Abstract: Abstract: Super-resolution technique is the task of reconstructing High-Resolution (HR) image from a sequence of Low-Resolution (LR) images. Quantization Constrain Set (QCS) was widely used as priori information about the coding process in super-resolution reconstruction of compressed video. An adaptive quantization constrain set (AQCS) was proposed by using statistical property of quantization errors based on theory of projection onto the Narrow Quantization Constrain Set (NQCS). A new smooth constrain set (SCS) was proposed by using the property of DCT transformed block edge. The experimental results showed that proposed AQCS outperformed traditional QCS in both Peak Signal to Noise Ratio (PSNR) and subjective image quality.

Key words: Compressed Video, Projection onto Convex Set, Super-Resolution Reconstruction, Quantization Constrain Set, Smooth Constrain Set

摘要： 摘要：超分辨率技术是使用低分辨率图像序列来重建高分辨率图像的技术。在压缩视频的超分辨率重建中，量化约束集(QCS)作为编码模型的先验信息被广泛采用。根据窄量化约束集(NQCS)理论，利用量化误差的统计特性，提出了一种改进量化约束集(AQCS)。根据DCT变换后块边界特性，提出了平滑约束集。实验结果表明，提出的基于改进量化约束集的压缩视频超分辨率重建算法较传统的量化约束集，在峰值信噪比(PSNR)和主观图像质量上有不同程度的提高。

关键词: 压缩视频, 凸集投影, 超分辨率重建, 量化约束集, 平滑约束集