基于卷积神经网络的数字图像模糊增强算法

Digital image fuzzy enhancement algorithm based on convolutional neural network

为了解决数字图像灰度信息丢失、边缘模糊导致的图像效果不佳等问题,提出基于卷积神经网络的数字图像模糊增强算法。以图像灰度级作为论域构建特征模糊集合,拟定模糊隶属度,基于模糊熵计算图像边缘灰度值,同时,构建不同光照环境下限制对比度均衡派生图、伽马转换派生图、对数转换派生图、两通道增强派生图,借此微小调整图像像素方差,强化细节信息的表达,增强隶属度;最后,通过卷积神经网络筛选图像内渡越点,以此调整图像内对比度、平均亮度与像素方差,实现数字图像的模糊增强。实验结果表明,采用本文方法增强数字图像模糊的效果较好、且不易受环境影响。

In order to solve the problems of gray information loss and poor image effect caused by edge blur of digital image,a digital image fuzzy enhancement algorithm based on convolution neural network is proposed.At the same time,restricted contrast balanced derivative image,gamma conversion derivative image,logarithmic conversion derivative image and two-channel enhancement derivative image are constructed under different lighting environments,so as to slightly adjust the variance of image pixels and enhance the expression of detail information.Finally,the convolution neural network is used to filter the transition points in the image,so as to adjust the contrast,average brightness and pixel variance in the image,and the fuzzy enhancement of the digital image was realized.Experimental results show that the proposed method can enhance the effect of digital image blur better,and is not easy to be affected by the environment.