基于5阶细胞神经网络图像浮雕算法研究

A study of the fifth-order CNN relief algorithm

针对如何使图像中的浮雕效果更具有立体感的问题,设计了关于图像的5阶细胞神经网络(CNN)的浮雕算法研究的方案,研究了该算法对于图像的立体化处理的过程,对于同一图像分别采用5阶CNN、3阶CNN算法和拉普拉斯算法进行对图像的立体化处理,并比较了几类算法的优劣.实验表明,基于5阶CNN浮雕算法更接近现实,直观效果更好,在实现图像的实时处理中,能够有效地生成具有艺术效果的浮雕图像.同时,将其应用在遥感图像的处理工作中,取得了很好的效果.

A new method named the fifth-order CNN relief algorithm is proposed,aiming at making the relief effect in the image more three-dimensional and revealing the relevant process. This research treats the same image by using the fifth-order CNN algorithm,the third-order CNN algorithm and the Laplace algorithm respectively,and compares their strengths and weaknesses. It concludes that this new algorithm can better reflect the truth,and produce better relief images with good artistic effects. Meanwhile,it will also produce good effects when applied to the remote-sensing images.