Noise model and image noise reduction method for focal plane reading circuit

Abstract

The invention discloses a noise model for a focal plane reading circuit. The noise model comprises an original output data array, a frame array, an original noise array, a row / column mean value array, a row / column stripe array, a random noise array and a row / column stripe noise array; in the array conversion mode, an original output data array is averaged by taking a frame as a unit and converted into a frame average array, the frame average array is subtracted from the original output data array and converted into an original noise array, and the original noise array is averaged by taking a row / column as a unit and converted into a row / column average array. Subtracting a column / row mean value from each row / column of the original noise array to convert into a row / column stripe array, subtracting a row / column mean value array from the row / column stripe array to convert into a random noise array, and subtracting the random noise array from the original noise array to convert into a row / column stripe noise array. The defect that a traditional model cannot independently analyze row and column noise characteristics is overcome; compared with a traditional noise model, the noise model is more visual, and noise characteristics of the focal plane readout circuit are easy to analyze.

Description

Technical Field

[0001] The invention belongs to the technical field of focal plane circuit testing and image non-uniformity correction, and relates to a noise model for a focal plane readout circuit and an image noise reduction method. Background Art

[0002] As detectors develop towards planar arrays and digital readout, focal plane detector technology has become an integral part of detector technology. The readout circuit of a focal plane detector often has non-uniformity in the row and column directions when outputting images or data, and also exhibits three-dimensional characteristics in time. Therefore, using traditional RMS noise and spatial noise as performance evaluation indicators can no longer characterize this three-dimensional characteristic of noise, and the noise model needs to be improved. The fixed stripe noise in the focal plane readout circuit mainly comes from the non-uniformity between materials, processes, or readout channels, while the stripe noise that varies with time is mainly related to the row and column scanning and frame scanning of the focal plane readout circuit.

[0003] The traditional three-dimensional noise model can preliminarily characterize the three-dimensional characteristics of the noise of the focal plane readout circuit, but the influence between the components is not eliminated during the calculation process, and the temporal and spatial noise of individual rows and columns cannot be calculated. Based on the improved noise model, the original output data is split into row and column noise components, and the grayscale value is mapped by the data array to obtain a denoised image with row and column stripes eliminated, which can provide a new idea for image non-uniformity correction. Summary of the invention

[0004] (I) Purpose of the invention

[0005] The purpose of the present invention is to provide a noise model and image noise reduction method for a focal plane readout circuit, so that the three-dimensional noise characteristics of individual rows and columns and pixels of the focal plane readout circuit can be tested and analyzed, and based on the improved noise model, the row and column stripes in the output image can be eliminated.

[0006] (II) Technical solution

[0007] In order to solve the above technical problems, the present invention provides a noise model for a focal plane readout circuit, which includes: an array and an array operation conversion method; the array includes an original output data array, a frame array, an original noise array, a row / column mean array, a row / column stripe array, a random noise array, and a row and column stripe noise array; the array conversion method includes averaging the original output data array in units of frames and converting it into a frame mean array, converting the original output data array minus the frame mean array into the original noise array, averaging the original noise array in units of rows / columns and converting it into a row / column mean array, converting each row / column of the original noise array minus the column / row mean into a row / column stripe array, converting the row / column stripe array minus the row / column mean array into a random noise array, and converting the original noise array minus the random noise array into a row and column stripe noise array.

[0008] The present invention also provides an image noise reduction method for a focal plane readout circuit, comprising the following steps:

[0009] Step S1: shading the pixels of the focal plane readout circuit, collecting the original output data of F frames, where F≥100;

[0010] Step S2: converting the multi-frame acquisition data into a frame mean array and an original noise array, calculating the frame-to-frame noise, and outputting the original noise image;

[0011] Step S3: the original noise array is converted into a column mean array and a row mean array to obtain a row fringe array and a column fringe array;

[0012] Step S4: The row / column stripe array and the row / column mean array are operated to obtain a random noise array, spatial random noise and time-space random noise are calculated, and a denoised image with row and column stripe fixed noise eliminated is output;

[0013] Step S5: The original noise array and the random noise array are operated to obtain a row and column stripe noise array, the spatial row / column noise and the temporal row / column noise are calculated, and a row and column stripe noise image is output.

