基于FPGA的自适应夜间图像增强算法研究

发布时间：2018-05-30 12:43

本文选题：FPGA + 亮度评估　；参考：《广东工业大学》2015年硕士论文

【摘要】：随着科技的不断创新与快速发展,视频图像系统作为人类认识与探索的重要信息源头已经遍布各个学科与领域。然而,在弱光或光照不均的环境下图像系统输出的视频信息具有躁点大、对比度低、灰度偏暗和灰度动态范围收缩等缺点。通过引入图像增强技术,能够有效地恢复和完善视频信息。有针对性的图像增强处理,能很大程度上提高视频图像的质量,以便于人眼观察与可利用信息的提取。但是实时的视频图像处理信息量较大,信号转换时序复杂等特点对视频图像系统的硬件要求比较高,且单纯依靠软件难以满足视频图像的实时处理。FPGA即现场可编程门阵列,以并行运算为主,在速度和性能上是MCU与DSP无法比拟的。作为用户可编程的数字集成电路,由于它具有速度快、性能强、稳定性高和成本低等优点,使其在图像处理上应用日益广泛。本文深入研究图像增强算法的理论基础和实际应用,通过对Retinex原理和Gamma校正等图像处理的理论知识作深入理解和探索,并分别对各图像算法和FPGA自身的特点进行系统性的考量和可实行性的分析,进而制定出了一个基于Retinex亮度评估下的自适应Gamma校正的实时夜间图像增强系统。传统的图像增强系统不能同时满足亮度过暗和光照不均等特殊场景下的图像增强。与传统的图像增强系统不一样,本文采用的算法具有自适应能力。针对实际场景中亮度的不同,该算法首先通过Retinex原理进行亮度评估,再根据评估后的亮度以自我调节Gamma参数的方式对图像像素实现高效的Gamma校正增强,最终达到大幅度提高图像可读性的目标。在实现的过程中,由于FPGA在Gamma校正的指数幂运算处理能力上存在一定局限。本文以消耗FPGA的内部资源为代价,通过查找表二次映射的方法有效代替了Gamma指数的仲裁与基于Gamma校正的指数幂运算,从而深度优化图像处理模块的设计,以提高系统的整体运算速度。最后借助基于FPGA的硬件开发平台以实现系统整体的仿真、调试和运行,进一步验证了基于该算法下的实时图像处理系统能有效地提高夜间视频的成像质量。
[Abstract]:With the continuous innovation and rapid development of science and technology, video image system as an important source of information for human understanding and exploration has spread all over various disciplines and fields. However, under the condition of weak light or uneven illumination, the video information outputted by the image system has some disadvantages, such as large point, low contrast, dark gray and shrinking dynamic range of gray. By introducing image enhancement technology, video information can be recovered and improved effectively. Targeted image enhancement can greatly improve the quality of video images, so as to facilitate human eye observation and extraction of available information. However, the real-time video image processing features such as large amount of information, complex signal conversion timing and other characteristics of the video image system hardware requirements are relatively high, and relying solely on software to meet the real-time processing of video images. FPGA, that is, field programmable gate array. Parallel operation is the main method, which is unparalleled in speed and performance between MCU and DSP. As a user programmable digital integrated circuit, it is widely used in image processing due to its high speed, high performance, high stability and low cost. In this paper, the theoretical foundation and practical application of image enhancement algorithm are deeply studied, and the theoretical knowledge of image processing such as Retinex principle and Gamma correction are deeply understood and explored. The characteristics of each image algorithm and FPGA are systematically considered and analyzed, and a real-time night image enhancement system based on adaptive Gamma correction based on Retinex luminance evaluation is developed. The traditional image enhancement system can not satisfy the image enhancement in the special scene of too dark brightness and uneven illumination. Different from the traditional image enhancement system, the algorithm adopted in this paper has adaptive ability. Aiming at the difference of brightness in the actual scene, the algorithm firstly evaluates the brightness by Retinex principle, and then realizes the efficient Gamma correction and enhancement of the image pixels by self-adjusting the Gamma parameters according to the evaluated brightness. Finally, the goal of greatly improving the readability of images is achieved. In the process of implementation, FPGA has some limitations on the processing ability of exponential power operation of Gamma correction. At the cost of consuming the internal resources of FPGA, this paper replaces the arbitration of Gamma exponent and exponential power operation based on Gamma correction effectively by lookup table quadratic mapping, so as to optimize the design of image processing module in depth. In order to improve the overall operation speed of the system. Finally, the hardware development platform based on FPGA is used to realize the simulation, debugging and running of the whole system, which further verifies that the real-time image processing system based on this algorithm can effectively improve the imaging quality of night video.
【学位授予单位】：广东工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN791;TP391.41

【相似文献】