Point spread function calibration system and method for digital imaging system based on star point target

Abstract

The present invention discloses a point spread function calibration system and method for a digital imaging system based on a star point target. In the system, the star point object space target is composed of a point light source array, which is fixedly mounted on an object plane base via a point light source adjustment component. The point light source array is evenly spaced on the object plane base, and point light sources in different regions are fixedly mounted on the object plane base via corresponding point light source adjustment components. The digital imaging system is mounted on a translation stage, and the point light source array is located in a quarter of the ideal object plane of the digital imaging system. The method utilizes the system to perform point spread function detection on the digital imaging system in a sub-field of view. The present invention realizes the construction of a point spread function calibration device for a digital imaging system, and uses Zernike ki to optimize the point spread function calculated from the real-shot image, thereby effectively calibrating the point spread function of the digital imaging system.

Description

Technical Field

[0001] The present invention belongs to the field of digital image processing and relates to a point spread function calibration system for a digital imaging system, in particular to a point spread function calibration system and method for a digital imaging system based on a star point object space target. Background Art

[0002] The point spread function (PSF) describes the response of a digital imaging system to a point light source or impulse function. The degree of energy dispersion is an important indicator of the system's image quality. Methods for measuring the PSF of actual optical systems can be primarily categorized into image deconvolution and real-time calibration. Image deconvolution requires extensive post-processing and computational complexity, making it inaccurate for dense calibration across the entire field of view. Furthermore, due to the ill-posed nature of the deconvolution problem, the estimated PSF is inaccurate. Real-time calibration uses a specific star-point object-space target to simulate the imaging result of a digital imaging system at an ideal point on the object plane. During this calibration process, noise in the measured image and lens vignetting are unavoidable. Therefore, to meet the requirements for dense PSF measurement and image quality assessment of digital imaging systems, it is necessary to develop targeted hardware devices (to simulate specific star-point object-space targets) and software algorithms (to address image noise and lens vignetting) to obtain the PSF of imaging systems. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies and address the problems encountered in the background art, the present invention proposes a point spread function calibration system and method for digital imaging systems based on star-point object space targets. This system uses a dataset composed of real-world star-point object space targets to simulate the imaging results of a digital imaging system for an ideal point on an object surface. During the calibration process, vignetting in the measured images and lens vignetting are unavoidable. Therefore, considering practical engineering requirements, the present invention ensures that the principal rays of a point light source converge at the center of the imaging system's optical axis.

[0004] In order to achieve the above objectives, the present invention adopts the following technical solutions:

[0005] 1. A Point Spread Function Calibration System for Digital Imaging Systems Based on Star Targets

[0006] The system includes an object plane base, a point light source array, a point light source adjustment component and a translation stage; the point light source array is fixedly mounted on the object plane base through the point light source adjustment component, the point light source array consists of multiple point light sources arranged in a matrix array on the object plane base at intervals, each point light source is fixedly mounted on the object plane base through a corresponding point light source adjustment component, a digital imaging system is fixedly mounted on the translation stage, the digital imaging system is arranged on one side of the object plane base where the point light source array is mounted, and the digital imaging system and the point light source array are arranged at intervals, and the point light source array serves as an object side star point target.

[0007] Each point light source in the point light source array is on the imaging plane of the digital imaging system. The full field of view of the digital imaging system is divided into four equal parts with the center of the optical axis and recorded as the first to fourth sub-fields of view respectively. The dividing line between two adjacent sub-fields of view is recorded as the field of view dividing line. The point light source array is located in one of the sub-fields of view of the digital imaging system. One of the corner points of the point light source array coincides with the center of the optical axis of the digital imaging system. The two sides of the point light source array are respectively on the two field of view dividing lines of the current sub-field of view, and the other corner point of the point light source array is set on the field of view edge line of the current sub-field of view or outside the current sub-field of view.

[0008] Take multiple segmented field of view angles of the current sub-field of view, and then calculate the radii of multiple corresponding arc dividing lines in the point light source array based on the segmented field of view angles, and then use the multiple arc dividing lines to divide the point light source array into point light source areas, and the heights of all point light sources in the point light source array are the same; in each point light source area, the main light rays of all point light sources are the same and the main light rays of the point light source where the middle arc is located are incident on the optical axis center of the digital imaging system.

[0009] The radius r of the arc dividing line θ The calculation formula is as follows:

[0010] r θ =|l|tanθ n1

[0011] Where, |l| is the distance from the optical axis center of the digital imaging system to the point light source array, θ n1 Indicates the n1th segmented field of view angle.

