Kampus Kejuruteraan (Engineering Campus) > Pusat Pengajian Kejuruteraaan Elektrik & Elektronik (School of Electrical & Electronic Engineering) > Thesis
    Depositing User: Mr Mohd Jasnizam Mohd Salleh
    Date Deposited: 12 Jun 2018 10:19
    Last Modified: 12 Jun 2018 10:19
    URI: http://eprints.usm.my/id/eprint/40753

    Actions (login required)

    View Item
    Share