软件
軟件
연건
SOFT WARE
2014年
2期
59-62
,共4页
图像去噪%非局部均值%结构相似性
圖像去譟%非跼部均值%結構相似性
도상거조%비국부균치%결구상사성
image denoising%non-local means%structure similarity
非局部均值(Non-Local Means，NLM)利用图像块之间灰度值的欧式距离确定权重，距离越小权重越大。不过，这种权重计算方法忽略了图像块之间的结构相似性。为解决这个问题，提出了一种结合图像结构信息的非局部均值去噪方法。该方法同时利用图像块之间的灰度距离、结构相似性来确定权重。实验结果表明该方法在PSNR、SSIM标准下均优于NLM及其一些改进算法。
非跼部均值(Non-Local Means，NLM)利用圖像塊之間灰度值的歐式距離確定權重，距離越小權重越大。不過，這種權重計算方法忽略瞭圖像塊之間的結構相似性。為解決這箇問題，提齣瞭一種結閤圖像結構信息的非跼部均值去譟方法。該方法同時利用圖像塊之間的灰度距離、結構相似性來確定權重。實驗結果錶明該方法在PSNR、SSIM標準下均優于NLM及其一些改進算法。
비국부균치(Non-Local Means，NLM)이용도상괴지간회도치적구식거리학정권중，거리월소권중월대。불과，저충권중계산방법홀략료도상괴지간적결구상사성。위해결저개문제，제출료일충결합도상결구신식적비국부균치거조방법。해방법동시이용도상괴지간적회도거리、결구상사성래학정권중。실험결과표명해방법재PSNR、SSIM표준하균우우NLM급기일사개진산법。
The weight of Non-Local means is determined by the Euclidean distance between the image patches. Smaller distance, larger weight. However, the weight obtained in this way ignores the structure similarity between the image patches. To address this problem, a structure based non-local means (SBNLM) is proposed. This approach utilizes both the intensity distance and structure similarity to determine the weight. Experimental results shows that this method is superior to NLM and some of its improved methods in PSNR and SSIM.
