摘要
传统舰船图像分解方法无法有效对图像超像素稀疏部分进行有效像素深度的恢复,导致分解图像缺失底层像素信息,噪声分解图像失真。为了解决上述问题,提出Moreau包络和深度恢复的舰船图像分解方法。首先通过引入Moreau包络算法,对图像分解区域进行超像素层的重构计算;然后利用稀疏深度恢复算法,对重构的超像素层待分解像素信息进行深度恢复优化。最后,通过像素点之间信息的关联特性,与神经网络学习特性,完成对恢复区域构成图像的分解。通过与传统神经网络分解方法的实例对比表明,提出方法分解后的图像输出效果,优于传统分解方法,更适合舰船图像分解的应用场景。
The traditional ship image decomposition method can not effectively restore the sparse part of the image super-pixel,and restore the effective pixel depth,resulting in the decomposition image missing the underlying pixel information,noise decomposition image distortion.In order to solve the above problems,the method of ship image decomposition of Moreau envelope and deep recovery is proposed.First,by introducing the Moreau envelope algorithm,the reconstruction calculation of the image decomposition area is carried out,then the reconstructed super-pixel layer is optimized by using the sparse depth recovery algorithm,and finally,the decomposition of the images that constitute the recovery area is completed by the correlation characteristics of the information between the pixels and the neural network learning characteristics.By comparing with the examples of the traditional neural network decomposition method,it is shown that the image output effect after the decomposition of the method is better than that of the traditional decomposition method,and more suitable for the application of ship image decomposition.
作者
曹英
CAO Ying(Wuxi Vocational College of Science and Technology College of Artificial Intelligence,Wuxi 214000,China)
出处
《舰船科学技术》
北大核心
2021年第16期148-150,共3页
Ship Science and Technology