[0012] The point light source adjustment assembly includes a bending base and a base height adjustment gasket, the base height adjustment gasket is fixedly mounted on the object plane base, one side of the lower surface of the bending base is fixedly mounted on the base height adjustment gasket, and the other side of the upper surface of the bending base is fixedly mounted with a point light source; the folding angle of the bending base is adjusted to adjust the emission angle of the main light of the corresponding point light source, and the height of the base height adjustment gasket is adjusted to ultimately make the heights of each point light source the same.

[0013] The height calculation formula of the base height adjustment gasket is as follows:

[0014] h n2 =s(sin40°-sinα n2 )

[0015] Among them, h n2 represents the height of the base height adjustment pad in the n2th point light source area, s is the distance between the center of the point light source and the bending line of the bending base, α n2Indicates the folding angle of the bent base in the n2th point light source area.

[0016] 2. A Point Spread Function Calibration Method for Digital Imaging Systems Based on Star Targets

[0017] The method adopts the point spread function calibration system, and the method includes the following steps:

[0018] 1) Place a plane mirror between the digital imaging system and the point light source array, parallel to and spaced from the point light source array. Move the translation stage until the reflected image of the digital imaging system is centered in the captured image. Record the position of the translation stage at this point as the initial position of the digital imaging system, and remove the plane mirror.

[0019] 2) Taking the initial position of the digital imaging system as the origin, capturing sensor images of the digital imaging system under different offsets of the point light source array in the current sub-field of view, and performing bad pixel correction, dark level subtraction, white balancing, and demosaicing on the sensor images under different offsets in the current sub-field of view to obtain linear color images under different offsets in the current sub-field of view;

[0020] 3) performing point spread function (PSF) detection on the linear color image under each offset condition in the current sub-field of view;

[0021] 4) calculating the optical transfer function of the point spread function under each offset condition in the current sub-field of view based on the point spread function PSF under each offset condition in the current sub-field of view, and optimizing a set of Zernike bases to minimize the mean square error between the optical transfer function reconstructed by the Zernike base and the optical transfer function of all point spread functions, and then performing an inverse Fourier transform on the optical transfer function reconstructed by the Zernike bases to obtain the point spread function of the current sub-field of view;

[0022] 5) Rotate the digital imaging system so that the point light source array is located in another sub-field of view, repeat 3) to 4) to obtain the point spread function of another sub-field of view, until the point spread functions of all sub-fields of view of the digital imaging system are obtained.

[0023] Said 3) is specifically:

[0024] 3.1) Determine the minimum field of view area in the pixel space of the linear color image based on the spacing between two adjacent point light sources in the point light source array. For each offset linear color image in the current sub-field of view, determine whether a measured point spread function (PSF) that meets the threshold condition exists within each minimum field of view area of the linear color image in the current offset condition. If so, execute 3.2). Otherwise, expand the current field of view area until a measured point spread function (PSF) that meets the threshold condition exists within the latest field of view area, and then execute 3.2).

[0025] 3.2) Calculating a histogram of the current field of view, and then using the histogram to count the number of pixels whose pixel values are greater than a pixel threshold, where the pixel threshold is 0.8 times the maximum value (max) of the current field of view, and marking pixels whose pixel values are greater than the pixel threshold as core pixels; when the number of core pixels is 1, using the current core pixel as the center pixel of the point spread function (PSF), intercepting a matrix of a preset size within the current field of view as the point spread function (PSF);

[0026] When the number of core pixels is greater than or equal to 2, the average value of the pixel space coordinates of all core pixels is calculated. When the deviations of the maximum and minimum values of the sum of the core pixel coordinates from the average value are greater than the preset offset threshold, the core pixel closest to the center of the current field of view area is taken as the center pixel of the point spread function PSF, and a matrix of a preset size is intercepted within the current field of view area as the point spread function PSF; otherwise, the pixel where the average value of the pixel space coordinates of all core pixels is located is taken as the center pixel of the point spread function PSF, and a matrix of a preset size is intercepted within the current field of view area as the point spread function PSF;

[0027] 3.3) Binarize the point spread function (PSF) of the current field of view at a pixel value 0.05 times the maximum pixel value (max) to obtain a binarized pixel region. Calculate the centroid of the binarized pixel region. Round the coordinates of the centroid to an integer and then translate the PSF so that the centroid of the PSF is within the sub-pixel range of the matrix center. This yields the final PSF for the current field of view.

[0028] 3.4) Repeat 3.1) to 3.3) to traverse all the field of view areas of the linear color image under the current offset condition to obtain the point spread function matrix of the linear color image under the current offset condition;

[0029] 3.5) Repeat 3.1) to 3.4) until the point spread function matrix corresponding to the linear color image under each offset condition in the current sub-field of view is calculated.

[0030] The threshold condition is that the maximum value (max) of pixels in the current field of view area is greater than 5000.

[0031] The beneficial effects of the present invention are:

[0032] The present invention designs a PSF calibration system and method for a digital imaging system based on a star-point object space target. The method is applicable to conventional digital imaging systems. The imaging results of an ideal point light source or impulse function on an object surface by a digital imaging system can be simulated, and the point spread function of the system can be quickly and densely calibrated. Since the influence of image noise is difficult to avoid during the calibration process, the method uses a digital imaging system to collect calibration images under various offset conditions multiple times, and then obtains the point spread function calibration results after noise removal by Zernike fitting. The calibrated point spread function can accurately reflect the degradation of the digital imaging system in the field of view, and can be used for various tasks such as image quality evaluation, imaging simulation, and non-blind image restoration. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Schematic diagram of the system of the present invention;

[0034] Figure 2 Schematic diagram of the overall process of the method of the present invention;

[0035] Figure 3 Schematic diagram of the object plane substrate for constructing a quarter field of view for the system of the present invention;

[0036] Figure 4 This is a schematic diagram of the installation of the LED point light source of the present invention

[0037] Figure 5 This is a rendering of the height correction effect of the bending base of the system of the present invention;

[0038] Figure 6 This is a specific flow chart of step 3) in the method of the present invention;

[0039] Figure 7 The results of calibrating the point spread functions of certain digital imaging systems using the method of the present invention (taking Model 1 and Model 2 as examples).

[0040] In the figure: object plane substrate 1, point light source array 2, bending base 3, base height adjustment gasket 4, and translation stage 5. DETAILED DESCRIPTION

[0041] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0042] like Figure 1As shown, a point spread function calibration system for a digital imaging system based on a star-point object space target includes an object plane substrate 1, a point light source array 2, a point light source adjustment assembly, and a translation stage 5; the translation stage 5 has a micrometer-level rotational range. The point light source array 2 is fixedly mounted on the object plane substrate 1 via the point light source adjustment assembly. The point light source array 2 comprises a plurality of point light sources arranged in a matrix array on the object plane substrate 1 at intervals. Each point light source is fixedly mounted on the object plane substrate 1 via a corresponding point light source adjustment assembly. In a specific embodiment, the point light sources are LEDs. A digital imaging system is fixedly mounted on the translation stage 5. The digital imaging system is arranged on one side of the object plane substrate 1 where the point light source array 2 is mounted, and is spaced apart between the digital imaging system and the point light source array 2. The point light source array 2 serves as the object space star point target.

[0043] like Figure 3 As shown, each point light source in the point light source array 2 is on the imaging plane of the digital imaging system, that is, the optical axis of the digital imaging system is perpendicular to the plane where the point light source array 2 is located; the full field of view of the digital imaging system is divided into four equal parts with the center of the optical axis and respectively recorded as the first to fourth sub-fields of view, and the dividing line between two adjacent sub-fields of view is recorded as the field of view dividing line. The point light source array 2 is located in one of the sub-fields of view of the digital imaging system, and one of the corner points of the point light source array 2 coincides with the center of the optical axis of the digital imaging system. The two sides of the point light source array 2 are respectively on the two field of view dividing lines of the current sub-field of view and the other corner point of the point light source array 2 is set on the field of view edge line of the current sub-field of view or outside the current sub-field of view. One of the corner points and the other corner point of the above-mentioned point light source array 2 are the diagonal points of the point light source array 2.

[0044] Take multiple segmented field of view angles of the current sub-field of view, and then calculate the radii of the corresponding multiple arc dividing lines in the point light source array 2 based on the segmented field of view angles, and then use the multiple arc dividing lines to divide the point light source array 2 into point light source areas. The heights of all point light sources in the point light source array 2 are the same; in each point light source area, the main light rays of all point light sources are the same and the main light rays of the point light source where the middle arc is located are incident on the center of the optical axis of the digital imaging system.

[0045] The radius of the arc dividing line r θ The calculation formula is as follows:

[0046] r θ =|l|tanθ n1

[0047] Where, |l| is the distance from the optical axis center of the digital imaging system to the point light source array, θ n1 Indicates the n1th segmented field of view, n1 = 1, 2, 3, 4, 5.

[0048] like Figure 4As shown, the point light source adjustment component includes a bending base 3 and a base height adjustment gasket 4, the base height adjustment gasket 4 is fixedly mounted on the object plane base 1, one side of the lower surface of the bending base 3 is fixedly mounted on the base height adjustment gasket 4, and the other side of the upper surface of the bending base 3 is fixedly mounted with a point light source; the folding angle of the bending base 3 is adjusted to adjust the emission angle of the main light of the corresponding point light source, and the folding angles of all the bending bases 3 in each point light source area are the same, plus the height adjustment of the base height adjustment gasket 4, ultimately make the heights of each point light source the same, as shown in FIG. Figure 5 shown.

[0049] The height calculation formula of the base height adjustment gasket 4 is as follows:

[0050] h n2 =s(sin40°-sinα n2 )

[0051] Among them, h n2 represents the height of the base height adjustment pad 4 in the n2th point light source area, s is the distance between the center of the point light source and the bending line of the bending base 3, α n2 Represents the folding angle of the bending base 3 in the n2-th point light source area, n2=1, 2, 3, 4, 5, 6.

[0052] In a specific implementation, the object plane substrate 1 is 2 meters by 2 meters. This 2 meter by 2 meter object plane is divided into four 1 meter by 1 meter planes. These four planes are then stitched together to form the object plane substrate for a quarter of the field of view of the digital imaging system. The distance between the point light source array 2 and the digital imaging system is 3 meters.

[0053] The focal length f and sensor pixel size d of the digital imaging system used in this embodiment are pixel The number of point light sources required for the array is n, which is 1.6um and designed on a 2m x 2m plane. total is 1600.

[0054] The calculation formula for the distance d between two adjacent point light sources in the point light source array 2 is as follows:

[0055]

[0056] Wherein, a is the side length of the object plane base, which is a=2000 mm in the present invention.

[0057] According to the number of point light sources n in the array total Calculate the array spacing d of the point light sources, d = 50 mm. Center the array point light sources on the plane according to the sampling interval distance.

[0058] The half-viewing angle θ of the digital imaging system used in this embodiment is 40°. The half-viewing angle is divided into an arithmetic progression with the first term being 0° and the tolerance being 8°. The 40° viewing range is divided into 6 point light source areas. The dividing line radius of the 6 point light source areas is used to calculate the segmented viewing angle θ. n They are 8°, 16°, 24°, 32°, and 40° respectively.

[0059] Each point light source area uses a 0402 SMD white LED. Due to the limited minimum exposure time of the digital imaging system used, testing has shown that adding a 4.7kΩ resistor in series provides the appropriate brightness for the white LEDs. Using bracketed exposure, the digital imaging system can capture PSFs of varying brightness. Each white LED is connected in series with a 4.7kΩ resistor to form a branch circuit, and all branches are then connected in parallel.

[0060] Ensure that the main light emitted by the point light source in each area converges to the center of the optical axis of the imaging system. In the six areas of the object plane divided by the dividing line, with the center of the field of view of the digital imaging system as the reference, the area closest to the camera to the area farthest from the camera are area 1, area 2, etc. The point light sources in the same area are all installed on the surface of the base accessory with the same bending angle. The base bending angle α used in the point light source area n2 is n2 =4°,12°,20°,28°,36°,40°.

[0061] Because the base bend angles used in different areas vary, the point light sources mounted on the inclined surface of the bent base will have a height difference Δh. Washers of different heights h are placed between the base and the object plane to ensure that all point light sources are at the same height. Using the center of the point light source mounted on a 40° bent base as a reference, the required washer heights h for bases with different bend angles are calculated. In this embodiment, s = 10mm. The calculated shim heights for the corresponding bent bases are: 5.73mm, 4.35mm, 3.00mm, 1.73mm, and 0.55mm, respectively.

[0062] Therefore, the 1 / 4 field of view object plane substrate composed of four 1 meter × 1 meter planes is divided into 6 parts. The first part is the field of view from 0° to 8°. The white light LED lamps in this area need to be installed on the inclined surface of the base with a 4° bend; the second part is the field of view from 8° to 16°. The white light LED lamps in this area need to be installed on the inclined surface of the base with a 12° bend; the third part is the field of view from 16° to 24°. The white light LEDs in this area need to be installed on the inclined surface of the base with a 20° bend; the fourth part is the field of view from 24° to 32°. The white light LEDs in this area need to be installed on the inclined surface of the base with a 28° bend; the fifth part is the field of view from 32° to 40°. The white light LEDs in this area need to be installed on the inclined surface of the base with a 36° bend; the sixth part is the field of view greater than 40°. The white light LEDs in this area need to be installed on the inclined surface of the base with a 40° bend.

[0063] The relative height deviation of the white light LED mounting base in different areas. The white light LED mounting base in each area needs to use washers of different heights to adjust the height difference Δh. Part 1 needs to use a washer with a height of 5.39mm, Part 2 needs to use a washer with a height of 4.01mm, Part 3 needs to use a washer with a height of 2.72mm, Part 4 needs to use a washer with a height of 1.53mm, and Part 5 needs to use a washer with a height of 0.47mm. The bent base is designed using sheet metal technology. All point light sources have the same installation position on the unbent base. When the base is bent into different angles, with the center of the point light source as the reference, the relative positions of the point light sources in different areas not only have height differences, but also have displacements in the horizontal direction. Similarly, using the point light source installed on the 40° bent base as a reference, the displacement x that needs to be corrected in the horizontal direction for bases using other bending angles is calculated. n They are 4.18mm, 3.55mm, 2.74mm, 1.76mm, and 0.62mm respectively.

[0064] like Figure 2 As shown, a method for calibrating the point spread function of a digital imaging system based on a star point object space target comprises the following steps:

[0065] 1) Place a plane mirror between the digital imaging system and the point light source array 2, parallel to and spaced from the point light source array 2. Move the translation stage 5 until the reflected image of the digital imaging system is at the center of the captured image, i.e., the center of the object plane of the point light source array 2 coincides with the center of the image plane of the digital imaging system. The position of the translation stage 5 at this point is recorded as the initial position of the digital imaging system, and the plane mirror is removed.

[0066] 2) Using the initial position of the digital imaging system as the origin, the digital imaging system is translated horizontally and vertically using a translation stage. After each translation, an image is captured. Multiple translations are performed to ultimately obtain sensor images of the digital imaging system under different offset conditions of the point light source array 2 in the current sub-field of view. The sensor images under different offset conditions in the current sub-field of view are subjected to bad pixel correction, dark level subtraction, white balancing, and demosaicing to obtain linear color images under different offset conditions in the current sub-field of view.

[0067] In a specific implementation, the horizontal translation amount is between -2mm and 2mm, and the vertical translation amount is between -1.5mm and 1.5mm. After each translation, a command is sent to the digital imaging system to capture an image, thereby obtaining a sensor image of the shooting device.

[0068] The captured sensor images (sensor images refer to unprocessed images recorded by the digital imaging system's sensor, typically in .dng format for mobile digital imaging systems such as mobile phones) are subjected to bad pixel correction, dark level subtraction, demosaicing, and white balancing to produce linear color images under different offset conditions. For bad pixel correction, the digital imaging system under test is first calibrated: two sensor images are captured with the digital imaging system facing a D65 light box, and then three sensor images are captured in a completely dark environment with the digital imaging system covered with a black cloth. The five images are read using the Python rawpy package and the rawpy.enhance.find_bad_pixels() function is used to generate a matrix containing bad pixel information for the digital imaging system's sensor. The matrix is in N×2 format, with the first dimension N representing the number of bad pixels and the second dimension 2 representing the pixel coordinates of the bad pixels. Once the bad pixel information matrix is obtained, the rawpy.enhance.repair_bad_pixels() function can be used to perform bad pixel correction on any input sensor image. In dark level subtraction, the dark level value black_level and image bit depth depth recorded in the sensor image header file can be used for processing. The processing formula is as follows:

[0069]

[0070] Among them, image i To complete the image after bad pixel correction, image s The image after dark level subtraction is completed. In demosaicing, the R, G, and B channels of the dark level subtraction image are interpolated separately through bilinear interpolation to obtain the demosaiced image. In white balance, the white balance value recorded in the sensor image header file can be used to process the image after dark level subtraction. The processing formula is as follows:

[0071]

[0072] Among them, R in , G in , B in are the three-channel pixel values of each pixel in the demosaiced image, m×n is the number of pixels in the image, and R out , G out , B out They are the three-channel pixel values of each pixel in the image after white balance, WB R , WB G , WB B They are the white balance gains of the three channels respectively.

[0073] 3) performing point spread function (PSF) detection on the linear color image under each offset condition in the current sub-field of view;

[0074] 3) Specifically:

[0075] 3.1) Determine the minimum field of view area in the pixel space of the linear color image based on the distance between two adjacent point light sources in the point light source array 2. For each offset linear color image in the current sub-field of view, the calculation flow chart of the corresponding point spread function PSF is as follows: Figure 6 As shown, determine whether there is a measured PSF that meets the threshold condition in each minimum field of view area of the linear color image under the current offset condition. If so, execute 3.2). Otherwise, expand the current field of view area until a measured PSF that meets the threshold condition exists in the latest field of view area, and then execute 3.2).

