计算机系统应用
計算機繫統應用
계산궤계통응용
APPLICATIONS OF THE COMPUTER SYSTEMS
2014年
8期
198-202
,共5页
艾木拉姑丽·卡得尔%吐尔洪江·阿布都克力木%阿卜杜如苏力·奥斯曼
艾木拉姑麗·卡得爾%吐爾洪江·阿佈都剋力木%阿蔔杜如囌力·奧斯曼
애목랍고려·잡득이%토이홍강·아포도극력목%아복두여소력·오사만
图像检索%二进小波变换%局部二值模式
圖像檢索%二進小波變換%跼部二值模式
도상검색%이진소파변환%국부이치모식
image retrieval%dyadic wavelet transform%uniform local binary pattern
纹理是一种反映像素的空间分布属性的图像特征,成为图像分析的重要手段。在图像检索领域,由于小波变换具有方向性、局部特性、多尺度特性,使得图像的纹理特征能得到更好的表征。局部二值模式是提取图像纹理特征的有效工具,具有旋转不变性和灰度不变性。利用性能良好的 Bubble 小波对图像进行二进小波变换得到四个子带图像,其中低频子带包含原图信息最多,对角线方向的子图包含信息最少,其余两个子图包含信息介于两者之间,因此选取低频和垂直,水平高频图像,然后对分量利用等价局部二值模式提取图像纹理特征,并进行图像检索。实验结果表明该算法不仅有效地结合了小波与局部二值模式的优势,而且具有较好的检索性能。
紋理是一種反映像素的空間分佈屬性的圖像特徵,成為圖像分析的重要手段。在圖像檢索領域,由于小波變換具有方嚮性、跼部特性、多呎度特性,使得圖像的紋理特徵能得到更好的錶徵。跼部二值模式是提取圖像紋理特徵的有效工具,具有鏇轉不變性和灰度不變性。利用性能良好的 Bubble 小波對圖像進行二進小波變換得到四箇子帶圖像,其中低頻子帶包含原圖信息最多,對角線方嚮的子圖包含信息最少,其餘兩箇子圖包含信息介于兩者之間,因此選取低頻和垂直,水平高頻圖像,然後對分量利用等價跼部二值模式提取圖像紋理特徵,併進行圖像檢索。實驗結果錶明該算法不僅有效地結閤瞭小波與跼部二值模式的優勢,而且具有較好的檢索性能。
문리시일충반영상소적공간분포속성적도상특정,성위도상분석적중요수단。재도상검색영역,유우소파변환구유방향성、국부특성、다척도특성,사득도상적문리특정능득도경호적표정。국부이치모식시제취도상문리특정적유효공구,구유선전불변성화회도불변성。이용성능량호적 Bubble 소파대도상진행이진소파변환득도사개자대도상,기중저빈자대포함원도신식최다,대각선방향적자도포함신식최소,기여량개자도포함신식개우량자지간,인차선취저빈화수직,수평고빈도상,연후대분량이용등개국부이치모식제취도상문리특정,병진행도상검색。실험결과표명해산법불부유효지결합료소파여국부이치모식적우세,이차구유교호적검색성능。
Texture is an image feature which reflects the pixel spatial distribution property and has become important means in image analysis. In the field of image retrieval, due to the directional wavelet transformation, local characteristics and multi-scale features can make the texture feature of image get better representation. Local binary pattern is an effective tools for extract texture characteristics of images and have rotation invariance and gray scale invariance. This paper uses Bubble wavelet which has good performance to do the image dyadic wavelet transformation. Four sub-band images are got. The low frequency sub-band contains maximum art-work informations. The diagonal directional graph contains minimum informations. The other two graphs contain informations between the max and the min. Hence, low frequency, vertical and horizontal high frequency image are to be selected first. Then, by means of the uniform local binary pattern, the three texture components of image are extracted, and the image retrieval proceeded. The experimental results show that the algorithm not only effectively combines the advantages of wavelet and local binary pattern, and also has better retrieval performance.