全像素双核成像技术及应用研究综述

全像素双核(dual-pixel, DP)自动对焦(dual-pixel CMOS auto focus, DP CMOS AF)采用混合检测自动对焦,其在每个像素配备两个光电二极管,使每个像素既参与对焦又参与成像,克服了传统相位检测自动对焦和反差检测自动对焦技术的缺点。根据离焦视差估计图像合焦镜头所需移动的距离,DP自动对焦具有更快的对焦速度和更高的对焦精度,因此广泛应用在手持设备中(如手机相机、单反相机等)。由于全像素双核传感器将每个像素分成两半,该传感器一次拍摄即可得到两幅图像。这两幅图像(全像素双核图像对)可以看做一个具有相同曝光时间和严格校正的小基线立体图像对。该图像对的视差与图像模糊程度相对应,只在离焦区域存在视差。全像素双核传感器不仅用于自动对焦,而且可以用于深度估计、散焦去模糊和反射去除等方面。本文系统地综述了全像素双核传感器的自动对焦、成像原理及研究现状,并进一步展望其未来发展。1)对自动对焦技术进行介绍,对比了传统对焦与全像素双核对焦;2)详细分析了全像素双核传感器的成像原理、成像模型及特点;3)系统地介绍了全像素双核在计算机视觉领域应用的最新进展,从深度估计、反射去除和离焦模糊去除等方面进行全面阐述及分析; 4)适当的数据集是基于深度学习方法的基础,对目前的全像素双核数据集进行了介绍;5)分析了全像素双核在计算机视觉领域面临的挑战与机遇,对未来的全像素双核应用方向进行了讨论与展望。     

一种基于最小熵准则的新的自聚焦优化算法

最小熵方法是合成孔径雷达自聚焦算法的一种,但此方法搜索最优多普勒调频率时需要在整个设置区间搜索,计算量较大。基于最小熵准则,利用图像熵与多普勒调频率的关系,提出了一种新的自聚焦优化算法。该方法利用二分法在整个设置区间能快速搜索到最优多普勒调频率,从而大幅度地降低了运算量。实测数据处理结果证明了此方法的有效性。     

提高显微镜自动对焦范围的频谱分析法的研究

提出了一种基于CCD(电荷耦合器件)模拟视频信号分析法的显微镜自动对焦改进方法.许多基于视频信号分析法的自动对焦系统仅考虑视频信号的高频分量,忽略低频分量在对焦过程中的变化.本文对模拟视频信号进行频谱分析,研究不同频率范围的信号对对焦评价曲线的影响,得出适合自动调焦的粗调/细调的频域范围,使系统有更广调焦范围的同时保持较高精度.最终通过实验验证,加入适当的滤波电路后调焦范围更广,40倍物镜下显微镜自动对焦系统的有效调焦范围可达±500μm.     

Autofocusing for tissue microscopy

This paper describes the implementation of autofocusing for tissue microscopy. We first investigate the suitability of several criterion functions for the evaluation of image sharpness. Since tissue sections are invariably stained, we also discuss the selection of the colour component on which autofocusing will be performed. In tissue microscopy, where a section generally comprises multiple layers, it is often not possible to obtain an image that is well focused over the field of view because of the limited depth of field of the objective. We describe focus enhancement algorithms, closely related to the autofocus system, which may be employed to obtain an entirely sharp image.     

A fast autofocus sharpness function of microvision system based on the Robert function and Gauss fitting

For the microvision system, a new autofocus evaluation function based on the Robert function is proposed by increasing the threshold value. Compared with the traditional evaluation function, the new focus function reduces the local extreme value and increases the steepness of the focusing curve. According to the characteristics of the focusing evaluation function, the focus curve can be divided into two stages: the gentle area and the steep area. In the gentle area, there will be set a large step‐length to realize the fast search. In the steep area, the data will be fitted by Gauss method, and on the basis of the fitting results, the motor of microvision system was directly driven to achieve the focal plane and this method has been improved in real‐time and accuracy.     

星载合成孔径雷达二维闪烁相位误差校正方法研究

分析了电离层闪烁效应对低频段星载合成孔径雷达(Synthetic Aperture Radar, SAR)方位向分辨率的影响, 并基于二维相位屏方法生成了距离向空变的闪烁相位误差.传统的相位梯度自聚焦(Phase Gradient Autofocus, PGA)方法难以适用于二维空变相位误差校正, 而本文基于闪烁相位误差在距离向具有连续性的特点, 在一定规则基础上将整幅图像划分为若干个子孔径, 对每个子孔径图像利用PGA方法进行相位误差估计, 再将得到的各子孔径相位误差进行插值运算, 从而得到整幅图像的闪烁相位误差.仿真结果表明:相比于传统PGA方法, 子孔径PGA方法可以有效解决二维空变闪烁相位误差对图像方位向分辨率的影响, 校正后的图像得到很好的恢复, 方位向分辨率明显提高.     

合成孔径雷达（SAR）的最新发展

全文分若干单元分别论述了机载2D-SAR自聚焦技术，降低斑点技术，MTI-SAR和3D-SAR（含机载和空载），对上述各种SAR分别讨论了工作原理，列出了多个实例以及应用情况。     

A novel autofocus algorithm based on maximum total variation criteria for SAR images

A novel autofocus algorithm for synthetic aperture radar (SAR) based on total variation is presented in this paper. The method, which starts with a complex phasedegraded SAR image, after the phase errors model is introduced into the range-compressed phase-history domain, carries out phase errors correction by changing the focus till the total variation of the azimuth profile is maximized. Compared with the minimum entropy autofocus algorithm, the autofocus algorithm has less computational complexity and is easier to implement. The simulation and the processing results of the measured data show the validity of the proposed method. Translated from Journal of Xidian University, 2006, 33(2): 167–172 [译自: 西安电子科技大学学报]     

基于二代小波算法的数字全息显微成像自动对焦方法

针对传统数字全息显微重建方法难以获取最佳重建距离,且依赖于传统PC平台难以便携化等问题,设计了一种嵌入式数字全息显微自动对焦系统。该系统不仅能方便的完成数字全息显微成像的自动对焦,获取最佳重建效果,而且具有便携化的特点。该系统采用同轴全息光路配合图像传感器获取全息图,利用优化后的角谱算法实现全息重建,再结合二代小波算法对重建结果进行图像质量评价,根据评价结果找到最佳重建距离,同时获取最佳重建结果,完成自动对焦。利用该系统采集一批分辨率板全息图进行自动对焦测试。实验表明,系统的重建速度为0.6s,分辨率为3.48um,自动对焦准确率达到90%以上,实验结果验证了系统的有效性。     

The Weibull manifold in low-level image processing: An application to automatic image focusing

In this paper, we introduce a novel framework for low-level image processing and analysis. First, we process images with very simple, difference-based filter functions. Second, we fit the 2-parameter Weibull distribution to the filtered output. This maps each image to the 2D Weibull manifold. Third, we exploit the information geometry of this manifold and solve low-level image processing tasks as minimisation problems on point sets. For a proof-of-concept example, we examine the image autofocusing task. We propose appropriate cost functions together with a simple implicitly-constrained manifold optimisation algorithm and show that our framework compares very favourably against common autofocus methods from literature. In particular, our approach exhibits the best overall performance in terms of combined speed and accuracy.