[0076] Because the pixel spacing between two adjacent LEDs on the object side, when converted to the image space sensor, is 50 pixels during device design, the field of view is initialized to a 50×50 pixel range on the linear color image. In the pixel space of the linear color image, the field of view Region(h,w) can be expressed as:

[0077] Region(h,w)=image rgb ((h-1)*50+1:h*50,(w-1)*50+1:w*50) where image rgb () is a linear color image, Region is the field of view, h is the longitudinal field of view number of the linear color image, and w is the horizontal field of view number of the linear color image (for example, when the resolution of the linear color image is 3000×4000, the serial number h is an integer value in the sequence [1,2,…,60(3000 / 50)], and the serial number w is an integer value in the sequence [1,2,…,80(4000 / 50)]).

[0078] The threshold condition is that the maximum pixel value (max) of the current field of view area is greater than 5000, and the pixel value range of the linear color image is between 0 and 65535.

[0079] 3.2) Calculate the histogram of the current field of view area, and then use the histogram to count the number of pixels whose pixel values are greater than the pixel threshold. The pixel threshold is 0.8 times the maximum value max of the pixels in the current field of view area. Pixels with pixel values greater than the pixel threshold are recorded as core pixels. When the number of core pixels is 1, only the neighborhood of the maximum value meets the conditions of the measured PSF. Therefore, the current core pixel can be directly used as the center pixel of the point spread function PSF, and a matrix of a preset size can be intercepted in the current field of view area as the point spread function PSF. In the specific implementation, the size of the point spread function PSF is 31×31 pixels.

[0080] When the number of core pixels is greater than or equal to 2, two situations may occur: 1. The pixel dispersion within the same PSF is large, and the core pixels greater than the threshold are all detected within the same PSF; 2. There are two or more PSFs in the field of view.

[0081] In order to determine whether it is the first case or the second case, it is necessary to calculate the pixel space coordinates of all core pixels (remember h r , w r ), when the deviations of the maximum and minimum values of the sum of the core pixel coordinates from the average value are all greater than a preset offset threshold, in a specific implementation, the preset offset threshold is 3 pixels, which is an empirical value. It can be considered that the field of view area is the second case, and the core pixel closest to the center of the current field of view area is used as the center pixel of the point spread function PSF, and then a matrix of a preset size is intercepted in the current field of view area as the point spread function PSF; otherwise, the pixel where the average value of the pixel space coordinates of all core pixels is located is used as the center pixel of the point spread function PSF, and then a matrix of a preset size is intercepted in the current field of view area as the point spread function PSF;

[0082] 3.3) Binarize the point spread function (PSF) of the current field of view at 0.05 times the maximum pixel value (max) to obtain a binarized pixel region. Calculate the centroid of the binarized pixel region. Round the coordinates of the centroid to an integer and then translate the entire PSF so that the centroid of the PSF is within the sub-pixel range of the matrix center. This yields the final PSF for the current field of view.

[0083] 3.4) Repeat 3.1) to 3.3) to traverse all the field of view areas of the linear color image under the current offset condition to obtain the point spread function matrix of the linear color image under the current offset condition;

[0084] 3.5) Repeat 3.1) to 3.4) until the point spread function matrix corresponding to the linear color image under each offset condition in the current sub-field of view is calculated.

[0085] 4) calculating the optical transfer function of the point spread function under each offset condition in the current sub-field of view based on the point spread function PSF under each offset condition in the current sub-field of view, and optimizing a set of Zernike bases to minimize the mean square error between the optical transfer function reconstructed by the Zernike base and the optical transfer function of all point spread functions, wherein the reconstructed optical transfer function is obtained by multiplying the optimized Zernike coefficients with the corresponding Zernike bases and then summing them, and then performing an inverse Fourier transform on the optical transfer function reconstructed by the Zernike bases to obtain the point spread function of the current sub-field of view;

[0086] Step 4 is as follows:

[0087] (4.1) For each offset condition obtained from a certain field of view, the optical transfer function (OTF) of the point spread function (PSF) is calculated using the following formula:

[0088] OTF = fftshift(fft2(PSF))

[0089] Where fft2() is a two-dimensional Fourier transform, and fftshift() swaps the high-frequency and low-frequency components of the fft matrix. After calculating the optical transfer functions for each offset, it is necessary to optimize a set of Zernike coefficients to minimize the mean square error between the reconstructed optical transfer function and the optical transfer functions of all point spread functions. The optimization equation can be expressed as:

[0090]

[0091] Where k is the vector representing the Zernike coefficients, OTF i Optical transfer function representing point spread function calculations for different offsets, is Zernike, which can be divided into two categories: odd and even, which can be expressed as:

[0092]

[0093]

[0094] in, For Chizerniky, is an even Zernike basis (the value of the Zernike basis number l can be positive m or -m), l, n and m represent the number of the Zernike basis, n ≥ m is a non-negative integer, and the value can be 0, 1, 2, etc., m is an integer, and the value can be 0, -1, 1, etc., r is the radial distance within the unit circle in the polar coordinate system, θ is the direction angle within the unit circle in the polar coordinate system, and the Zernike parameter It can be expressed as follows:

[0095]

[0096] This optimization equation can be directly differentiated with respect to the variable i, and the optimal solution of i can be obtained by the Newton-Raphson method or the damped least squares method;

[0097] (4.2) The optimized Zernike coefficients are multiplied by the corresponding Zernike basis and then summed to obtain the reconstructed optical transfer function OTF recon :

[0098]

[0099] Performing inverse Fourier transform on the reconstructed optical transfer function can obtain the point spread function of the field of view:

[0100] PSF=ifft2(ifftshift(OTF))

[0101] Among them, ifftshift() represents the inverse Fourier transform

[0102] 5) Rotate the digital imaging system so that the point light source array (2) is located in another sub-field of view, and repeat 3) to 4) to obtain the point spread function of another sub-field of view, until the point spread functions of all sub-fields of view of the digital imaging system are obtained. Through this set of processes, the present invention calibrates the main camera optical systems of multiple commercial mobile phones, and the RGB channel point spread functions of a certain field of view of model 1 and model 2 are obtained as follows: Figure 7 shown.

Claims

1. A point spread function calibration system for a digital imaging system based on a star target, characterized by: The invention comprises an object plane base (1), a point light source array (2), a point light source adjustment component and a displacement stage (5); the point light source array (2) is fixedly mounted on the object plane base (1) via the point light source adjustment component; the point light source array (2) comprises a plurality of point light sources arranged at intervals on the object plane base (1) in the form of a matrix array; each point light source is fixedly mounted on the object plane base (1) via a corresponding point light source adjustment component; a digital imaging system is fixedly mounted on the displacement stage (5); the digital imaging system is arranged on one side of the object plane base (1) where the point light source array (2) is mounted, and the digital imaging system and the point light source array (2) are arranged at intervals; the point light source array (2) serves as an object side star point target; A plurality of segmented viewing angles of the current sub-viewing field are obtained, and then the radii of a plurality of arc segmentation lines corresponding to the point light source array (2) are calculated based on the segmented viewing angles, and then the point light source array (2) is divided into point light source areas using the plurality of arc segmentation lines, wherein all point light sources of the point light source array (2) have the same height; in each point light source area, the main light rays of all point light sources have the same exit angle, and the main light rays of the point light source where the middle arc is located are incident on the optical axis center of the digital imaging system.

2. The point spread function calibration system for a digital imaging system based on a star target according to claim 1, characterized in that: Each point light source in the point light source array (2) is on the imaging plane of the digital imaging system, the full field of view of the digital imaging system is equally divided into four parts with the center of the optical axis and respectively recorded as the first to fourth sub-fields of view, and the dividing line between two adjacent sub-fields of view is recorded as the field of view dividing line, the point light source array (2) is located in one of the sub-fields of view of the digital imaging system, one of the corner points of the point light source array (2) coincides with the center of the optical axis of the digital imaging system, two sides of the point light source array (2) are respectively on two field of view dividing lines of the current sub-field of view, and another corner point of the point light source array (2) is arranged on the field of view edge line of the current sub-field of view or outside the current sub-field of view.

3. The point spread function calibration system for a digital imaging system based on star point targets according to claim 1, characterized in that: The radius r of the arc dividing line θ The calculation formula is as follows: r θ =|l|tanθ n1 Where, |l| is the distance from the optical axis center of the digital imaging system to the point light source array, θ n1 Indicates the n1th segmented field of view angle.

4. The point spread function calibration system for a digital imaging system based on star point targets according to claim 1, characterized in that: The point light source adjustment component comprises a bending base (3) and a base height adjustment gasket (4); the base height adjustment gasket (4) is fixedly mounted on the object plane base (1); one side of the lower surface of the bending base (3) is fixedly mounted on the base height adjustment gasket (4); and the other side of the upper surface of the bending base (3) is fixedly mounted with a point light source; the folding angle of the bending base (3) is adjusted to adjust the emission angle of the main light of the corresponding point light source; the height of the base height adjustment gasket (4) is adjusted to finally make the heights of the various point light sources the same.

