Hybrid method to obtain interest region and non interest region for color based image retrieval
Content based image retrieval (CBIR) has become one of the most active research areas in the past few years. Many indexing techniques are based on global feature distributions. However, these global distributions have limited discriminating power because they are unable to capture...
保存先:
| 主要な著者: | , , |
|---|---|
| フォーマット: | 論文 |
| 言語: | English |
| 出版事項: |
International Center for Scientific Research and Studies
2015
|
| 主題: | |
| オンライン･アクセス: | http://eprints.unisza.edu.my/6887/1/FH02-FIK-15-04678.jpg http://eprints.unisza.edu.my/6887/ |
| タグ: |
タグ追加
タグなし, このレコードへの初めてのタグを付けませんか!
|
| 要約: | Content based image retrieval (CBIR) has become one of the most active research areas in the past few years. Many indexing techniques are based on global feature distributions. However, these global distributions have limited discriminating power because they are unable to capture local image information. In this paper, the new proposed method based on local image to classify the Interest Region (IR) and Non Interest Region (NIR) of images. To develop this, the integration of clustering and user intervention was applied. Clustering process is obtaining several regions, meanwhile to ascertain the location of the center of images through user intervention. Several experiments are conducted using different weight (ω, γ) of IR and NIR. Subsequently average color moment is extracted from this region (IR and NIR) in CIE Lab color model. To investigate the performance, new distance is proposed based on Euclidean distance. Experimental results show the proposed method more efficient in image retrieval. |
|---|