论文部分内容阅读
针对基于双向深度图像绘制技术(Double-sided Depth-image Based Rendering,Double-sided DIBR)中产生的空洞,重采样,重叠问题,为提高虚拟图像的合成质量,提出一种改进的正反向映射技术。该技术主要有四点贡献。一提出一种深度差值估计法。二在3D-warping过程中使用改进的基于Z-buffer的OPFD算法,有效解决重采样和重叠问题。三对深度虚拟图像运用改进的基于背景空洞填补算法消除空洞。四改进反向映射过程,通过判断投影后的图像和辅助彩色参考图像被遮挡信息背景的一致性,选择不同的空洞填补算法填补彩色虚拟图像中的空洞从而达到更好的填补效果。实验结果表明,改进技术在降低算法复杂度的同时,在主观图像质量与客观峰值信噪比(Peak Signal to Noise Ratio,PSNR)以及结构相似(Structural SIMilarity,SSIM)都有所提高。
To solve the problems of voids, resamples and overlaps in Double-sided DIBR (Double-sided Depth-image Based Rendering), an improved forward and backward mapping is proposed to improve the quality of virtual images. technology. This technology mainly has four contributions. First proposed a depth difference estimation method. Second, the improved Z-buffer-based OPFD algorithm is used in the 3D-warping process to effectively solve the resampling and overlay problems. Three pairs of depth virtual images utilize an improved background-based hole-filling algorithm to eliminate holes. Four improved reverse mapping process, by judging the projection of the image and auxiliary color reference image is occluded information background consistency, choose different hole filling algorithm to fill the hole in the color virtual image so as to achieve a better fill effect. The experimental results show that the proposed technique improves the subjective image quality and PSNR and Structural SIMilarity (SSIM) while reducing the complexity of the algorithm.