国产bbaaaaa片,成年美女黄网站色视频免费,成年黄大片,а天堂中文最新一区二区三区,成人精品视频一区二区三区尤物

首頁(yè)> 外文學(xué)位 >Algorithms for image classification using wavelets and fusion methods.
【24h】

Algorithms for image classification using wavelets and fusion methods.

機(jī)譯:使用小波和融合方法進(jìn)行圖像分類的算法。

獲取原文
獲取原文并翻譯 | 示例

摘要

This work presents a set of algorithms for feature extraction and image classification using wavelet transform and fusion methods. First we use wavelet transform as a feature extraction method. We present some basic definitions and theoretical background of the problem and proposed methodologies for solving it.; We decompose the image to a second level decomposition giving us the coefficients of approximation, diagonal, horizontal and vertical. We used nine different wavelet filters in order to test how the classification performed depending on the filters, images and coefficients used. The algorithms used for image classification are Euclidean Distance, Maximum Likelihood and K-Nearest Neighbor. Then all of the results from the algorithms individually go to a fusion center where we use Majority Voting.; Images from the Brodatz album and an image of Mayaguez Bay are used. Different experiments were done trying to aim at the best results possible. We got really interesting results for some of the textures. Some of them gave us really good results for the classifiers and filters, but the others did not. We normalized all of the samples in order to achieve a better discrimination.
機(jī)譯:這項(xiàng)工作提出了一組使用小波變換和融合方法進(jìn)行特征提取和圖像分類的算法。首先,我們使用小波變換作為特征提取方法。我們提供了該問(wèn)題的一些基本定義和理論背景,并提出了解決該問(wèn)題的方法。我們將圖像分解為第二級(jí)分解,從而給出近似系數(shù),對(duì)角線,水平和垂直系數(shù)。我們使用了九種不同的小波濾波器,以測(cè)試如何根據(jù)所使用的濾波器,圖像和系數(shù)進(jìn)行分類。用于圖像分類的算法是歐氏距離,最大似然和K最近鄰。然后,這些算法的所有結(jié)果將分別進(jìn)入融合中心,在此我們將使用多數(shù)投票。使用來(lái)自Brodatz專輯的圖像和Mayaguez Bay的圖像。為了達(dá)到最佳結(jié)果,進(jìn)行了各種實(shí)驗(yàn)。對(duì)于某些紋理,我們獲得了非常有趣的結(jié)果。其中一些為分類器和過(guò)濾器提供了非常好的結(jié)果,而其他的則沒(méi)有。我們對(duì)所有樣本進(jìn)行了歸一化,以實(shí)現(xiàn)更好的區(qū)分度。

著錄項(xiàng)

相似文獻(xiàn)

  • 外文文獻(xiàn)
  • 中文文獻(xiàn)
  • 專利
獲取原文

客服郵箱:kefu@zhangqiaokeyan.com

京公網(wǎng)安備:11010802029741號(hào) ICP備案號(hào):京ICP備15016152號(hào)-6 六維聯(lián)合信息科技 (北京) 有限公司?版權(quán)所有
  • 客服微信

  • 服務(wù)號(hào)