基于图像分辨率增强算法的场景生成技术

针对高分辨率真实感的虚拟环境及场景浏览时变焦观察的需要,研究了基于图像的高分辨率场景生成技术。图像分辨率增强技术是指利用已采样的信息来重新构建分辨率更高的场景图像,包含单帧图像的分辨率增强和多帧图像序列的分辨率增强两种技术。对于单帧图像的分辨率增强技术,提出了一种基于熵变分的图像分辨率增强算法。该算法在贝叶斯估计和最大熵原理的基础上,将图像像素点梯度信息应用到图像分辨率增强中,从而建立起一种基于图像梯度信息的各向异性自适应分辨率增强算法。对于多帧图像序列的超分辨率复原技术,在单帧熵变分模型的基础上,将双边滤波技术引入到图像超分辨率复原中,建立了一种基于广义熵变分的图像超分辨率复原模型,提出了一种基于几何距离和梯度信息的双重加权各向异性分辨率增强算法。实验结果表明:使用本文算法得到的高分辨率复原图像具有较高的峰值信噪比和视觉质量,与传统图像分辨率增强算法相比具有一定的优势。

Virtual Scene Genaration Technology Based on Algorithm of Image Resolution Enhancement

The scene generation technology based on algorithm of image resolution enhancement,aiming at the needs of high image resolution and zooming browse in virtual environment was studied.The algorithm of image resolution enhancement contains two situations.First,for one frame,a novel image zooming algorithm based on maximum a posteriori probability and maximum entropy principle was presented,which anisotropy interpolate pixels using gradient information self-adaptively.Second,in the case of multi-frames,super-re...