基于多分辨率局部結(jié)構(gòu)化信息熵的魯棒多模圖像融合算法
Multiresolution Based Local Structured Information Entropy for Robust Multimodal Image Fusion
-
摘要: 目前的圖像融合算法不能區(qū)分噪聲和視覺上有意義的圖像特征，往往將噪聲當(dāng)作有意義的信息傳輸?shù)饺诤辖Y(jié)果中。針對這一問題，該文基于復(fù)數(shù)小波變換(CWT)，將圖像的結(jié)構(gòu)化特征表現(xiàn)在不同尺度和方向上，定義了兩種結(jié)構(gòu)化信息熵，表達(dá)局部圖像結(jié)構(gòu)化程度：帶內(nèi)結(jié)構(gòu)化信息熵，以及考慮帶間特征相關(guān)性的結(jié)構(gòu)化信息熵。利用定義的兩種測度，在圖像融合之前對輸入加權(quán)處理，使視覺上有意義的信息在融合結(jié)果中自適應(yīng)地增強(qiáng)，而噪聲自適應(yīng)地抑制。通過對融合算法仿真結(jié)果的主觀比較和客觀性能分析，展示了本文提出的圖像融合算法的優(yōu)越性。Abstract: The updated image fusion schemes could not identify meaningful image features from noises, the input noise is treated as valid information and transferred into the fused output. After complex wavelet transformation (CWT), structured information is decomposed into varying scales and directions. Based on CWT, two structured information entropies, intra-band structured information entropy and inter-band structured information entropy, are formulated to express the structurization level of image features. Preceding the image fusion process, the metrics are employed to weight all inputs. As a result, the perceptual salient inputs are enhanced while the noise inputs are de-emphasized adaptively. Comparing the visual aesthetics of fusion results and analyzing the performance objectively, show the good performance of the proposed image fusion algorithm.
-
計量
- 文章訪問數(shù): 2880
- HTML全文瀏覽量: 80
- PDF下載量: 1119
- 被引次數(shù): 0
下載:
