CN103595923A - Double-camera high-speed imaging system and method based on code exposure - Google Patents

Abstract

The invention provides a double-camera high-speed imaging system and method based on code exposure. The double-camera high-speed imaging system is mainly composed of a high-speed camera with low spatial resolution, a code exposure camera with high resolution, and a beam splitting system. The beam splitting system can be achieved by adopting a half-reflection semi-permeable lens, the high-speed camera and the code exposure camera are placed at positions, equally spaced from the lens, in a reflective light path and an incident light path respectively, scenery light is distributed to the two cameras with the same view field and the same angle, and the scenes shot by the two cameras are the same. According to the double-camera high-speed imaging system and method based on code exposure, by utilizing the beam splitting system and external triggering signals, the two cameras can shoot the same scene synchronously for imaging, and scene shooting with a higher frame rate and more pictures can be achieved more easily.

Description

Translated fromChinese

一种基于曝光编码的双机高速成像系统及方法A dual-camera high-speed imaging system and method based on exposure coding

技术领域technical field

该发明涉及一种基于曝光编码的双机高速成像系统及方法。该系统及方法利用一台低分辨率的高速相机和一台高分辨率的曝光编码相机，通过图像处理和计算成像技术实现高速成像。The invention relates to a two-camera high-speed imaging system and method based on exposure coding. The system and method utilize a low-resolution high-speed camera and a high-resolution exposure coding camera to realize high-speed imaging through image processing and computational imaging technology.

背景技术Background technique

高速成像技术是实验物理研究过程中的一种重要的观察和测量手段。许多快速物理过程的精确观测，要求高速成像系统在具备高时间分辨的同时，能够具有较高的空间分辨率。现有的高速成像技术主要有转镜式成像系统、高速CCD/CMOS成像系统、多分幅成像系统、分立器件阵列成像系统等。转镜式成像系统采用机械式转镜结构，通过将景物光快速投影到成像胶片上，实现高速成像，此类系统目前可实现数十万帧每秒的成像速度；但由于成像介质采用传统的胶片，导致图像的获取周期较长，后期图像数据的处理较困难。多分幅成像系统基于光学分光系统，将景物光分别投射到多个成像设备，利用特殊的高速电子快门实现多个成像设备的超高速成像；但由于多路光学分光结构较复杂，且对景物光具有衰减作用，此类系统通常很难实现较多画幅数的成像，系统造价昂贵。基于分立器件阵列的成像系统具有极高的时间响应，但是由于集成度较低导致系统的空间分辨率不高，且由于系统规模较大，造价昂贵。基于高速CCD/CMOS的成像系统是目前发展最快的一类高速成像系统，尤其随着CCD/CMOS等图像传感器技术的不断发展，此类系统的成像速度和空间分辨能力不断提高。但是此类系统的成像速度和空间分辨率是一对相互制约的性能指标，在提高成像速度时，不可避免的要牺牲空间分辨率，而在获得全分辨率时，其成像速度则会受到制约。在对空间分辨率和成像速度都有较高要求的应用中，需要采取创新的成像方法。High-speed imaging technology is an important means of observation and measurement in the process of experimental physics research. The precise observation of many fast physical processes requires a high-speed imaging system to have high spatial resolution while having high time resolution. The existing high-speed imaging technologies mainly include rotating mirror imaging systems, high-speed CCD/CMOS imaging systems, multi-framing imaging systems, and discrete device array imaging systems. The rotating mirror imaging system adopts a mechanical rotating mirror structure to quickly project the scene light onto the imaging film to achieve high-speed imaging. This type of system can currently achieve an imaging speed of hundreds of thousands of frames per second; however, due to the imaging medium using traditional Film, resulting in a longer image acquisition cycle and more difficult post-processing of image data. The multi-framing imaging system is based on the optical beam splitting system, which projects the scene light to multiple imaging devices respectively, and uses a special high-speed electronic shutter to realize ultra-high-speed imaging of multiple imaging devices; With attenuation, such systems are usually difficult to achieve imaging with a large number of frames, and the system is expensive. Imaging systems based on discrete device arrays have extremely high time response, but the spatial resolution of the system is not high due to low integration, and the system is expensive due to its large scale. Imaging systems based on high-speed CCD/CMOS are currently the fastest-growing type of high-speed imaging systems. Especially with the continuous development of image sensor technologies such as CCD/CMOS, the imaging speed and spatial resolution of such systems continue to increase. However, the imaging speed and spatial resolution of this type of system are a pair of mutually restrictive performance indicators. When the imaging speed is increased, the spatial resolution is inevitably sacrificed, and when the full resolution is obtained, the imaging speed will be restricted. . In applications where high spatial resolution and imaging speed are required, innovative imaging methods are required.

随着计算技术的不断进步，传统的成像方法正在不断得到改进。通过在成像系统中引入预先设定的编码信息，并利用这些信息对所拍摄的编码图像进行解码，可以获得更加丰富的图像信息。利用计算技术提升成像系统性能和功能的计算成像技术正在飞速发展，并不断推动着现代摄影技术的革新和进步。本发明利用曝光编码技术结合传统高速成像技术，实现高空间分辨率的高速成像。Traditional imaging methods are being continuously improved with advances in computing technology. By introducing preset coding information into the imaging system and using the information to decode the captured coded images, more abundant image information can be obtained. Computational imaging technology, which uses computing technology to improve the performance and functions of imaging systems, is developing rapidly, and continues to promote the innovation and progress of modern photography technology. The invention utilizes the exposure encoding technology combined with the traditional high-speed imaging technology to realize high-speed imaging with high spatial resolution.

发明内容Contents of the invention

本发明旨在解决成像系统的高帧频与高分辨率相互制约的矛盾，利用曝光编码技术结合高速成像技术，实现高分辨率的双机高速成像系统。The invention aims to solve the contradiction between high frame frequency and high resolution of the imaging system, and realizes a dual-machine high-speed imaging system with high resolution by combining exposure encoding technology with high-speed imaging technology.

