林宗男臺灣大學:電機工程學研究所金雅祥Chin, Ya-HsienYa-HsienChin2007-11-262018-07-062007-11-262018-07-062007http://ntur.lib.ntu.edu.tw//handle/246246/53402影像增強是一個非常廣泛的領域,最常見的應用就是我們日常生活中的數位相片。由於拍攝環境條件的不同或是拍攝者的差異,使得照出來的影像常常有局部過暗或過亮情形發生。或是在自動辨識系統,常用到的前處理以用來取出較有鑑別度的特徵,讓我們更能精確地去做分類。有些應用於室內的監視系統,因為在室內光源不足情況下,使的某些影像中會產生偏暗的區域,以致無法有效的達到監控的目的。因此就需要透過影像增強的技術來對這些影像做矯正。 因此我們希望能對這些低對比的影像區塊作適當的局部增強,達到當用肉眼去觀測時能符合使用者的需求,而不至於破壞影像中其餘的影像區塊。由於人的感官對於影像中的亮度較為敏銳,而彩色影像RGB是屬於高相關的色彩空間,不易於處理。所以我們將影像轉換至YCBCR的色彩空間,針對亮度(Luminance)來做調整。藉由增加影像評估方程式中低對比影像區塊的成分,和四個非線性曲線的組合,透過基因演算法來求得適合的轉換曲線,以用來增強局部低對比的影像。另外再使用空間擴散濾波器對邊緣的像素作一個較平滑的處理,讓整張影像看起來會較自然。一方面達到局部增強的效果,一方面也不至於會破壞其於高對比的影像區塊。Image enhancement ranges over numerous subjects. The most common application is digital photos taken in our daily life. Due to the different environmental conditions or the photographers, it frequently leads into too dark or bright images in local area. Either in certain Pattern Recognition(PR) system, by the way of pre-processing as known as image enhancement we could obtain more distinguishing features in order to classify them further precisely. As for some indoor surveillance systems, we fail to monitor efficiently because of the poor light source generating over dark areas in a few images. Thus, we will modify this through the technology of image enhancement. Hence we wish that we could make some proper local enhancements as to meet the demands of naked-eyed observations without destroying the rest parts. Owing to the higher sensitivity of human beings to luminance and RGB color space is higher correlated not easy to process, we usually transfer images to YCbCr color space then do certain adjustments in accordance with the luminance. Through increasing the weight of low contrast area and GA optimization for solution space, we could attain the proper transfer curve for enhancing local low contrast images. Together with a use of Spatial Diffusion Filter processing the boundary problems arising from two different transfer curves, we can achieve the goal of local enhancement without destroying the quality of rest areas in the meanwhile.第一章、 簡介 6 第二章、影像增強的方法: 12 一、:色彩空間轉換 12 二、:基因演算法(Genetic Algorithm) 14 三、:局部影像增強的限制 16 第三章、提出的架構: 19 一、局部增強: 19 二、空間擴散濾波器: 21 三、影像增強流程: 24 第四章、實驗的結果及數據: 26 第五章、結論: 38 參考文獻 395807320 bytesapplication/pdfen-US影像增強影像品質評估基因演算法image enhancementimage quality evaluationgenetic algorithm權重式局部影像增強Weight-Based Local Image Enhancementthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/53402/1/ntu-96-J93921052-1.pdf