点击次数：

论文类型：会议论文

发表时间：2017-01-01

收录刊物：Scopus

页面范围：1-5

摘要：Nowadays learning based approaches have been widely used in image processing and have achieved better results than classical methods. The core of the learning approach is data and the performance of the learning model can be greatly improved by the modality-specific training examples. However, few learning based methods study training example searching to optimize the data driven model. In our paper, we use Gaussian Process Regression (GPR) to learn the relationship between the hazy image and the transmission map. In order to optimize the learning model, we propose a training data searching method which adapts to the GPR model. For the given test examples, we first use k-dimensional tree to select training examples neighboring to the inputs. Then, based on the optimized GPR, we establish the relationship between hazy features and transmissions, and produce the transmission map of the hazy image for dehazing. Experimental results on the hazy image dataset show the effectiveness of the proposed method compared with the state-of-the-art dehazing methods. ? 2016 IEEE.