WEKO3
アイテム
空間的なテクスチャ解析によるコンプレックスネットワークに基づくテクスチャ解析の改善
https://doi.org/10.18997/00006988
https://doi.org/10.18997/0000698811637fe5-ee23-4c88-a530-3772b36e3e5e
| 名前 / ファイル | ライセンス | アクション |
|---|---|---|
sei_k_329.pdf (25.4 MB)
|
|
| アイテムタイプ | 学位論文 = Thesis or Dissertation(1) | |||||||
|---|---|---|---|---|---|---|---|---|
| 公開日 | 2018-11-28 | |||||||
| 資源タイプ | ||||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_db06 | |||||||
| 資源タイプ | doctoral thesis | |||||||
| タイトル | ||||||||
| タイトル | Enhancement of Complex Network-based Texture Characterization by Spatial Texture Analysis | |||||||
| 言語 | en | |||||||
| タイトル | ||||||||
| タイトル | 空間的なテクスチャ解析によるコンプレックスネットワークに基づくテクスチャ解析の改善 | |||||||
| 言語 | ja | |||||||
| 言語 | ||||||||
| 言語 | eng | |||||||
| 著者 |
Thewsuwan, Srisupang
× Thewsuwan, Srisupang
|
|||||||
| 抄録 | ||||||||
| 内容記述タイプ | Abstract | |||||||
| 内容記述 | This thesis proposes a new texture analysis model which enhanced from traditional complex network-based model for texture characterization via spatial texture analysis. The conceptual framework of the proposed model is to synergize between pattern recognition and graph theory research areas. The results of experiment show that the proposed model can capture robust textural information under various uncontrolled environments using standard texture databases. Texture analysis has played an important role in the last few decades. There are a growing number of techniques described in the literature, one of new area research is a complex network for texture characterization, which has developed in recent years. Inspired by the human brain system, the relation among structure texture elements on an image can be derived using the complex network model. Compared to the task of texture classification, development of the original complex network model is required in order to improve classification performance in environment variations. To fulfill this requirement, the enhancing complex network by spatial texture analysis (i.e., spatial distribution and spatial relation) has been achieved in this thesis. The proposed approach addresses the above requirement by investigating and modifying the original complex network model by extracting more discriminative information. A new graph connectivity measurement has been devised, including local spatial pattern mapping, which is denoted as a LSPM, to encode and describe local spatial arrangement of pixels. To the best of the author's knowledge, as investigated in this thesis, the encoding spatial information which has been adapted within the original complex network model presented here were first proposed and reported by the author. The essence of this proposed graph connectivity measurement describes the spatial structure of local image texture cause it can effectively capture and detect micro-structures (e.g., edges, lines, spots) information which is critical being used to distinguish various pattern structures and invariant uncontrolled environments. Moreover, the graph-based representation has been investigated for improving the performance of texture classification. Spatial vector property has been comprised of deterministic graph modeling which decomposing the two component of the magnitude and the direction. Then, the proposed hybrid-based complex network comprises the enhancing graph-based representation, and the new graph connectivity measurement has been devised as an enhancing complex network-based model for texture characterization in this thesis. The experiments are evaluated by using four standard texture databases include Brodatz, UIUC, KTH-TIPS, and UMD. The experimental results are presented in terms of classification rate in this thesis to demonstrate that: firstly, the proposed graph connectivity measurement (LSPM) approach achieved on-average 86.25%, 77.25%, 89.38% and 94.06% respectively based on four databases. Secondly, the proposed graph-based spatial property approach achieved on-average 90.92%, 87.92%, 96.56% and 92.65%, respectively; finally, the hybrid-based complex network model achieved on-average 88.92%, 85.46%, 95.14% and 95.52% respectively. Accordingly, this thesis has advanced the original complex network-based model for texture characterization. | |||||||
| 言語 | en | |||||||
| 目次 | ||||||||
| 内容記述タイプ | TableOfContents | |||||||
| 内容記述 | 1 Introduction||2 Literature Review||3 Complex Network Model and Spatial Information||4 Graph-based Representation in Texture Analysis||5 Hybrid-based Complex Network Model||6 Conclusions | |||||||
| 備考 | ||||||||
| 内容記述タイプ | Other | |||||||
| 内容記述 | 九州工業大学博士学位論文 学位記番号：生工博甲第329号 学位授与年月日：平成30年9月21日 | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Complex network | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Texture characterization | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Texture representation | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Texture analysis | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Spatial texture analysis | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | LBP | |||||||
| アドバイザー | ||||||||
| 堀尾, 恵一 | ||||||||
| 学位授与番号 | ||||||||
| 学位授与番号 | 甲第329号 | |||||||
| 学位名 | ||||||||
| 学位名 | 博士（工学） | |||||||
| 学位授与年月日 | ||||||||
| 学位授与年月日 | 2018-09-21 | |||||||
| 学位授与機関 | ||||||||
| 学位授与機関識別子Scheme | kakenhi | |||||||
| 学位授与機関識別子 | 17104 | |||||||
| 学位授与機関名 | 九州工業大学 | |||||||
| 学位授与年度 | ||||||||
| 内容記述タイプ | Other | |||||||
| 内容記述 | 平成30年度 | |||||||
| 出版タイプ | ||||||||
| 出版タイプ | VoR | |||||||
| 出版タイプResource | http://purl.org/coar/version/c_970fb48d4fbd8a85 | |||||||
| アクセス権 | ||||||||
| アクセス権 | open access | |||||||
| アクセス権URI | http://purl.org/coar/access_right/c_abf2 | |||||||
| ID登録 | ||||||||
| ID登録 | 10.18997/00006988 | |||||||
| ID登録タイプ | JaLC | |||||||
