基于机器学习的超分辨率问题的研究

Machine Learning Based Super-resolution and Its Implementation in Computer Vision

本文给出了一种基于机器学习的处理计算机视觉超分辨率问题的方法,可使原本模糊的图像变得清晰(通过从低分辨率图像中估计出高频细节信息).该方法以马尔可夫网络为基础体系,从大量的例子中学习并获得网络参数,利用贝叶斯信任传播机制为所要处理的图像找到一个后验概率的局部最大值,即为从图像中获得相应的景物而进行问题估计。从而生成一个与所提供的图像相符合的合成的景物世界、实验证明。

This paper show a machine learning-based method for 'super-resolution'in low-level vision, and it can show good results (estimating high frequency details from a low resolution image). It model that world with a Markov network, learning the network parameters from the examples. Bayesian belief propagation allows it to efficiently find a local maximum of the posterior probability for the scene, given the image. It generate a synthetic world of scenes and their corresponding rendered images.