压缩图像的变换域凸集投影法超分辨率重建

Super-resolution Reconstruction of the Compressed Image Using Transform POCS

在压缩格式中,视频序列被表示为运动矢量和传输系数的组合,而这些信息并没有被传统的超分辨率重建算法利用。在文中,直接利用量化间隔信息,采用凸集投影(POCS)的方法,给出了一种基于运动补偿的,离散余弦(DCT)域的算法模型,重建一幅高分辨率图像。实验表明,该方法能得到比普通解码或空间域重建更好的效果。

Compression schemes describe a video sequence through a combination of motion vectors and transform coefficients which are not utilized by the traditional super-resolution reconstruction formulation. This paper proposes a motion-compensated DCT domain formulation by directly using quantization-bound information and applying the POCS technique and creates a high quality image. Experiments show that this formulation works better than the ordinary decoding or space domain reconstruction.