本文選題：圖像檢索 + 顏色特征�。� 參考：《重慶理工大學》2012年碩士論文

【摘要】：隨著網(wǎng)絡技術、多媒體技術和數(shù)碼技術的飛速發(fā)展，網(wǎng)絡中的圖像資源日益豐富起來。為了滿足廣大網(wǎng)絡用戶對圖像檢索的要求，各種基于Web的圖像搜索引擎如雨后春筍般冒了出來。通常，人們在搜索圖像時，最關心的是搜索結(jié)果是否符合用戶的檢索要求。而檢索結(jié)果的準確性，是由圖像匹配算法的優(yōu)劣來決定的。 首先，從基于顏色特征的量化算法、描述方法以及匹配算法入手，針對現(xiàn)有的一些只考慮顏色總體比例而忽略顏色具體空間分布的圖像匹配算法的不足，提出了一種新的匹配算法，即基于分塊的不規(guī)則圖形相似比較的圖像匹配算法。具體算法的創(chuàng)新之處在于： （1）改進算法融入了顏色的空間信息，使得檢索結(jié)果更為準確。常用的顏色特征表示方法，如顏色直方圖法、顏色主色法等，這些算法都只考慮了圖像中顏色的整體比例，而忽略了圖像中各種顏色具體的空間分布信息。這就導致了兩幅具有相同顏色直方圖的圖像，可能因為其各自顏色分布不同，而使得兩幅圖像內(nèi)容相差很大。 （2）改進算法具有自動分塊的思想，它可以根據(jù)圖像中物體的特征自動分塊，，然后再取兩幅圖像中各塊進行匹配，提高了檢索的準確性。某些改進的圖像匹配算法，如基于分塊的顏色直方圖法或基于分塊的顏色主色法等，通過對圖像分塊而考慮了顏色的空間信息。但是該類方法都是采用固定分塊的方式來處理圖像，而且其分塊的數(shù)目沒有結(jié)合實際圖像中物體的特征來確定。一旦設定一個具體的分塊數(shù)目后，不同的圖像可能因為設定的塊數(shù)與圖像內(nèi)容有很大的差別而使得圖像的分割造成誤差。 其次，改進算法與其他算法進行了性能比較，設計并對系統(tǒng)進行了實現(xiàn)，然后通過實際檢索結(jié)果的對比，對改進算法的性能及實用性進行了比較。實驗證明，改進算法對圖像的旋轉(zhuǎn)、平移、尺寸等變化不相關，具有很好的穩(wěn)定性，能夠準確的檢索出用戶需求的圖片。
[Abstract]:With the rapid development of network technology, multimedia technology and digital technology, the image resources in the network are becoming more and more abundant. In order to meet the requirements of the vast number of web users for image retrieval, a variety of Web-based image search engines sprang up. Usually, when people search for images, they are most concerned about whether the search results meet the user's requirements. The accuracy of retrieval results is determined by the merits and demerits of image matching algorithm. First of all, starting with quantization algorithm based on color feature, description method and matching algorithm, aiming at the shortcomings of some existing image matching algorithms which only consider the proportion of color and ignore the specific spatial distribution of color. In this paper, a new matching algorithm is proposed, that is, an image matching algorithm based on block similarity comparison of irregular graphics. The innovations of the algorithm are as follows: (1) the improved algorithm incorporates color spatial information to make the retrieval results more accurate. The commonly used color feature representation methods, such as color histogram method, color master color method and so on, all of these algorithms only consider the overall proportion of the color in the image, but ignore the spatial distribution information of various colors in the image. As a result, two images with the same color histogram may differ greatly because of their different color distribution. (2) the improved algorithm has the idea of automatic partitioning. It can be automatically divided into blocks according to the features of the objects in the image, and then each block of the two images can be matched to improve the accuracy of the retrieval. Some improved image matching algorithms, such as the color histogram method based on block or the color master color method based on block, consider the spatial information of color by dividing the image into blocks. However, all of these methods deal with the image in a fixed block way, and the number of blocks is not determined by the characteristics of the object in the actual image. Once a specific number of blocks is set, different images may cause errors due to the difference between the number of blocks set and the content of the image. Secondly, the performance of the improved algorithm is compared with that of other algorithms, and the system is designed and implemented. Then, the performance and practicability of the improved algorithm are compared by comparing the actual retrieval results. Experiments show that the improved algorithm is not related to the image rotation, translation, size and other changes, has good stability, can accurately retrieve the user's needs of the picture.
【學位授予單位】：重慶理工大學
【學位級別】：碩士
【學位授予年份】：2012
【分類號】：TP391.41

【參考文獻】

相關期刊論文 前5條

1 李進;陳念;馬帥軍;明慧;;基于顏色的圖像檢索方法研究[J];軟件導刊;2010年04期

2 王濤,胡事民,孫家廣;基于顏色-空間特征的圖像檢索[J];軟件學報;2002年10期

3 何亞r

本文編號：2078930