本发明的技术解决方案为：Technical solution of the present invention is:

一种基于曝光编码的双机高速成像系统，其特殊之处在于：包括分光系统1、高空间分辨率相机3、低空间分辨率相机2、参数设置及同步控制单元5、解码单元4；A two-machine high-speed imaging system based on exposure coding, which is special in that it includes aspectroscopic system 1, a highspatial resolution camera 3, a lowspatial resolution camera 2, a parameter setting andsynchronization control unit 5, and adecoding unit 4;

所述分光系统将景物光以相同的视场大小和角度分配给高空间分辨率相机和低空间分辨率相机；The spectroscopic system distributes the scene light to the high spatial resolution camera and the low spatial resolution camera with the same field of view size and angle;

所述高空间分辨率相机采用编码曝光；The high spatial resolution camera adopts coded exposure;

所述低空间分辨率相机具有高速成像能力；The low spatial resolution camera has high-speed imaging capability;

所述参数设置及同步控制单元用于高空间分辨率相机和低空间分辨率相机的光通量和光强度的占比分配、触发控制，高空间分辨率相机的参数设置，低空间分辨率相机的参数设置；The parameter setting and synchronous control unit are used for the proportion distribution and trigger control of the luminous flux and light intensity of the high spatial resolution camera and the low spatial resolution camera, the parameter setting of the high spatial resolution camera, and the parameter setting of the low spatial resolution camera ;

所述解码单元用于从低空间分辨率相机拍摄的图像中提取景物图像点扩散函数，以及利用点扩散函数对高空间分辨率相机的曝光编码图像进行解码还原。The decoding unit is used to extract the point spread function of the scene image from the image taken by the low spatial resolution camera, and use the point spread function to decode and restore the exposure coded image of the high spatial resolution camera.

上述高空间分辨率相机的参数设置包括根据所拍摄景物对相机时间分辨能力的要求设置曝光时间片长度T；根据所拍摄景物的变化过程持续时间或者说拍摄者所关心的变化持续时间设置曝光全过程的时间片个数N；根据不同拍摄对象场景变化的时间域特征设置曝光二进制编码B₁B₂B₃…B_N；The parameter setting of the above-mentioned high spatial resolution camera includes setting the exposure time slice length T according to the requirements of the photographed scene on the camera’s time resolution capability; The number of time slices in the process is N; according to the time domain characteristics of different shooting object scenes, the exposure binary code B₁ B₂ B₃ ... B_N is set;

所述编码曝光包括以下步骤：The coded exposure comprises the following steps:

1）编码曝光阶段：电子快门的曝光控制信号受曝光编码的控制，若时间片TI对应的编码BI为1，则在该时间片内，电子快门打开，像素单元光敏区接受景物光并产生相应的感应电荷；若BI为0，则该时间片，电子快门不打开，等待一个时间片长度后进行下一时间片的曝光操作；1) Coded exposure stage: The exposure control signal of the electronic shutter is controlled by the exposure code. If the code BI corresponding to the time slice TI is 1, then within the time slice, the electronic shutter is opened, and the photosensitive area of the pixel unit receives the scene light and generates a corresponding The induced charge; if BI is 0, the electronic shutter is not opened for this time slice, and the exposure operation of the next time slice is performed after waiting for a time slice length;

2）曝光转移阶段：进入下一时间片的曝光，对图像传感器的像素光敏区重复上述的步骤1的曝光过程，若上一时间片有曝光，则进行光敏区感应电荷向存储区的转移过程；如此反复，经过N个时间片曝光过程后，图像传感器经过一次完整的编码曝光，N个时间片所获得的图像信息累积存储在像素存储区中；2) Exposure transfer stage: Enter the exposure of the next time slice, repeat the exposure process ofstep 1 above for the photosensitive area of the pixel of the image sensor, if there is exposure in the previous time slice, the transfer process of the induced charge in the photosensitive area to the storage area is carried out ; Repeatedly, after N time slices exposure process, the image sensor undergoes a complete encoding exposure, and the image information obtained by N time slices is accumulated and stored in the pixel storage area;

3）将图像传感器中存储的图像数据输出，获得一副曝光编码图像；3) Output the image data stored in the image sensor to obtain an exposure coded image;

所述解码还原包括以下步骤：The decoding restore comprises the following steps:

1）首先根据景物图像的点扩散函数和时间片信息进行点扩散函数实例化，获得对应时间片的反卷积矩阵；1) First, instantiate the point spread function according to the point spread function and time slice information of the scene image, and obtain the deconvolution matrix corresponding to the time slice;

2）使用时间片对应的反卷积矩阵对曝光编码图像进行反卷积运算，得到对应时间片的过程图像；2) Use the deconvolution matrix corresponding to the time slice to deconvolute the exposure coded image to obtain the process image corresponding to the time slice;

3）对N个时间片重复步骤1和2，即得到每个时间片对应的过程图像。3) Repeat steps 1 and 2 for N time slices to obtain the process image corresponding to each time slice.

上述低空间分辨率相机的参数设置包括低分辨率高速相机三次曝光成像时刻T_S、T_M、T_E的设置。The parameter setting of the above-mentioned low spatial resolution camera includes the setting of the three exposure imaging moments T_S ,_TM , and_TE of the low resolution high-speed camera.

一种基于曝光编码的双机高速成像方法，其特殊之处在于：包括如下步骤：A dual-camera high-speed imaging method based on exposure coding, which is special in that it includes the following steps:

1）参数设置；1) Parameter setting;

其中，高空间分辨率相机参数包括根据所拍摄景物对相机时间分辨能力的要求设置曝光时间片长度T、根据所拍摄景物的变化过程持续时间或者说拍摄者所关心的变化持续时间设置曝光全过程的时间片个数N、根据不同拍摄对象场景变化的时间域特征设置曝光二进制编码B₁B₂B₃…B_N；Among them, the high spatial resolution camera parameters include setting the exposure time slice length T according to the requirements of the captured scene on the camera's time resolution capability, and setting the entire exposure process according to the duration of the change process of the captured scene or the duration of the change that the photographer is concerned about. The number of time slices N, setting the exposure binary code B₁ B₂ B₃ ... B_N according to the time domain characteristics of different shooting object scenes;

低空间分辨率相机参数包括景物快速变化的开始阶段、过程中以及接近结束阶段所对应的三次曝光成像时刻T_S、T_M、T_E；Low spatial resolution camera parameters include the three exposure imaging moments T_S ,_TM , and_TE corresponding to the beginning stage, process, and near-end stage of rapid scene changes;

分光系统对高空间分辨率相机和低空间分辨率相机的光通量和光强度的占比分配；The distribution of the light flux and light intensity of the high spatial resolution camera and the low spatial resolution camera by the spectroscopic system;

2）两台相机根据步骤1设置的拍摄条件同步进行拍摄：2) The two cameras shoot synchronously according to the shooting conditions set in step 1:

2.1)高空间分辨率相机的曝光、转移及输出：2.1) Exposure, transfer and output of high spatial resolution camera:

2.1.1)编码曝光阶段：电子快门的曝光控制信号受曝光编码的控制，若时间片T_I对应的编码B_I为1，则在该时间片内，电子快门打开，像素单元光敏区接受景物光并产生相应的感应电荷；若B_I为0，则该时间片，电子快门不打开，等待一个时间片长度后进行下一时间片的曝光操作；2.1.1) Coded exposure stage: the exposure control signal of the electronic shutter is controlled by the exposure code, if the code B_I corresponding to the time slice T_I is 1, then within the time slice, the electronic shutter is opened, and the photosensitive area of the pixel unit receives the scene Light and generate corresponding induced charges; if B_I is 0, then the time slice, the electronic shutter is not opened, and the exposure operation of the next time slice is performed after waiting for a time slice length;

2.1.2)曝光转移阶段：进入下一时间片的曝光，对图像传感器的像素光敏区重复上述的步骤2.1.1的曝光过程，若上一时间片有曝光，则进行光敏区感应电荷向存储区的转移过程；如此反复，经过N个时间片曝光过程后，图像传感器经过一次完整的编码曝光，N个时间片所获得的图像信息累积存储在像素存储区中；2.1.2) Exposure transfer stage: Enter the exposure of the next time slice, repeat the exposure process of the above step 2.1.1 for the photosensitive area of the pixel of the image sensor, if there is exposure in the previous time slice, the photosensitive area will be stored in the induced charge direction The transfer process of the area; so repeated, after N time slice exposure process, the image sensor undergoes a complete encoding exposure, and the image information obtained by N time slices is accumulated and stored in the pixel storage area;

2.1.3)将图像传感器中存储的图像数据输出，获得一副曝光编码图像；2.1.3) Outputting the image data stored in the image sensor to obtain an exposure coded image;

2.2)低空间分辨率相机的曝光、转移及输出：2.2) Exposure, transfer and output of low spatial resolution cameras:

低空间分辨率相机在设定好的T_S、T_M、T_E时刻，对景物光分别曝光一个时间片T长度，生成三幅低分辨率的高速图像并输出；即开始帧FrameS、过程帧FrameM和结束帧FrameE；The low-spatial-resolution camera exposes the scene light for a time slice T at the set time T_S , T_M , and T_E , and generates three low-resolution high-speed images and outputs them; that is, the start frame FrameS and the process frame FrameM and end frame FrameE;

3）解码还原：3) Decoding restore:

3.1)获取点扩散函数：3.1) Obtain the point spread function:

3.1.1)对开始帧、过程帧和结束帧图像，进行不同灰度阈值条件下的图像轮廓提取，或者进行灰度梯度计算得到图像的灰度梯度图；3.1.1) For the start frame, process frame and end frame images, perform image contour extraction under different gray threshold conditions, or perform gray gradient calculation to obtain the gray gradient map of the image;

3.1.2)根据轮廓提取结果和灰度梯度信息，对开始帧与过程帧、过程帧与结束帧分别进行模型匹配，获取景物图像不同灰度阈值条件下，在图像中的位置或扩散信息；3.1.2) According to the contour extraction result and the gray gradient information, model matching is performed on the start frame and the process frame, and the process frame and the end frame respectively, and the position or diffusion information in the image is obtained under the condition of different gray threshold values of the scene image;

3.1.3)根据三帧图像拍摄的时间片信息T_S、T_M、T_E以及景物在不同图像帧中的位置和扩散信息，分别计算出景物图像在开始帧、过程帧与结束帧三个时间片之间的两个点扩散函数PSF₁、PSF₂；3.1.3) According to the time slice information T_S ,_TM ,_TE of the three frames of images and the position and diffusion information of the scene in different image frames, calculate the three frames of the scene image in the start frame, process frame and end frame respectively. Two point spread functions PSF₁ , PSF₂ between time slices;

3.2)解码还原：3.2) Decoding restoration:

3.2.1)首先根据景物图像的点扩散函数和时间片信息进行点扩散函数实例化，获得对应时间片的反卷积矩阵；3.2.1) First, instantiate the point spread function according to the point spread function and time slice information of the scene image, and obtain the deconvolution matrix corresponding to the time slice;

3.2.2)使用时间片对应的反卷积矩阵对曝光编码图像进行反卷积运算，得到对应时间片的景物图像；3.2.2) Use the deconvolution matrix corresponding to the time slice to perform deconvolution operation on the exposure coded image to obtain the scene image corresponding to the time slice;

3.2.3)对N个时间片重复步骤3.2.1和3.2.2，即得到每个时间片对应的景物图像。3.2.3) Repeat steps 3.2.1 and 3.2.2 for N time slices to obtain the scene image corresponding to each time slice.

上述对时间片信息进行点扩散函数实例化和反卷积运算的具体步骤如下：The specific steps of instantiating the point spread function and deconvolution operation on the time slice information are as follows:

1)对时间片T_i从T₁到T_N迭代使用本算法1) Use this algorithm iteratively for the time slice T_i from T₁ to T_N

2)若时间片T_i位于T_S和T_M之间，则采用PSF₁对该时间片的点扩散函数进行实例化，得到PSF_i；利用PSF_i对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；i=i+1；2) If the time slice T_i is between_TS and_TM , use PSF₁ to instantiate the point spread function of the time slice to obtain PSF_i ; use PSF_i to deconvolute the exposure coded image to obtain restore image Frame_i ; i=i+1;

3)若时间片T_i位于T_M和T_E之间，则采用PSF₂对该时间片的点扩散函数进行实例化，得到PSF_i；利用PSF_i对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；i=i+1；3) If the time slice T_i is located between_TM and_TE , use PSF₂ to instantiate the point spread function of the time slice to obtain PSF_i ; use PSF_i to deconvolute the exposure coded image to obtain restore image Frame_i ; i=i+1;

4)若时间片T_i位于T₁和T_S之间，采用PSF1进行点扩散函数实例化；利用PSF_i对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；若i=S-1，对恢复图像与开始帧图像FrameS进行抽样误差分析，若误差大于误差容限W_E，则取i=i+1，重新实例化点扩散函数并对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；若i<S-1，则对恢复图像Frame_i与上一帧图像Frame_i+1图像进行误差分析，若误差大于误差容限W_E，则取i+1，重新实例化点扩散函数并对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；i=i-1；4) If the time slice T_i is between T₁ and T_S , use PSF1 to instantiate the point spread function; use PSF_i to deconvolute the exposure coded image to obtain the restored image Frame_i ; if i=S-1 , analyze the sampling error between the restored image and the start frame image FrameS, if the error is greater than the error tolerance W_E , then take i=i+1, re-instantiate the point spread function and deconvolve the exposure coded image to obtain the restored Image Frame_i ; if i<S-1, then analyze the error between the restored image Frame_i and the previous frame image Frame_i+1 , if the error is greater than the error tolerance W_E , take i+1, and re-instantiate the point Diffusion function and perform deconvolution operation on the exposure coded image to obtain the restored image Frame_i ; i=i-1;

5)若时间片T_i位于T_E和T_N之间，采用PSF2进行点扩散函数实例化；利用PSF_i对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；若i=E+1，对恢复图像与开始帧图像FrameE进行抽样误差分析，若误差大于误差容限W_E，则取i=i-1，重新实例化点扩散函数并对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；若i>E+1，则对恢复图像Frame_i与上一帧图像Frame_i-1图像进行误差分析，若误差大于误差容限W_E，则取i-1，重新实例化点扩散函数并对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；i=i+1。5) If the time slice T_i is between T_E and T_N , use PSF2 to instantiate the point spread function; use PSF_i to deconvolute the exposure coded image to obtain the restored image Frame_i ; if i=E+1 , analyze the sampling error between the restored image and the starting frame image FrameE, if the error is greater than the error tolerance W_E , then take i=i-1, re-instantiate the point spread function and deconvolve the exposure coded image to obtain the restored Image Frame_i ; if i>E+1, analyze the error between the restored image Frame_i and the previous frame image Frame_i-1 , if the error is greater than the error tolerance W_E , take i-1, and re-instantiate the point Diffusion function and perform deconvolution operation on the exposure coded image to obtain the restored image Frame_i ; i=i+1.

上述误差分析是以两幅图像之间的像素值均方差为依据。The above error analysis is based on the mean square error of the pixel values between the two images.

上述抽样误差分析具体为：对图像Frame_i与图像Framej进行抽样误差分析，首先对高分辨率的图像进行低分辨率抽样，抽样采取像素值抽样或者像素值求平均的方法；然后对抽样后的图像采用误差分析。The above-mentioned sampling error analysis is specifically as follows: the sampling error analysis is performed on the image Frame_i and the image Framej, first, low-resolution sampling is performed on the high-resolution image, and sampling adopts the method of pixel value sampling or pixel value averaging; Images were analyzed using error analysis.

本发明所具有的有益效果：The beneficial effects that the present invention has:

1、利用分光系统和外部触发信号，两台相机可以同步的对同一景物进行拍摄成像，有利于实现更高帧频率、更多画幅数的景物拍摄。1. Using the spectroscopic system and external trigger signal, the two cameras can simultaneously shoot and image the same scene, which is conducive to achieving higher frame frequency and more frames.

2、低分辨率的高速相机分别在拍摄过程的开始、中间和结束时，拍摄三副曝光时间长度为T的图像，可用于对分辨率要求不高、画幅数不多的高速景物拍摄。2. A low-resolution high-speed camera shoots three images with an exposure time length of T at the beginning, middle and end of the shooting process, which can be used for high-speed scene shooting with low resolution requirements and a small number of frames.

3、高分辨率的曝光编码相机，可在景物拍摄的过程中，以时间片为单位进行全过程的编码曝光，便于捕捉景物快速变化的光场信息，为软件恢复清晰景物图像提供信息基础。3. The high-resolution exposure coded camera can perform coded exposure of the whole process in units of time slices during the shooting of the scene, which is convenient for capturing the rapidly changing light field information of the scene, and provides the information basis for the software to restore clear scene images.

4、基于三副低分辨率的景物图像，有利于软件算法计算出较精确的景物图像点扩散函数，为曝光编码图像的解码恢复提供必要条件。4. Based on three sets of low-resolution scene images, it is beneficial for the software algorithm to calculate a more accurate point spread function of the scene image, which provides necessary conditions for the decoding and restoration of the exposure coded image.

5、成像系统的性能即时间片T的长度取决于图像传感器的电子快门速度通常在微秒量级，而不受限于传感器的数据输出带宽，因此可实现极高的拍摄速度和帧频率，基于带电子快门的科学级CCD，可实现微秒级高速成像。5. The performance of the imaging system, that is, the length of the time slice T depends on the electronic shutter speed of the image sensor, which is usually on the order of microseconds, and is not limited by the data output bandwidth of the sensor, so it can achieve extremely high shooting speed and frame frequency. Based on scientific-grade CCD with electronic shutter, it can realize microsecond-level high-speed imaging.

6、利用高像素的曝光编码相机，通过软件算法的解码恢复，可实现高分辨率的超高速成像，突破传统成像系统中提高帧频率必然带来分辨率降低的技术难题。6. Utilizing a high-pixel exposure coded camera and decoding and restoring software algorithms, high-resolution ultra-high-speed imaging can be achieved, breaking through the technical problem of resolution reduction that is inevitably caused by increasing the frame rate in traditional imaging systems.

附图说明Description of drawings

图1是双机高速成像系统结构示意图；Figure 1 is a schematic diagram of the structure of a dual-machine high-speed imaging system;

图2是曝光编码成像过程示意图；Fig. 2 is a schematic diagram of exposure coding imaging process;

图3是基于曝光编码的双机高速成像拍摄过程示意图；Fig. 3 is a schematic diagram of the shooting process of dual-camera high-speed imaging based on exposure coding;

图4是曝光编码图像解码恢复过程示意图；Fig. 4 is a schematic diagram of the process of decoding and restoring the exposure coded image;

具体实施方式Detailed ways

本发明方法提出了一种新颖的双机高速成像方法，该方法所提成像系统结构如图1所示，它的特征具体可描述为：The inventive method proposes a novel two-machine high-speed imaging method, the proposed imaging system structure of the method is as shown in Figure 1, and its features can be specifically described as:

1]成像系统主要由一台低空间分辨率的高速相机、一台高分辨率的曝光编码相机和分光系统构成。分光系统可以采用半反半透透镜实现，在反射光和投射光路径上距离透镜距离相同的地方分别放置高速相机和曝光编码相机，景物光将以相同的视场大小和角度分配给两台相机，使得两台相机拍摄的景物相同；1] The imaging system is mainly composed of a high-speed camera with low spatial resolution, a high-resolution exposure coding camera and a spectroscopic system. The spectroscopic system can be realized by using a semi-reflective and semi-transparent lens. A high-speed camera and an exposure coding camera are respectively placed at the same distance from the lens on the reflected light and projected light paths. The scene light will be distributed to the two cameras with the same field of view and angle. , so that the scenes captured by the two cameras are the same;

2]两台相机接受相同的外部触发信号，同步对景物进行拍摄成像；2] The two cameras receive the same external trigger signal, and simultaneously shoot and image the scene;

3]成像系统将一次景物拍摄的全过程划分成N个等分的时间片，时间片长度T决定系统的帧频率和成像速度，该时间片长度取决于运动场景的拍摄时间精度要求或受限于图像传感器电子快门速度；3] The imaging system divides the whole process of shooting a scene into N equal time slices. The length of the time slice T determines the frame frequency and imaging speed of the system. The length of the time slice depends on the shooting time accuracy requirements or limitations of the moving scene. Based on the electronic shutter speed of the image sensor;

4]低空间分辨率的高速相机在景物拍摄过程中，分别在拍摄开始、过程中和拍摄结束时，拍摄三幅曝光时间长度为T的低分辨率图像，即开始帧FrameS、过程帧FrameM和结束帧FrameE；4] A high-speed camera with low spatial resolution shoots three low-resolution images with an exposure time length of T at the start, process and end of the scene shooting, namely the start frame FrameS, process frame FrameM and end frame FrameE;

5]高分辨率的曝光编码相机则对景物拍摄的全过程进行曝光编码成像，曝光编码的时间精度为T，生成一副包含拍摄全过程光学信息的曝光编码图像；5] The high-resolution exposure coding camera performs exposure coding imaging on the whole process of shooting the scene, and the time precision of the exposure coding is T, and generates an exposure coding image containing the optical information of the whole shooting process;

6]最后由软件算法利用低分辨率图像从曝光编码图像中恢复出多幅清晰的高分辨率景物过程图像，算法主要有两个功能，即图像点扩散函数的提取和曝光编码图像的解码恢复，算法的主要步骤描述如下：6] Finally, the software algorithm uses the low-resolution image to restore multiple clear high-resolution scene process images from the exposure coded image. The algorithm mainly has two functions, namely, the extraction of the image point spread function and the decoding and restoration of the exposure coded image , the main steps of the algorithm are described as follows:

6.1 对三幅低分辨率高速图像进行轮廓提取和灰度梯度计算；6.1 Perform contour extraction and gray gradient calculation on three low-resolution high-speed images;

6.2 分别对开始帧与过程帧、过程帧与结束帧计算景物图像的点扩散函数；6.2 Calculate the point spread function of the scene image for the start frame and process frame, process frame and end frame respectively;

6.3 根据时间片取值不同，分别实例化各个时间片的点扩散函数对应的反卷积矩阵；6.3 According to the different values of the time slice, instantiate the deconvolution matrix corresponding to the point spread function of each time slice;

6.4 对曝光编码图像进行反卷积计算，得到高分辨率的恢复图像；6.4 Perform deconvolution calculation on the exposure coded image to obtain a high-resolution restored image;

6.5 对恢复图像进行误差分析并进行迭代处理。6.5 Perform error analysis and iterative processing on the restored image.

通过分光系统，两台相机对所拍摄景物的视场大小和角度相同，即两台相机“所见”相同。Through the spectroscopic system, the two cameras have the same field of view and angle of the captured scene, that is, the two cameras "see" the same.

一次景物拍摄的全过程被成像系统等分成时间长度为T的N个时间片T₁、T₂、T₃…T_N，时间长度T代表单幅图像的最短曝光时间长度，其倒数1/T代表成像系统的帧频率。The whole process of a scene shooting is equally divided into N time slices T₁ , T₂ , T₃ ... T_N by the imaging system. The time length T represents the shortest exposure time length of a single image, and its reciprocal 1/T Represents the frame frequency of the imaging system.

低分辨率的高速相机在拍摄过程的T_S、T_M、T_E的三个时间片对景物进行高速成像，生成三幅低分辨率的高速图像，其中S、M、E为介于1和N之间的整数。The low-resolution high-speed camera performs high-speed imaging on the scene in the three time slices of T_S ,_TM , and_TE during the shooting process, and generates three low-resolution high-speed images, where S, M, and E are between 1 and An integer between N.

曝光编码相机可以对拍摄过程的N个时间片T₁、T₂、T₃…T_N，根据给定的N位二进制编码b₁b₂b₃…b_N进行编码曝光，即二进制码第i位即b_i为0时，T_i时间片不曝光，若第i位为1，则T_i时间片曝光，其曝光过程的示意图如图2所示。Exposure encoding The camera can encode and expose N time slices T₁ , T₂ , T₃ ... T_N in the shooting process according to the given N-bit binary code b₁ b₂ b₃ ... b_N , that is, the i-th binary code When the bit_i is 0, the T_i time slice is not exposed, and if the i-th bit is 1, the T_i time slice is exposed. The schematic diagram of the exposure process is shown in Figure 2.

软件算法利用三副低分辨率图像计算出景物图象较精确的点扩散函数，并基于此点扩散函数和曝光编码信息，从曝光编码图像中解码恢复出N幅清晰的高分辨率景物图像。The software algorithm uses three low-resolution images to calculate the more accurate point spread function of the scene image, and based on the point spread function and exposure coding information, decodes and recovers N clear high-resolution scene images from the exposure coded image.

下面对本发明的具体实施方式作进一步的详细介绍。The specific implementation manners of the present invention will be further described in detail below.

步骤1：拍摄条件设置阶段，主要包括分光系统、控制编码和外部触发信号等的设置。Step 1: The shooting condition setting stage mainly includes the setting of the spectroscopic system, control code and external trigger signal.

分光系统的设置主要是指通过调节分光系统，使得景物光以相同的视场大小和角度分配给两台相机，并且对总的景物光进行合理分配。由于分光系统在拍摄过程中所获得总的光通量是一定的，因此需要在两台相机之间进行光的分配。对于低分辨率高速相机，由于其高速成像过程中，单帧图像的曝光时间较短，可适当增大其光路的光通量和光强度，以提高其成像的亮度；而曝光编码相机，则是对拍摄全过程进行曝光，因此可适当降低其成像光路的光通量和光强度的占比。The setting of the spectroscopic system mainly refers to adjusting the spectroscopic system so that the scene light is distributed to the two cameras with the same field of view size and angle, and the total scene light is reasonably distributed. Since the total luminous flux obtained by the spectroscopic system during shooting is certain, it is necessary to distribute the light between the two cameras. For low-resolution high-speed cameras, due to the short exposure time of a single frame image during its high-speed imaging process, the luminous flux and light intensity of its optical path can be appropriately increased to improve the brightness of its imaging; Exposure is carried out during the whole process, so the proportion of luminous flux and light intensity of its imaging optical path can be appropriately reduced.

控制编码的设置主要包括曝光时间片长度T的设置、曝光全过程时间片个数N的设置、低分辨率高速相机三次曝光成像时刻T_S、T_M、T_E的设置以及曝光编码相机二进制编码B₁B₂B₃…B_N的设置。曝光时间片长度T通常根据所拍摄景物对相机时间分辨能力的要求进行设置，即根据景物变化的速度进行相应设置，使得景物在时间T内的成像不模糊；曝光全过程的时间片个数N则取决于所拍摄景物的变化过程持续时间或者说拍摄者所关心的变化持续时间；高速相机三次曝光成像的时刻T_S、T_M、T_E则通常设置为景物快速变化的开始，过程中以及接近结束，有利于点扩散函数的精确计算和获取；曝光编码相机二进制编码B₁B₂B₃…B_N的设置则是根据不同拍摄对象场景变化的时间域特征进行设置，以便软件算法最大限度的恢复出清晰的景物图像。The setting of the control code mainly includes the setting of the length T of the exposure time slice, the setting of the number of time slices N in the whole exposure process, the setting of the imaging time T_S ,_TM ,_TE of the three exposures of the low-resolution high-speed camera, and the binary code of the exposure code camera. B₁ B₂ B₃ …B_N settings. The length of the exposure time slice T is usually set according to the requirements of the time resolution capability of the camera for the scene to be photographed, that is, it is set according to the speed of the change of the scene, so that the image of the scene within the time T is not blurred; the number of time slices in the whole exposure process is N It depends on the duration of the change process of the scene being photographed or the duration of the change that the photographer is concerned about; the time T_S ,_TM , and_TE of the high-speed camera’s three-exposure imaging is usually set as the beginning of the rapid change of the scene, during the process and Near the end, it is conducive to the accurate calculation and acquisition of point spread function; the setting of exposure coding camera binary code B₁ B₂ B₃ ... B_N is set according to the time domain characteristics of different shooting object scenes, so that the software algorithm can maximize The recovery of a clear scene image.

步骤2：双机同步拍摄阶段，包括低分辨率高速相机三帧高速成像和高分辨率曝光编码相机的曝光编码成像。Step 2: Two-machine synchronous shooting stage, including three-frame high-speed imaging of the low-resolution high-speed camera and exposure coding imaging of the high-resolution exposure coding camera.

如图3所示，在接受外部触发信号后，两台相机根据步骤1设置的拍摄条件同步进行拍摄：低分辨率高速相机在设定好的T_S、T_M、T_E时刻，对景物光分别曝光一个时间片长度，生成三幅较低分辨率的高速图像，即开始帧FrameS、过程帧FrameM和结束帧FrameE。曝光编码相机则根据设定的二进制编码B₁B₂B₃…B_N，对整个拍摄过程进行编码曝光，生成一副曝光编码图像。As shown in Figure 3, after receiving the external trigger signal, the two cameras shoot synchronously according to the shooting conditions set in step 1: the low-resolution high-speed camera captures the scene light at the set T_S , T_M , and T_E Expose one time slice length respectively to generate three low-resolution high-speed images, namely the start frame FrameS, the process frame FrameM and the end frame FrameE. The exposure coded camera performs coded exposure on the entire shooting process according to the set binary codes B₁ B₂ B₃ ...B_N , and generates an exposure coded image.

步骤3：图像解码恢复阶段，包括基于三帧低分辨率图像的景物图像点扩散函数提取和基于点扩散函数的曝光编码图像解码恢复，软件算法的过程如图4所示，具体步骤描述如下：Step 3: The stage of image decoding and restoration, including the extraction of point spread functions of scene images based on three frames of low-resolution images and the decoding and restoration of exposure coded images based on point spread functions. The process of the software algorithm is shown in Figure 4, and the specific steps are described as follows:

3.1算法第一步获取景物图像过程中的点扩散函数（PointSpreadFunction，PSF），点扩散函数表示景物图像中各个像素点随时间运动变化的函数，由于三幅图像分别在景物快速变化的开始、过程中和接近结束时拍摄，因此可以比较好的反应景物的运动变化规律，软件算法可以提取出较为精确的图像点扩散函数。3.1 The first step of the algorithm is the point spread function (PointSpreadFunction, PSF) in the process of obtaining the scene image. The point spread function represents the function of each pixel in the scene image changing with time. Since the three images are at the beginning and process of the rapid change of the scene respectively The shooting is near the end of neutralization, so it can better reflect the movement changes of the scene, and the software algorithm can extract a more accurate image point spread function.

1）首先对开始帧、过程帧和结束帧等三帧较低分辨率图像，进行不同灰度阈值条件下的图像轮廓提取，或者进行灰度梯度计算得到图像的恢复梯度图；1) Firstly, for the three frames of lower-resolution images including the start frame, process frame and end frame, the image contour extraction under different gray threshold conditions is performed, or the gray gradient calculation is performed to obtain the restoration gradient map of the image;

2）然后根据轮廓提取结果和灰度梯度信息，对开始帧与过程帧、过程帧与结束帧分别进行模型匹配，获取景物图像不同灰度阈值条件下，在图像中的位置或扩散信息；2) Then, according to the contour extraction results and the gray gradient information, model matching is performed on the start frame and the process frame, and the process frame and the end frame respectively, and the position or diffusion information in the image is obtained under the condition of different gray threshold values of the scene image;

3）最后根据三帧图像拍摄的时间片信息T_S、T_M、T_E以及景物在不同图像帧中的位置和扩散信息，分别计算出景物图像在开始帧、过程帧与结束帧三个时间片之间的两个点扩散函数PSF₁、PSF₂；3) Finally, according to the time slice information T_S ,_TM ,_TE captured by the three frames of images and the position and diffusion information of the scene in different image frames, the three times of the scene image in the start frame, process frame and end frame are calculated respectively Two point spread functions PSF₁ , PSF₂ between the slices;

3.2 软件算法第二步则是利用已经获取的点扩散函数PSF₁、PSF₂和图像的时间编码信息进行曝光编码图像的解码恢复。曝光编码图像相当于多幅时间片图像的叠加生成的一副模糊图像，在已知点扩散函数的前提下，对模糊图像的恢复，可利用点扩散函数与模糊图像进行反卷积运算；由于曝光编码图像与点扩散函数之间存在相对时间关系，且曝光编码图像与低分辨率高速图像所拍摄景物相同，因此本软件算法将曝光编码图像的时间编码信息以及三副低分辨率高速图像作为已知条件，引入解码算法过程，以提高算法解码图像质量，算法设定误差容限W_E作为衡量恢复图像正确性的判断。3.2 Software Algorithm The second step is to use the obtained point spread function PSF₁ , PSF₂ and the time coding information of the image to decode and restore the exposure coded image. The exposure coded image is equivalent to a pair of blurred images generated by the superposition of multiple time-slice images. On the premise of known point spread function, the recovery of the blurred image can be deconvolved with the blurred image by using the point spread function; because There is a relative time relationship between the exposure coded image and the point spread function, and the scene captured by the exposure coded image and the low-resolution high-speed image is the same, so the software algorithm uses the time code information of the exposure coded image and three low-resolution high-speed images as Given the known conditions, the decoding algorithm process is introduced to improve the quality of the algorithm-decoded image, and the algorithm sets the error tolerance W_E as a judgment to measure the correctness of the restored image.

1）对时间片T_i从T₁到T_N迭代使用本算法1) Use this algorithm iteratively for time slice T_i from T₁ to T_N

2）若时间片T_i位于T_S和T_M之间，则采用PSF₁对该时间片的点扩散函数进行实例化，得到PSF_i；利用PSF_i对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；i=i+1；2) If the time slice T_i is located between_TS and_TM , use PSF₁ to instantiate the point spread function of the time slice to obtain PSF_i ; use PSF_i to deconvolute the exposure coded image to obtain restore image Frame_i ; i=i+1;

3）若时间片T_i位于T_M和T_E之间，则采用PSF₂对该时间片的点扩散函数进行实例化，得到PSF_i；利用PSF_i对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；i=i+1；3) If the time slice T_i is located between_TM and_TE , use PSF₂ to instantiate the point spread function of the time slice to obtain PSF_i ; use PSF_i to deconvolute the exposure coded image to obtain restore image Frame_i ; i=i+1;

4）若时间片T_i位于T₁和T_S之间，采用PSF1进行点扩散函数实例化；利用PSF_i对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；若i=S-1，对恢复图像与开始帧图像FrameS进行抽样误差分析，若误差大于误差容限W_E，则取i=i+1，重新实例化点扩散函数并对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；若i<S-1，则对恢复图像Frame_i与上一帧图像Frame_i+1图像进行误差分析，若误差大于误差容限W_E，则取i+1，重新实例化点扩散函数并对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；i=i-1；4) If the time slice T_i is between T₁ and T_S , use PSF1 to instantiate the point spread function; use PSF_i to deconvolute the exposure coded image to obtain the restored image Frame_i ; if i=S-1 , analyze the sampling error between the restored image and the start frame image FrameS, if the error is greater than the error tolerance W_E , then take i=i+1, re-instantiate the point spread function and deconvolve the exposure coded image to obtain the restored Image Frame_i ; if i<S-1, then analyze the error between the restored image Frame_i and the previous frame image Frame_i+1 , if the error is greater than the error tolerance W_E , take i+1, and re-instantiate the point Diffusion function and perform deconvolution operation on the exposure coded image to obtain the restored image Frame_i ; i=i-1;

5）若时间片T_i位于T_E和T_N之间，采用PSF2进行点扩散函数实例化；利用PSF_i对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；若i=E+1，对恢复图像与开始帧图像FrameE进行抽样误差分析，若误差大于误差容限W_E，则取i=i-1，重新实例化点扩散函数并对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；若i>E+1，则对恢复图像Frame_i与上一帧图像Frame_i-1图像进行误差分析，若误差大于误差容限W_E，则取i-1，重新实例化点扩散函数并对曝光编码图像进行反卷积运算，得到恢复图像Frame_i；i=i+1；5) If the time slice T_i is between T_E and T_N , use PSF2 to instantiate the point spread function; use PSF_i to deconvolute the exposure coded image to obtain the restored image Frame_i ; if i=E+1 , analyze the sampling error between the restored image and the starting frame image FrameE, if the error is greater than the error tolerance W_E , then take i=i-1, re-instantiate the point spread function and deconvolve the exposure coded image to obtain the restored Image Frame_i ; if i>E+1, analyze the error between the restored image Frame_i and the previous frame image Frame_i-1 , if the error is greater than the error tolerance W_E , take i-1, and re-instantiate the point Diffusion function and perform deconvolution operation on the exposure coded image to obtain the restored image Frame_i ; i=i+1;

6）误差分析以两幅图像之间的像素值均方差为依据；6) The error analysis is based on the mean square error of the pixel values between the two images;

7）抽样误差分析：对图像Frame_i与图像Framej进行抽样误差分析，首先对高分辨率的图像进行低分辨率抽样，抽样可以采取像素值抽样或者像素值求平均等方法；然后对抽样后的图像采用误差分析；7) Sampling error analysis: Sampling error analysis is performed on image Frame_i and image Framej. First, low-resolution sampling is performed on high-resolution images. The sampling method can be pixel value sampling or pixel value averaging; and then the sampled The image adopts error analysis;

本发明原理：Principle of the present invention:

本发明将计算成像技术中的曝光编码成像技术与传统高速成像技术相结合，提出了一种新颖的双机高速成像方法。曝光编码成像技术所面临的最大问题是图像点扩散函数的获取，而传统高速成像技术目前面临的则是高帧频与高分辨率的两难问题。通过将低分辨率高速成像技术与高分辨率曝光编码成像相结合，很好的解决了景物点扩散函数精确提取难题，通过对曝光编码图像的解码恢复，实现了超高速成像系统的高分辨率成像能力。The invention combines the exposure coding imaging technology in the computational imaging technology with the traditional high-speed imaging technology, and proposes a novel two-machine high-speed imaging method. The biggest problem facing exposure coding imaging technology is the acquisition of image point spread function, while traditional high-speed imaging technology is currently facing the dilemma of high frame rate and high resolution. By combining low-resolution high-speed imaging technology with high-resolution exposure coding imaging, the problem of accurate extraction of scene point spread functions is well solved. Through decoding and restoration of exposure coding images, the high resolution of ultra-high-speed imaging systems is realized. imaging capabilities.

Claims (7)

1. the two-shipper high speed imaging system based on exposure coding, is characterized in that: comprise beam splitting system, high spatial resolution camera, low spatial resolution camera, parameter setting and synchronous control unit, decoding unit;

Described beam splitting system by scene light with identical visual field size and angular distribution to high spatial resolution camera and low spatial resolution camera;

Described high spatial resolution camera adopts coding exposure;

Described low spatial resolution camera has high speed imaging ability;

Described parameter setting and synchronous control unit are for high spatial resolution camera and the luminous flux of low spatial resolution camera and the accounting of luminous intensity is distributed, triggering is controlled, the parameter setting of high spatial resolution camera, the parameter setting of low spatial resolution camera;

Described decoding unit extracts scene image point spread function for the image of taking from low spatial resolution camera, and utilizes point spread function to the reduction of decoding of the exposure coded image of high spatial resolution camera.

2. the two-shipper high speed imaging system based on exposure coding according to claim 1, is characterized in that:

The parameter setting of described high spatial resolution camera comprises according to captured scenery the requirement of camera time resolution is arranged to time for exposure sheet length T; The timeslice number N of exposure overall process was set according to photographer is concerned about in other words variation duration change procedure duration of captured scenery; According to the time domain characteristic of different reference object scene changes, exposure binary coding B is set₁b₂b₃b_n;

Described coding exposure comprises the following steps:

1) coding exposure stage: the exposure control signal of electronic shutter is subject to the control of exposure coding, if coding BI corresponding to timeslice TI is 1,, in this timeslice, electronic shutter is opened, and pixel cell photosensitive area is accepted scene light and produced corresponding charge inducing; If BI is 0, this timeslice, electronic shutter is not opened, and waits for the exposing operation of carrying out next timeslice after a timeslice length;

2) exposure transition phase: enter the exposure of next timeslice, the photosensitive regions of pixels of imageing sensor is repeated to the exposure process of above-mentioned step 1, if a upper timeslice has exposure, carry out photosensitive area charge inducing to the transfer process of memory block; So repeatedly, after N timeslice exposure process, imageing sensor is through the exposure of once complete coding, and the image information that N timeslice obtains is accumulated and is stored in pixel memory area;

3), by the view data output of storing in imageing sensor, obtain a secondary exposure coded image;

Described decoding reduction comprises the following steps:

1) first according to the point spread function of scene image and timeslice information, carry out point spread function instantiation, obtain the deconvolution matrix of corresponding timeslice;

2) deconvolution matrix corresponding to sheet service time carries out de-convolution operation to exposure coded image, obtains the procedural image of corresponding timeslice;

3), to N timeslice repeating step 1 and 2, obtain procedural image corresponding to each timeslice.

3. the two-shipper high speed imaging system based on exposure coding according to claim 1 and 2, is characterized in that:

The parameter setting of described low spatial resolution camera comprises low resolution high speed camera three times exposure images T constantly_s, T_m, T_esetting.

4. the two-shipper high speed imaging method based on exposure coding, is characterized in that: comprise the steps:

1) parameter setting;

Wherein, high spatial resolution camera parameter comprise according to captured scenery to the requirement of camera time resolution arrange time for exposure sheet length T, according to photographer is concerned about in other words variation duration change procedure duration of captured scenery arrange exposure overall process timeslice number N, according to the time domain characteristic of different reference object scene changes, exposure binary coding B is set₁b₂b₃b_n;

Low spatial resolution camera parameter comprises in fast-changing incipient stage of scenery, process and approaches corresponding three times exposure images of ending phase T constantly_s, T_m, T_e;

Beam splitting system is to high spatial resolution camera and the luminous flux of low spatial resolution camera and the distribution of the accounting of luminous intensity;

2) two cameras are synchronously taken according to the shooting condition of step 1 setting:

2.1) exposure of high spatial resolution camera, transfer and output:

2.1.1) coding exposure stage: the exposure control signal of electronic shutter is subject to the control of exposure coding, if timeslice T_icorresponding coding B_ibe 1,, in this timeslice, electronic shutter is opened, and pixel cell photosensitive area is accepted scene light and produced corresponding charge inducing; If B_ibe 0, this timeslice, electronic shutter is not opened, and waits for the exposing operation of carrying out next timeslice after a timeslice length;

2.1.2) exposure transition phase: enter the exposure of next timeslice, the photosensitive regions of pixels of imageing sensor is repeated to the exposure process of above-mentioned step 2.1.1, if a upper timeslice has exposure, carry out photosensitive area charge inducing to the transfer process of memory block; So repeatedly, after N timeslice exposure process, imageing sensor is through the exposure of once complete coding, and the image information that N timeslice obtains is accumulated and is stored in pixel memory area;

2.1.3), by the view data output of storing in imageing sensor, obtain a secondary exposure coded image;

2.2) exposure of low spatial resolution camera, transfer and output:

The T that low spatial resolution camera is setting_s, T_m, T_econstantly, to the scene light timeslice T length of exposing respectively, generate high speed image the output of three width low resolution; Be start frame FrameS, process frame FrameM and end frame FrameE;

3) decoding reduction:

3.1) acquisition point spread function:

3.1.1), to start frame, process frame and end frame image, the image outline carrying out under different gray threshold conditions extracts, or carries out the shade of gray figure that shade of gray calculates image;

3.1.2) according to profile, extract result and shade of gray information, start frame and process frame, process frame and end frame are carried out respectively to Model Matching, obtain under the different gray threshold conditions of scene image the position in image or diffuse information;

3.1.3) the timeslice information T taking according to three two field pictures_s, T_m, T_eand position and the diffuse information of scenery in different images frame, calculate respectively two the point spread function PSFs of scene image between start frame, process frame and three timeslices of end frame₁, PSF₂;

3.2) decoding reduction:

3.2.1) first according to the point spread function of scene image and timeslice information, carry out point spread function instantiation, obtain the deconvolution matrix of corresponding timeslice;

3.2.2) deconvolution matrix corresponding to sheet service time carries out de-convolution operation to exposure coded image, obtains the scene image of corresponding timeslice;

3.2.3), to N timeslice repeating step 3.2.1 and 3.2.2, obtain scene image corresponding to each timeslice.

5. the two-shipper high speed imaging method based on exposure coding according to claim 4, is characterized in that:

It is described that timeslice information is carried out to the concrete steps of point spread function instantiation and de-convolution operation is as follows:

1) to timeslice T_ifrom T₁to T_niteration is used this algorithm

2) if timeslice T_ibe positioned at T_sand T_mbetween, adopt PSF₁the point spread function of this timeslice is carried out to instantiation, obtain PSF_i; Utilize PSF_iexposure coded image is carried out to de-convolution operation, and image Frame is restored_i; I=i+1;

3) if timeslice T_ibe positioned at T_mand T_ebetween, adopt PSF₂the point spread function of this timeslice is carried out to instantiation, obtain PSF_i; Utilize PSF_iexposure coded image is carried out to de-convolution operation, and image Frame is restored_i; I=i+1;

4) if timeslice T_ibe positioned at T₁and T_sbetween, adopt PSF1 to carry out point spread function instantiation; Utilize PSF_iexposure coded image is carried out to de-convolution operation, and image Frame is restored_i; If i=S-1, carries out Error Analysis of Sampling to Recovery image and beginning frame image FrameS, if error is greater than error margin W_e, get i=i+1, again instantiation point spread function and to exposure coded image carry out de-convolution operation, image Frame is restored_i; If i<S-1, to Recovery image Frame_iwith previous frame image Frame_i+1image carries out error analysis, if error is greater than error margin W_e, get i+1, again instantiation point spread function and to exposure coded image carry out de-convolution operation, image Frame is restored_i; I=i-1;

5) if timeslice T_ibe positioned at T_eand T_nbetween, adopt PSF2 to carry out point spread function instantiation; Utilize PSF_iexposure coded image is carried out to de-convolution operation, and image Frame is restored_i; If i=E+1, carries out Error Analysis of Sampling to Recovery image and beginning frame image FrameE, if error is greater than error margin W_e, get i=i-1, again instantiation point spread function and to exposure coded image carry out de-convolution operation, image Frame is restored_i; If i>E+1, to Recovery image Frame_iwith previous frame image Frame_i-1image carries out error analysis, if error is greater than error margin W_e, get i-1, again instantiation point spread function and to exposure coded image carry out de-convolution operation, image Frame is restored_i; I=i+1.

6. the two-shipper high speed imaging method based on exposure coding according to claim 4, is characterized in that:

Described error analysis is that the pixel value mean square deviation of take between two width images is foundation.

7. the two-shipper high speed imaging method based on exposure coding according to claim 4, is characterized in that:

Described Error Analysis of Sampling is specially: to image Frame_icarry out Error Analysis of Sampling with image Framej, first high-resolution image is carried out to low resolution sampling, the method that sampling takes pixel value sampling or pixel value to be averaging; Then the image after sampling is adopted to error analysis.

Images