TDI-CMOS成像系统研究及FPN校正方法设计
发布时间：2018-11-02 15:02
【摘要】：传统线阵图像传感器采用一维像素阵列结构,通过扫描运动获取二维图像,其成像性能受限于运动速度和光照强度。时间延迟积分(TDI)图像传感器采用二维像素阵列结构,通过扫描的方式实现不同行像素对同一景物的多次曝光,等效延长曝光时间,从而有效提高图像传感器的信噪比和灵敏度。TDI-CMOS图像传感器具有与CMOS工艺兼容、高集成度、低功耗、抗辐射等优点,逐渐成为研究热点。在TDI-CMOS图像传感器的研制过程中,对传感器的功能测试和性能分析是极其重要的一个环节。因此,关于TDI-CMOS图像传感器成像质量及其成像系统的研究具有重要意义。为了消除TDI-CMOS图像传感器输出图像中的固定模式噪声(FPN),本文首先研究了TDI-CMOS图像传感器的工作原理,然后在分析FPN来源、噪声特点及其对成像质量的影响的基础上,提出了TDI-CMOS图像传感器的噪声模型,并根据该模型设计了一种基于灰度值补偿的FPN校正方法。该方法首先在均匀光下采集大量样本图像,然后基于所有样本图像的行均值向量和列均值向量分别估计行FPN和列FPN,最后,通过像素原始灰度值加上对应的行FPN估计值校正行FPN、通过像素原始灰度值减去对应的列FPN估计值校正列FPN。为了验证提出的校正方法的有效性,首先基于128级模拟域累加TDICMOS图像传感器设计了一套成像系统,研究了系统设计中的关键技术,具体完成了图像传感器硬件电路设计、片上可编程系统(SoPC)设计及履带机械运动装置设计。其中,SoPC是成像系统的嵌入式主控系统,基于FPGA开发板设计。然后使用设计的成像系统采集100幅样本图像,并基于MATLAB软件完成对FPN的估计,最后进行FPN校正。实验结果表明,使用提出的方法对均匀光照下拍摄的图像进行FPN校正后,其行均值标准差从5.6798 LSB减小到了0.4214 LSB,其列均值标准差从15.2080 LSB减小到了13.4623 LSB;而实际测试图像中的行FPN和列FPN也都得到了有效校正。本文设计的方法可以有效消除TDI-CMOS图像传感器输出图像中的FPN。
[Abstract]:The traditional linear image sensor uses one-dimensional pixel array structure to obtain two-dimensional images by scanning motion. Its imaging performance is limited by the speed of motion and the intensity of illumination. The time-delay integral (TDI) image sensor uses two-dimensional pixel array structure to realize multiple exposures of different pixels to the same scene by scanning, which can effectively prolong the exposure time. In order to improve the SNR and sensitivity of the image sensor, TDI-CMOS image sensor has become a hot research area because of its compatibility with CMOS process, high integration, low power consumption, radiation resistance and so on. In the development of TDI-CMOS image sensor, the function test and performance analysis of the sensor is very important. Therefore, it is of great significance to study the imaging quality and imaging system of TDI-CMOS image sensor. In order to eliminate the fixed mode noise in the output image of TDI-CMOS image sensor, this paper first studies the working principle of TDI-CMOS image sensor, then analyzes the source of FPN, the characteristics of noise and its influence on the image quality. The noise model of TDI-CMOS image sensor is proposed and a FPN correction method based on gray value compensation is designed according to the model. Firstly, a large number of sample images are collected under uniform light, and then row FPN and column FPN, are estimated based on row mean vector and column mean vector of all sample images, respectively. Correction row FPN, by pixel original gray value plus corresponding row FPN estimation line FPN, subtractive corresponding column FPN estimate value correction column FPN. In order to verify the validity of the proposed method, an imaging system is designed based on 128-level analog domain accumulative TDICMOS image sensor. The key techniques in the system design are studied, and the hardware circuit of the image sensor is designed. On-chip programmable system (SoPC) design and track mechanical motion device design. Among them, SoPC is the embedded main control system of the imaging system, based on the design of FPGA development board. Then we use the designed imaging system to collect 100 sample images and complete the estimation of FPN based on MATLAB software. Finally, FPN correction is carried out. The experimental results show that the row mean standard deviation decreases from 5.6798 LSB to 0.4214 LSB, and the column mean standard deviation decreases from 15.2080 LSB to 13.4623 LSB; after the proposed method is used for FPN correction of images taken under uniform illumination. The row FPN and column FPN in the actual test image are also corrected effectively. The method designed in this paper can effectively eliminate the FPN. in the output image of TDI-CMOS image sensor.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;TP212
本文编号：2306137
[Abstract]:The traditional linear image sensor uses one-dimensional pixel array structure to obtain two-dimensional images by scanning motion. Its imaging performance is limited by the speed of motion and the intensity of illumination. The time-delay integral (TDI) image sensor uses two-dimensional pixel array structure to realize multiple exposures of different pixels to the same scene by scanning, which can effectively prolong the exposure time. In order to improve the SNR and sensitivity of the image sensor, TDI-CMOS image sensor has become a hot research area because of its compatibility with CMOS process, high integration, low power consumption, radiation resistance and so on. In the development of TDI-CMOS image sensor, the function test and performance analysis of the sensor is very important. Therefore, it is of great significance to study the imaging quality and imaging system of TDI-CMOS image sensor. In order to eliminate the fixed mode noise in the output image of TDI-CMOS image sensor, this paper first studies the working principle of TDI-CMOS image sensor, then analyzes the source of FPN, the characteristics of noise and its influence on the image quality. The noise model of TDI-CMOS image sensor is proposed and a FPN correction method based on gray value compensation is designed according to the model. Firstly, a large number of sample images are collected under uniform light, and then row FPN and column FPN, are estimated based on row mean vector and column mean vector of all sample images, respectively. Correction row FPN, by pixel original gray value plus corresponding row FPN estimation line FPN, subtractive corresponding column FPN estimate value correction column FPN. In order to verify the validity of the proposed method, an imaging system is designed based on 128-level analog domain accumulative TDICMOS image sensor. The key techniques in the system design are studied, and the hardware circuit of the image sensor is designed. On-chip programmable system (SoPC) design and track mechanical motion device design. Among them, SoPC is the embedded main control system of the imaging system, based on the design of FPGA development board. Then we use the designed imaging system to collect 100 sample images and complete the estimation of FPN based on MATLAB software. Finally, FPN correction is carried out. The experimental results show that the row mean standard deviation decreases from 5.6798 LSB to 0.4214 LSB, and the column mean standard deviation decreases from 15.2080 LSB to 13.4623 LSB; after the proposed method is used for FPN correction of images taken under uniform illumination. The row FPN and column FPN in the actual test image are also corrected effectively. The method designed in this paper can effectively eliminate the FPN. in the output image of TDI-CMOS image sensor.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;TP212
【参考文献】
相关期刊论文 前2条
1 徐超;姚素英;徐江涛;李玲霞;;In-Pixel Charge Addition Scheme Applied in Time-Delay Integration CMOS Image Sensors[J];Transactions of Tianjin University;2013年02期
2 桑美贞;徐江涛;聂凯明;姚素英;;TDI型CMOS图像传感器时序控制设计与实现[J];传感技术学报;2011年12期
，本文编号：2306137
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2306137.html
