多媒体视觉VR重构算法的改进与仿真

传统图像重构噪声偏大,图像离群点较多,导致重构效果与实际图像偏差大,研究提出多媒体视觉VR重构算法。利用双目偏移定位重新校正VR图像,对重构后的VR图像进行去噪处理,完成基于高通滤波矩阵加强的重构图像纹路细节特征,采用三维重构相关性检索方法,获得合群点与离群点,确定其最大最小深度范围,对错误识别和峰值进行重新定位,完成算法设计。为验证研究的有效性,进行实验验证。通过实验结果数据分析可证明,研究算法大大提升了重构图像的真实度,图像角点检测准确,重构后图像噪声更小。

Improvement and Simulation of VR Reconstruction Algorithm for Multimedia Vision

Traditionally, the reconstructed image has many outliers, leading to large deviation between the reconstruction effect and the actual image. Therefore, a multimedia visual VR reconstruction algorithm was proposed. Firstly, the binocular visual positioning was used to recalibrate VR image and reduce the noise from reconstructed VR image, so as to strengthen the texture detail feature of reconstructed image based on high-pass filter matrix. Secondly, the relevance retrieval method based on 3 D reconstruction was used to get gregarious points and outliers, and determine maximum-minimum depth range. Moreover, the peak value and error recognition was located again, so as to complete the algorithm design. Simulation results verify that the proposed algorithm greatly improves the authenticity of reconstructed images. The corner point detection of image is more accurate and the noise of reconstructed image is smaller.