基于亚像素补偿和视觉感知的超低码率压缩算法研究

With the emerging of big data for high resolution images and videos, an effective and efficient way to handle the super low bit-rate compression of big data is in high demand, particularly considering the big data in transportation monitoring , criminal surveillance, etc. existing algorithms usually achieve low bit-rate compression with the sacrifice of fine details. Nevertheless, quite a lot of fine details may be of vital importance for image/video contents understanding, such as the character information in surveillance videos. In this proposal, we are target to solve the two problems in super low bit-rate compression that "serious loss of fine details" and "superior high complexity in reconstruction", by proposing effective and efficient algorithms as well as the corresponding platforms based on subpixel rendering and visual perception. More specifically, we will take full advantage of the fact that each pixel in LCD consists of three independent space-addressable subpixels (red, green, blue), combine adjacent subpixels, and utilize the characteristics of human visual perception, to increase the individual reconstruction units during sampling, compression, restoration. Meanwhile, we will investigate the frequency-domain characteristics of apparent luminance details and color distortion in subpixel rendering, with which a multi-channel adaptive super low bit-rate compression algorithm will be particularly proposed to achieve high apparent resolution for fine but important detailed contents. With the proposed schemes, we will provide an effective real-time super low bit-rate compression platform, serving for the compression, transmission, and display of the mass data in high resolution image/video.

伴随着图像视频数据的海量化，我们急需解决海量数据的超低码率实时压缩问题，尤其在车船监控、犯罪监控等领域。现有压缩方法通常以牺牲细节内容来实现低码率，然而图像视频中重要细节往往对图像视频内容的理解起决定性的作用，比如监控视频中的文字信息等。本课题针对超低码率压缩下的严重细节丢失及重建复杂度较高的问题，研究基于亚像素补偿和视觉感知的新型超低码率压缩算法及应用平台。为增强图像视频中重要细节内容的视觉分辨率，本课题拟充分利用LCD上每个像素由3个独立空间可寻址的亚像素(红绿蓝)组成，通过自由控制组合相邻亚像素，结合人类视觉感知特性，增加在超低码率下的采样、压缩、复原中重建点数目；并通过研究亚像素级别的灰度细节与色度失真的频域特性，设计多通道自适应的超低码率压缩算法，有效的避免丰富细节内容的严重丢失。本课题通过上述研究，将实现一个高效实时的超低码率压缩平台，切实服务于海量高清数据的压缩、传输、显示。

伴随着图像视频数据的海量化，我们急需解决海量数据的超低码率实时压缩问题，尤其在车船监控、犯罪监控等领域。现有压缩方法通常以牺牲细节内容来实现低码率，然而图像视频中重要细节往往对图像视频内容的理解起决定性的作用，比如监控视频中的文字信息等。本课题针对超低码率压缩下的严重细节丢失及重建复杂度较高的问题，研究基于亚像素补偿和视觉感知的新型超低码率压缩算法及应用平台。为增强图像视频中重要细节内容的视觉分辨率，本课题拟充分利用LCD上每个像素由3个独立空间可寻址的亚像素(红绿蓝)组成，通过自由控制组合相邻亚像素，结合人类视觉感知特性，增加在超低码率下的采样、压缩、复原中重建点数目；并通过研究亚像素级别的灰度细节与色度失真的频域特性，设计多通道自适应的超低码率压缩算法，有效的避免丰富细节内容的严重丢失。本课题通过上述研究，将实现一个高效实时的超低码率压缩平台，切实服务于海量高清数据的压缩、传输、显示。