基于全变分的全向图像稀疏重构算法

Sparse Reconstruction for Omnidirectional Image Based on Total Variation

折反射全向成像由于曲面镜的反射作用,导致全向图像存在严重变形,传统的梯度计算方法在全向图像中不能很好地符合折反射成像的特点.为了从压缩采样数据快速有效地重构全向图像,提出了一种结合全向图像特征的全变分模型——全向全变分,并在基于TV范数进行全向图像重构时,采用全向全变分作为目标函数,进行模型的求解.实验结果验证了本文算法的有效性和可行性,其重构结果的主客观效果明显优于传统TV模型.

Because of the distortions produced by the reflection of a mirror ,catadioptric omnidirectional images cannot be processed similarly to classical perspective images .In this paper ,we propose to define a new model named omnidirectional total variation （Omni-TV） ,which reflects the omnidirectional image structure features .In order to reconstruct the images from compres-sive samples ,the Omni-TV is used as the subject function during the image reconstruction .The simulation results show that the om-nidirectional images could be reconstructed effectively and accurately .Comparing with classical TV minimization model ,the images , which are recovered based on Omni-TV model ,can provide higher quality both in subjective evaluation and objective evaluation .