面向高清显示的图像增强技术研究

More and more handheld devices are equipped with large and high definition screens in recent years. But the image/video display results are usually blurred due to their resolution is lower than the screens’ physical resolution. In this project, we aim to develop a kind of rapid image enhancement technology, the details are as follows. Firstly, we develop a kind of piecewise interpolation method with discontinuity points, and give a uniform formula for image interpolation. Secondly, we convert image data to geometric mesh adaptively, and interpolate the image data with newly proposed subdivision based on the piecewise interpolation method. Thirdly, we study the constrained subvision methods with normal vector control, curvature control or local interpolation control. Meanwhile, we apply the constrained subdivion to image interpolation to keep the sharp image edges by means of controlling its normal vectors or curvatures. Lastly, we develop a fast image enhancement system on Android platform. The research result of this project will improve the users’ visual experience in handheld devices and other embedded devices. Hence, it is of realistic meaning.

目前，显示设备呈现大尺寸、高分辨率的特点，但由于多数图像、视频等数字内容的分辨率不能与设备分辨率匹配，常常产生显示模糊不清的问题。本课题拟以几何处理方法为工具，开发一种支持高清显示的快速图像增强技术，主要研究内容为：1、研究含有间断点的分段连续插值方法，并建立图像边缘自适应的图像插值统一算法；2、研究将图像转换为几何网格的自适应算法，提出新的几何网格插值格式，对图像数据进行插值；3、研究对已有细分格式施加法向量、曲率、局部插值等约束条件，构造约束细分格式的方法，并应用于图像放大中保持图像边缘清晰；4、在Android平台上实现图像增强系统。本课题研究的较好完成，将给大量手持设备、嵌入式设备的显示模块提供软件支持，给用户带来更好的视觉体验，具有一定的应用价值。

本项目面对日益普及的手机等手持设备，研究低复杂度的图像/视频清晰度的增强方法。提出了不连续B样条、有理格式细分等几何造型方法，并将这些理论方法应用于图像增强中，弥补了由数字图像/视频分辨率低于设备物理分辨率而产生的边缘模糊、目标虚化等问题，明显加快了运算速度，并在手机、平板、机顶盒等设备上实现该技术。