[0014] In step S1, the process of obtaining the original output data is as follows:

[0015] After the focal plane readout circuit works, F frames of digital output data are collected through the host computer and the acquisition card. The digital output data is the original output data. Each frame of the collected original output data contains M×N pixels. The output value of each pixel in the data array of the collected multiple frames of original output data is expressed as g(i,j,k), where i=1~M, j=1~N, and k=1~F.

[0016] In step S1, the calculation formula for the total average value S of all pixel outputs is:

[0017]

[0018] In step S2, the acquisition of the frame mean array and the original noise array includes:

[0019] The frame mean array is obtained by averaging all pixel outputs of each frame of the collected multi-frame image data, and the formula is:

[0020]

[0021] The original noise array of the detector is obtained by subtracting the frame average of the corresponding frame from all pixel data of each frame of the collected multi-frame image array. The formula is:

[0022]

[0023] The frame-to-frame noise is calculated by the mean array, and the formula is:

[0024]

[0025] In step S3, the process of obtaining the column / row mean array and the row / column stripe array includes:

[0026] All columns / rows of each frame of the original noise array are averaged to obtain the column / row average array, and the formula is:

[0027]

[0028]

[0029] The row / column stripe array of the focal plane readout circuit is obtained by subtracting the column / row average of the corresponding frame from all rows / columns of each frame of the original noise array, and the formula is:

[0030]

[0031]

[0032] In step S4, the random noise array, spatial random noise and temporal-spatial random noise are used to obtain a denoised image that eliminates row and column stripe fixed noise, including:

[0033] The random noise array is obtained by subtracting the row / column average of the corresponding frame from all the columns / rows of each frame of the row / column stripe array, and the formula is:

[0034]

[0035]

[0036] The total average value of all pixel outputs of random noise is given by:

[0037]

[0038] The spatial random noise and the temporal-spatial random noise are calculated by the random noise array, and the formula is:

[0039]

[0040]

[0041] Through the random noise array, the output values ​​of the pixels in the array are mapped to the grayscale space of the corresponding coordinates, and the obtained image is the denoised image that eliminates the fixed noise of row and column stripes.

[0042] The comparison process between the denoised image and the original image includes: the output values ​​of the pixels in the original output data array are mapped to the grayscale space of the corresponding coordinates, which is the original image; the denoised image is the output values ​​of the pixels in the random noise array.

[0043] In step S5, the process of obtaining the row and column stripe noise array, the spatial row / column noise and the temporal row / column noise, and the row and column stripe noise image includes:

[0044] The row and column stripe noise array is obtained by subtracting the corresponding pixel data in the random noise array from the pixel data in the original noise array. The formula is:

[0045] N stripe (i,j,k)=N(i,j,k)-N random (i,j,k)

[0046] The formula for calculating the total average value of all pixel outputs of the row and column stripe noise array is:

[0047]

[0048] The spatial row / column noise and the temporal row / column noise are calculated by the row and column stripe noise array, and the formula is:

[0049]

[0050]

[0051]

[0052]

[0053] Through the row and column stripe noise array, the output values ​​of the pixels in the array are mapped to the grayscale space of the corresponding coordinates to obtain a noise image of the row and column stripes.

[0054] (III) Beneficial effects

[0055] The noise model and image noise reduction method for the focal plane readout circuit provided by the above technical solution have the following beneficial effects:

[0056] 1. The noise model designed by the present invention adopts the ideas of matrix and mean subtraction, splits the original output data into different arrays, and then calculates the temporal and spatial non-uniformity of different noise components through different arrays. In the process of calculation, the influence of other noise components is eliminated, which is more accurate and has stronger isolation.

[0057] 2. The noise model designed by the present invention calculates noise components through different arrays and can analyze noise characteristics separately in units of frames, pixels, rows and columns, thus making up for the deficiency that traditional models cannot analyze row and column noise characteristics separately.

