在過(guò)去幾年中,神經(jīng)生物學(xué)中的顯著性檢測(cè)成為一項(xiàng)重要研究。在處理圖像時(shí),由于存在一些人類興趣區(qū)域（ROI）等重要數(shù)據(jù),人們使用不同注意力水平感知圖像信息。視覺顯著性區(qū)域檢測(cè)是一種實(shí)現(xiàn)圖像ROI的有效方式。顯著性被定義為“一種專注于圖像中最有價(jià)值部分的注意機(jī)制。一些認(rèn)知和交互系統(tǒng)用來(lái)模擬顯著性模型。盡管目前存在各種用于顯著性檢測(cè)的先進(jìn)算法,但相對(duì)于無(wú)限制和復(fù)雜場(chǎng)景中時(shí)間成本計(jì)算和顯著對(duì)象分割,在性能改進(jìn)方面仍具有挑戰(zhàn)性。本文重點(diǎn)研究了圖像中視覺顯著性區(qū)域的檢測(cè)與融合方法。顯著性檢測(cè)在圖像視頻壓縮、目標(biāo)識(shí)別、圖像編輯、圖像縮略圖創(chuàng)建,圖片拼貼,圖像重定向和圖像檢索等不同圖像處理問(wèn)題中有廣泛的應(yīng)用。本文主要基于視覺顯著性檢測(cè),在自然圖像中使用自底向上的方法進(jìn)行重要顯著物體的檢測(cè)和分割。在本文中,我們研究了三種新型的自底向上的顯著性檢測(cè)方法,以解決目前顯著性檢測(cè)算法存在的問(wèn)題。首先,在第一章中我們對(duì)顯著性檢測(cè)算法在不同圖像處理問(wèn)題中的應(yīng)用進(jìn)行了概述。第二章簡(jiǎn)要介紹了現(xiàn)有的自底向上的視覺顯著性檢測(cè)方法。接下來(lái),我們提出了兩種創(chuàng)新的自底向上的顯著性檢測(cè)算法以更好地估計(jì)突出目標(biāo),以及一種新的圖像顯著區(qū)... 

【文章來(lái)源】：大連理工大學(xué)遼寧省 211工程院校 985工程院校 教育部直屬院校

【文章頁(yè)數(shù)】：105 頁(yè)

【學(xué)位級(jí)別】：博士

【文章目錄】：
Abstract
摘要
1 General Introduction
    1.1 Background
    1.2 Application of Saliency Detection Algorithms in Different Image ProcessingProblems
        1.2.1 Saliency Based Facial Features Detection
        1.2.2 Saliency Based Image and Video Segmentation
        1.2.3 Saliency Based Image Cropping
        1.2.4 Saliency Based Image and Video Compression
    1.3 Scope
    1.4 Contributions
    1.5 Dissertation Organization
2 Literature Review, Motivations and Evaluation Matrices
    2.1 Introduction
    2.2 Overview of Previous Salient Region Detection Methods
    2.3 Addressed Problems and Motivations
    2.4 Evaluation Matrices and datasets
        2.4.1 Datasets
        2.4.2 Evaluation Matrices
    2.5 Conclusion
3 CFBF-SRD:Color Frequency Features and Bayesian Framework Based Salient RegionDetection
    3.1 Introduction
    3.2 Formulation and Related Work of CFBF-SRD
        3.2.1 Log-Gabor Filter
        3.2.2 Bayesian Framework for Saliency
    3.3 Main Steps of CFBF-SRD
    3.4 Experimental Classification Results and Analysis
        3.4.1 Dataset and Parameter Settings
        3.4.2 Graphical Representation
        3.4.3 Computational Time Cost
        3.4.4 Segmentation by Adaptive Thresholding
        3.4.5 Failure Cases
    3.5 Discussion
    3.6 Conclusion
4 SAMM-SRD:Surroundedness and Absorption Markov Model Based Salient RegionDetection
    4.1 Introduction
    4.2 Related Work
        4.2.1 Surroundedness based Eye Fixation Prediction
        4.2.2 The Absorption Markov Chain: Review
    4.3 Proposed SAMM-SRD Algorithm
        4.3.1 Eye Fixation Prediction
        4.3.2 Graph Model Construction
        4.3.3 Construct Transfer Matrix
        4.3.4 Detect initial Saliency Map S_1
        4.3.5 Detect Initial Saliency Map S_2
        4.3.6 Fusion
        4.3.7 Smoothing
    4.4 Experiments
        4.4.1 Evaluation of Experimental Results
        4.4.2 Computational Time Cost
        4.4.3 Adaptive Thresholding based Segmentation
        4.4.4 Comparison
    4.5 Conclusion
5 DSET-SRF: DS-Evidence Theory Based Salient Regions Fusion
    5.1 Introduction
    5.2 Related Work
    5.3 Proposed DSET-SRF Algorithm
        5.3.1 DS-Evidence Theory:Review
        5.3.2 Main Steps of DSET-SRF Algorithm
    5.4 Experiments and Results
        5.4.1 Data-Sets
        5.4.2 Evaluation Metrics
        5.4.3 Performance Comparison
    5.5 Conclusions
6 Summary and Future Work
    6.1 Introduction
    6.2 Summary
    6.3 Future Work
Abstract of Innovation Points
References
Publications and Research Achievements During Ph.D. Period
Acknowledgement
About the Author



本文編號(hào)：3500001