5. The point spread function calibration system for a digital imaging system based on star point targets according to claim 4, characterized in that: The height calculation formula of the base height adjustment pad (4) is as follows: h n2 =s(sin40°-sinα n2 ) Among them, h n2 represents the height of the base height adjustment pad (4) in the n2th point light source area, s is the distance between the center of the point light source and the bending line of the bending base (3), α n2 Indicates the folding angle of the bending base (3) in the n2-th point light source area.

6. A method for calibrating the point spread function of a digital imaging system based on a star target, characterized in that: The method adopts the point spread function calibration system according to any one of claims 1 to 5, and the method comprises the following steps: 1) placing a plane mirror between the digital imaging system and the point light source array (2), the plane mirror being parallel to the point light source array (2) and spaced apart, moving the translation stage (5) until the reflected image of the digital imaging system is at the center of the captured image, recording the position of the translation stage (5) at this time as the initial position of the digital imaging system, and removing the plane mirror; 2) Taking the initial position of the digital imaging system as the origin, photographing and obtaining the sensor images of the digital imaging system under different offset conditions of the point light source array (2) in the current sub-field of view, and performing bad pixel correction, dark level subtraction, white balance, and demosaicing on the sensor images under different offset conditions in the current sub-field of view to obtain linear color images under different offset conditions in the current sub-field of view; 3) performing point spread function (PSF) detection on the linear color image under each offset condition in the current sub-field of view; 4) calculating the optical transfer function of the point spread function under each offset condition in the current sub-field of view based on the point spread function PSF under each offset condition in the current sub-field of view, and optimizing a set of Zernike bases to minimize the mean square error between the optical transfer function reconstructed by the Zernike base and the optical transfer function of all point spread functions, and then performing an inverse Fourier transform on the optical transfer function reconstructed by the Zernike bases to obtain the point spread function of the current sub-field of view; 5) Rotate the digital imaging system so that the point light source array (2) is located in another sub-field of view, repeat 3) to 4) to obtain the point spread function of another sub-field of view, until the point spread functions of all sub-fields of view of the digital imaging system are obtained.

7. The method for calibrating point spread function of a digital imaging system based on star target according to claim 6, characterized in that: Said 3) is specifically: 3.1) Determine the minimum field of view area in the pixel space of the linear color image based on the spacing between two adjacent point light sources in the point light source array (2); for each linear color image in each offset condition in the current sub-field of view, determine whether there is a measured point spread function (PSF) that meets the threshold condition in each minimum field of view area of the linear color image in the current offset condition; if so, execute 3.2); otherwise, expand the current field of view area until there is a measured point spread function (PSF) that meets the threshold condition in the latest field of view area, and then execute 3.2); 3.2) Calculating a histogram of the current field of view, and then using the histogram to count the number of pixels whose pixel values are greater than a pixel threshold, where the pixel threshold is 0.8 times the maximum value (max) of the current field of view, and marking pixels whose pixel values are greater than the pixel threshold as core pixels; when the number of core pixels is 1, using the current core pixel as the center pixel of the point spread function (PSF), intercepting a matrix of a preset size within the current field of view as the point spread function (PSF); When the number of core pixels is greater than or equal to 2, the average value of the pixel space coordinates of all core pixels is calculated. When the deviations of the maximum and minimum values of the sum of the core pixel coordinates from the average value are greater than the preset offset threshold, the core pixel closest to the center of the current field of view area is taken as the center pixel of the point spread function PSF, and a matrix of a preset size is intercepted within the current field of view area as the point spread function PSF; otherwise, the pixel where the average value of the pixel space coordinates of all core pixels is located is taken as the center pixel of the point spread function PSF, and a matrix of a preset size is intercepted within the current field of view area as the point spread function PSF; 3.3) Binarize the point spread function (PSF) of the current field of view at a pixel value 0.05 times the maximum pixel value (max) to obtain a binarized pixel region. Calculate the centroid of the binarized pixel region. Round the coordinates of the centroid to an integer and then translate the PSF so that the centroid of the PSF is within the sub-pixel range of the matrix center. This yields the final PSF for the current field of view. 3.4) Repeat 3.1) to 3.3) to traverse all the field of view areas of the linear color image under the current offset condition to obtain the point spread function matrix of the linear color image under the current offset condition; 3.5) Repeat 3.1) to 3.4) until the point spread function matrix corresponding to the linear color image under each offset condition in the current sub-field of view is calculated.

8. The method for calibrating point spread function of a digital imaging system based on star point targets according to claim 7, characterized in that: The threshold condition is that the maximum value (max) of pixels in the current field of view area is greater than 5000.

Images