以四元樹為基礎抽取圖片物件特徵 之 影像檢索 專題J組 指導教授：曾修宜教授 組員 ： 楊智宇 91156207 黃文宣 91156250 劉濬毅 91155315
TABLE OF CONTENT Introduction Image Features 。Color 。Texture Quadtree Decomposition Representative feature extraction 。Vector quantization (VQ) algorithm 。 Representative feature extraction of objects using VQ Implementation
Introduction Content-Based Image Retrieval An application of computer vision to the image retrieval problem, that is, the problem of searching for digital images in large databases
Introduction Target 我們想要達到的目的為： 1. 讓程式透過Quadtree可以 縮小人類感官與電腦處理的差距 2. 增加search系列套圖的成功率 3. 減低圖片處理的資料量
Image Features - Color HSV 。Hue distinguish colors 。Saturation the percentage of white light that is added to a pure color 。Value perceived light intensity The HSV color model, which is similar to human perception, is most frequently used for retrieval.
Image Features - Color RGB (Red Green Blue)
Image Features - Texture Angular Second Moment = Contrast = Correlation = Variance = Entropy =
Quadtree Decomposition An image is divided into rectangular blocks 7x7 is the smallest block we define 7
Representative feature extraction Vector Quantization (VQ) algorithm 主要概念： 配合前述的Quadtree，把圖片重新分割為block 再把切割出來的block配合之前的Texture公式對應到 一個8-dimension vector space中的座標 再配合Vector Quantization把vector分群
Representative feature extraction Representative feature extraction of objects using VQ
Representative feature extraction Representative feature extraction of objects using VQ
Implementation A Simple Follow Chart
Implementation Search Result