毛雪 董歡 彭哲 何欞瑜
摘 要:圖像復原有很多經(jīng)典的方法,例如逆濾波,維納濾波,最大熵復原等,然而傳統(tǒng)的方法在解決函數(shù)逼近問題時存在著缺點,BP神經(jīng)網(wǎng)絡因為其高速的并行計算能力和自適應學習能力在這方面顯示出優(yōu)勢。BP網(wǎng)絡本質(zhì)上是一種輸入和輸出之間的精確的數(shù)學表達式,不需要知道兩者之間的明確的關系,只需要用已知的模式對其進行訓練,網(wǎng)絡就具有輸入與輸出的映射能力,對許多點擴散函數(shù)(PSF)的變化難以掌握,所以我們采用神經(jīng)網(wǎng)絡對其進行復原。
關鍵詞:運動模糊圖像;BP神經(jīng)網(wǎng)絡;圖像復原
Image restoration has a lot of the classic methods,such as inverse filter,wiener filter,maximum entropy restoration,etc.,but the traditional method in solving the problem of function approximation exist shortcomings,the BP neural network because of its high-speed parallel computing capacity and adaptive learning ability shows advantages in this respect.BP network is essentially a precise mathematical expression between the input and output,do not need to know the clear relationship between the two,only need to use a known pattern of training,the network will have the input and output mapping ability,for many the change of the point spread function(PSF)is difficult to grasp,so we use neural network to recover.
引言:現(xiàn)已有很多復原圖像的數(shù)字圖像處理技術 ,但其使用的條件苛刻,現(xiàn)實生活中使用過程麻煩,而且效果不理想。題目假設了模糊圖像中中的全部景觀都以同一個速度運動,我們給出了勻速運動模糊圖像復原的模型,利用維納濾波對其進行復原,為了得到更為良好的圖像復原,我們建立了BP神經(jīng)網(wǎng)絡的圖像復原模型,利用Sigmoid函數(shù) 進行傳輸,便能得到優(yōu)秀的復原圖像。
1傳統(tǒng)圖像退化數(shù)學模型
參考文獻
[1] 任金凡,運動模糊圖像復原算法的研究,電子科技大學,2014-04-23
(作者單位:西華大學電氣與電子信息學院)