【摘要】：顯著性檢測是圖像處理和計算機視覺領域的一個重要研究內容。本文探討利用獨立通道自適應縮放圖像的多尺度超圖進行顯著性檢測的方法。所做工作對圖像處理和計算機視覺領域的發(fā)展具有重要意義。目前經典的顯著性檢測方法有很多,但已有方法幾乎都沒有考慮圖像的R,G,B通道值差異對目標顯著性檢測結果的影響。本文將圖像的R,G,B通道值差異引入基于超圖建模的顯著性檢測方法中,提出了一種利用獨立通道自適應縮放進行顯著性檢測的方法。具體研究內容包括:(1)考慮到人眼對R,G,B三原色的敏感度差異,對圖像像素值進行獨立通道自適應縮放,得到初始圖像的獨立通道自適應縮放圖像;(2)將像素值的獨立通道自適應縮放方法引入超頂點和超邊構造中,并將自適應縮放因子與固定尺度經驗值相結合得到自適應的多尺度超邊和超頂點,從而設計基于獨立通道自適應縮放圖像的自適應多尺度超圖;(3)將所設計的基于獨立通道自適應縮放圖像的自適應多尺度超圖引入顯著性檢測方法中,設計基于獨立通道自適應縮放圖像超圖的顯著性檢測方法,并利用圖像分割實例展示所設計的顯著性檢測方法的有效性。首先利用梯度圖計算基于獨立通道自適應縮放圖像的超頂點和超邊的顯著度,得到多尺度超圖的顯著性圖,然后對多尺度超圖的顯著性圖進行融合,得到最終的基于獨立通道自適應縮放圖像超圖的顯著性圖,最后將所設計的顯著性檢測方法應用于圖像分割實例以說明其有效性;(4)將所設計的基于獨立通道自適應縮放圖像超圖的顯著性檢測方法在公開可獲得的圖像集MSRA-1000、SOD、SED和ImgSal-5上進行了測試,并與先前的6種經典顯著性檢測方法進行了比較。大量實驗表明,所提出的方法對于R,G,B通道值范圍差異較窄的圖像在一定程度上改善了顯著性檢測效果。本文提出的基于獨立通道自適應縮放圖像的顯著性檢測方法可應用于圖像分割、目標識別、圖像自適應壓縮、基于內容的圖像檢索等許多圖像處理和計算機視覺應用中,有助于改善其性能、提高其效率。
[Abstract]:Salience detection is an important research content in the field of image processing and computer vision. In this paper, we discuss a method to detect salience by using the multi-scale hypergraph of an adaptive scaling image with independent channels. The work done is of great significance to the development of image processing and computer vision. At present, there are many classical significance detection methods, but almost all of the existing methods do not consider the influence of the difference of RGG channel value on the target significance detection results. In this paper, the difference of RG B channel value of image is introduced into the significance detection method based on hypergraph modeling, and a new method of significance detection based on adaptive scaling of independent channel is proposed. The specific research contents are as follows: (1) considering the sensitivity difference of human eyes to RGZB, the pixel value of the image is scaled by independent channel adaptively, and the original image can be scaled by independent channel adaptively; (2) the independent channel adaptive scaling method of pixel value is introduced into hypervertex and hyper-edge construction, and the adaptive multi-scale super-edge and hypervertex are obtained by combining the adaptive scaling factor with the fixed scale empirical value. An adaptive multi-scale hypergraph based on independent channel adaptive scaling image is designed. (3) the adaptive multi-scale hypergraph based on the independent channel adaptive scaling image is introduced into the salience detection method, and the significance detection method based on the independent channel adaptive scaling image hypergraph is designed. An example of image segmentation is used to demonstrate the effectiveness of the proposed salience detection method. Firstly, using gradient graph to calculate the saliency of hypervertex and super-edge based on independent channel adaptive scaling image, the significance graph of multi-scale hypergraph is obtained, and then the significance graph of multi-scale hypergraph is fused. Finally, the final saliency graph based on independent channel adaptive scaling image hypergraph is obtained. Finally, the proposed saliency detection method is applied to the image segmentation example to demonstrate its effectiveness. (4) the salience detection method based on independent channel adaptive scaling image hypergraph is tested on the publicly available image sets MSRA-1000,SOD,SED and ImgSal-5. And compared with the previous six classical significance detection methods. A large number of experiments show that the proposed method can improve the significant detection effect to some extent for the images with narrow range of RGN B channel values. The salience detection method based on independent channel adaptive scaling image can be used in many image processing and computer vision applications, such as image segmentation, target recognition, image adaptive compression, content-based image retrieval and so on. It is helpful to improve its performance and improve its efficiency.
【學位授予單位】：山東大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TP391.41

【參考文獻】

相關博士學位論文 前1條

1 景慧昀;視覺顯著性檢測關鍵技術研究[D];哈爾濱工業(yè)大學;2014年

，

本文編號：2302799