查詢結果分析
相關文獻
- 以影像正規化及面積比對做形狀之辨認
- Invariant Image Matching by Compacting and Moment Normalization
- 認知型態與影像查詢比對關係探討(2)
- 位移組合概念發展之診斷與案例分析
- Efficient Block Matching Algorithm for Video Coding Using the Generalized Edge Feature Matching Technique
- A Fast Motion Estimation Vector Algorithm Using Successive Refinement by Subsampling
- Proton Chemical Shift Imaging of the Hippocampus in Patients with Complex Partial Seizures
- 數位音訊廣播系統之反交錯器及介面控制積體電路設計與實現
- 移動向量估測法之簡介
- 銲線機視覺辨識系統設計技術
頁籤選單縮合
題名 | 以影像正規化及面積比對做形狀之辨認=Shape Recognition Using Image Normalization and Area Matching |
---|---|
作者姓名(中文) | 游國幹; | 書刊名 | 龍華學報 |
卷期 | 16 民89.12 |
頁次 | 頁171-180 |
分類號 | 312.13 |
關鍵詞 | 形狀辨認; 位移; 放大縮小及旋轉之不變性; 比對; PSRI; Shape recognition; Position; Scale; Rotation invariance; Matching; |
語文 | 中文(Chinese) |
中文摘要 | 形狀辨認(Shape recognition)在電腦視覺(Computer vision)上是非常重要的,而位移、放大縮小及旋轉之不變性(Position, scale, and rotation invariance; PSRI)為形狀辨認之三大主要工作,本文提出一影像(Image)正規化(Normalization)方法將平面圖形(Pattern)轉換至某一特定位置、大小及方向以完成PSRI,接著用一簡單的面積比對(Matching)方法去辨認形狀。影像正規化方法分三個步驟,首先將圖形之中心點(Centroid)移至影像中心點以達成位移之不變性;然後旋轉圖形使其中心點至半徑加權平均點(Radius weighted mean point)所畫出的形狀特定(Shape-specific)方向線與正X座標軸同方向以達成旋轉之不變性;最後,用調整圖像的面積至整個影像面積的1/16以達成放大縮小之不變性。本法不需太多的運算,因此易於實現。 |
英文摘要 | Shape recognition is very important in computer vision. Position, scale, and rotation invariance (PSRI) are the three main tasks in shape recognition. In this paper, we propose a normalization algorithm that transforms a planar pattern into a specific position, size and direction to achieve PSRI, and then we can use a simple area matching method to recognize the shape. In the normalization algorithm, we first shift the pattern such that its centroid is translated to image center to achieve position invariance. Then we rotate the pattern such that its shape-specific direction from the centroid to the radius weighted mean point aligns with the positive X coordinate to achieve rotation invariance. Finally, we adjust the pattern area to 1/16 of the image size to achieve scale invariance. The algorithm does not need much computation; therefore, it is very easy to implement it. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。