改进的基于卷积神经网络的图像超分辨率方法

图像超分辨率是计算机视觉领域的经典问题。使用深度神经网络来解决图像超分辨率的问题目前得到越来越多的研究学者的关注和青睐。为改善基于卷积神经网络的图像超分辨率方法的图像生成效果,提出一种改进的方法。在神经网络层中加深网络层数,并且针对加深网络可能出现的退化现象引入残差网络结构,并将图像上采样步骤放入网络中。实验表明,在与传统的插值法和原始的基于卷积神经网络方法的对比中,该优化方法生成的图像观感更加锐利清晰、细节丰富,而且无论在峰值信噪比和结构相似性上均有明显提高,验证了该方法的有效性。

AN IMPROVED IMAGE SUPER-RESOLUTION METHOD BASED ON CONVOLUTION NEURAL NETWORK

Image super-resolution is a classical problem in the field of computer vision. The use of depth neural network to solve the problem of image super-resolution has attracted more and more attention and favor of researchers. In order to improve the image generation effect of image super-resolution method based on convolution neural network, an improved method was proposed. It mainly increased the number of network layers in the neural network layer, introduced residual network structure for the degradation phenomenon that might occur in the deepening network, and put the image sampling steps into the network. Experiments show that compared with the traditional interpolation method and the original method based on convolutional neural network, the optimization method generates sharper and clear image with richer details. And it has significant improvement both in peak signal-to-noise ratio and structural similarity, which verifies the effectiveness of this method.