基于FPGA的高清视频电子稳像算法研究

发布时间：2018-05-03 22:40

本文选题：电子稳像 + 平移运动估计　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：电子稳像依托于数字图像处理技术来实现对视频序列的去抖,以提高画面的质量。该技术使用数字处理器来完成图像的稳定,可以充分利用现代大规模以及超过规模集成电路的优势将稳像系统小型化和低功耗化。由于图像数据量比较大,因此在计算时也就会消耗很多的时间,算法的计算速度和复杂度应该被着重考虑。本文从电子稳像系统中图像运动参数估计精度、算法计算复杂度以及在嵌入式平台现场可编程门阵列(Field-Programmable Gate Array,FPGA)上实现为主要考虑点进行研究。本文首先介绍了电子稳像系统的基本原理,将图像的运动分为平移运动和旋转运动分别进行研究。在图像平移运动参数估计中,本文通过实验分析了代表点匹配法、位平面法、灰度投影法和相位相关法的优缺点,这些方法在参数估计精度、计算时间复杂度和抗干扰性方面无法同时得到很好的结果。本文对块匹配法进行优化,使其保留了参数估计精度高和抗小幅度旋转优势的同时极大降低了运算量。其次,针对图像复合运动参数估计对圆投影进行扩展得到基于模板滤波的圆环投影算法。当图像的运动包含旋转时使其像素相对位置发生了变化,文中选择了具有旋转不变特性的灰度直方图和极坐标投影进行了实验分析,并将圆投影进行了扩展得到基于模板滤波的圆环投影算法。该方法具有旋转不变特性,而且计算时间复杂度也比前两种方法大幅度下降。最后,本文设计并实现了电子稳像硬件系统。硬件系统包含摄像机、图像采集卡和显示器。文中对图像采集卡电路设计和Verilog程序设计进行了详细的介绍,在FPGA上实现了SDI数据流解析、SSRAM读写控制、图像运动参数估计、图像运动参数滤波等功能模块,并验证了各模块的正确性。在实际测试中,该电子稳像硬件系统可以实现对分辨率为1080p,帧率为30fps的视频进行实时稳像。
[Abstract]:Electronic image stabilization is based on digital image processing technology to achieve video sequence shake, in order to improve the quality of the picture. The digital processor is used to stabilize the image, which can make full use of the advantages of modern large-scale and larger integrated circuits to miniaturize and reduce the power consumption of the image stabilization system. Because of the large amount of image data, the computational speed and complexity of the algorithm should be considered. In this paper, the estimation accuracy of image motion parameters, the computational complexity of the algorithm and the implementation on the field programmable gate array (Field-Programmable Gate ArrayFPGA) of an embedded platform are studied. In this paper, the basic principle of electronic image stabilization system is introduced, and the motion of image is divided into translation motion and rotation motion. In this paper, the advantages and disadvantages of the representative point matching method, bit plane method, gray projection method and phase correlation method are analyzed through experiments. The computational time complexity and anti-interference can not get good results at the same time. In this paper, the block matching method is optimized to keep the high precision of parameter estimation and the advantage of anti-rotation of small amplitude, at the same time greatly reducing the computational complexity. Secondly, the circular projection algorithm based on template filter is obtained by extending the circular projection based on image composite motion parameter estimation. When the motion of the image includes rotation, the relative position of the pixel changes. In this paper, the gray histogram and polar projection with rotation invariant property are selected for experimental analysis. The circular projection algorithm based on template filter is obtained by extending the circular projection. The method has the property of rotation invariance, and the computational time complexity is much lower than that of the former two methods. Finally, the hardware system of electronic image stabilization is designed and implemented. The hardware system includes a camera, an image capture card and a display. In this paper, the circuit design and Verilog program design of image acquisition card are introduced in detail. The function modules of SDI data stream parsing and reading and writing control, image motion parameter estimation, image motion parameter filtering and so on are implemented on FPGA. The correctness of each module is verified. In the actual test, the electronic image stabilization hardware system can achieve real-time image stabilization for video with a resolution of 1080p and frame rate of 30fps.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】