[0058] 3. The method of the present invention is concise and easy to implement. It can be designed in various programming languages. Based on the idea of ​​matrix, the converted data output can obtain the required noise image and noise reduction image. Compared with the traditional noise model, it is more intuitive and easy to analyze the noise characteristics of the focal plane readout circuit. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 Schematic diagram of the process steps of the method of the present invention.

[0060] Figure 2 It is a schematic diagram of the noise model of the present invention for the focal plane readout circuit.

[0061] Figure 3 It is a schematic diagram of the image noise reduction effect of the image noise reduction method for a focal plane readout circuit in a specific implementation of the present invention.

[0062] Figure 4 It is a schematic diagram of experimental data in a specific implementation of the method of the present invention.

[0063] Figure 5 Figures a and b in the middle are the distribution diagrams of the row and column spatial noise of the focal plane readout circuit respectively. DETAILED DESCRIPTION

[0064] In order to make the purpose, content and advantages of the present invention more clear, the specific implementation methods of the present invention are further described in detail below in conjunction with the drawings and examples.

[0065] like Figure 2As shown, the noise model for the focal plane readout circuit of the present embodiment includes an array and an array operation conversion method; the array includes an original output data array, a frame array, an original noise array, a row / column mean array, a row / column stripe array, a random noise array, and a row and column stripe noise array; the array conversion method includes converting the original output data array into a frame mean array by averaging the original output data array in units of frames, converting the original noise array into a row / column mean array by subtracting the frame mean array, converting the original noise array into a row / column mean array by averaging the original noise array in units of rows / columns, converting each row / column of the original noise array into a row / column stripe array by subtracting the column / row mean, converting the row / column stripe array into a random noise array by subtracting the row / column mean array, and converting the original noise array into a row and column stripe noise array by subtracting the random noise array.

[0066] like Figure 1 As shown, the image noise reduction method for the focal plane readout circuit of this embodiment includes the following steps:

[0067] Step S1: shading the pixels of the focal plane readout circuit, collecting the original output data of F frames, where F≥100;

[0068] Step S2: converting the multi-frame acquisition data into a frame mean array and an original noise array, calculating the frame-to-frame noise, and outputting the original noise image;

[0069] Step S3: the original noise array is converted into a column mean array and a row mean array to obtain a row fringe array and a column fringe array;

[0070] Step S4: The row / column stripe array and the row / column mean array are operated to obtain a random noise array, spatial random noise and time-space random noise are calculated, and a denoised image with row and column stripe fixed noise eliminated is output;

[0071] Step S5: The original noise array and the random noise array are operated to obtain a row and column stripe noise array, the spatial row / column noise and the temporal row / column noise are calculated, and a row and column stripe noise image is output.

[0072] In step S1, the process of obtaining the original output data is as follows:

[0073] After the focal plane readout circuit works, the host computer and the acquisition card collect F frames of digital output data, which are the original output data. Each frame of the collected original output data contains M×N pixels, and the output value of each pixel in the data array of the collected multiple frames of original output data is expressed as g(i,j,k), where i=1~M, j=1~N, and k=1~F.

[0074] The calculation formula for the total average value S of all pixel outputs is:

[0075]

[0076] In step S2, the acquisition of the frame mean array and the original noise array includes:

[0077] The frame mean array is obtained by averaging all pixel outputs of each frame of the collected multi-frame image data. The specific formula is:

[0078]

[0079] The original noise array of the detector is obtained by subtracting the frame average of the corresponding frame from all pixel data of each frame of the collected multi-frame image array. The specific formula is:

[0080]

[0081] The frame-to-frame noise is calculated by the mean array, and the specific formula is:

[0082]

[0083] In step S3, the process of obtaining the column / row mean array and the row / column stripe array includes:

[0084] All columns / rows of each frame of the original noise array are averaged to obtain the column / row average array, and the specific formula is:

[0085]

[0086]

[0087] The row / column stripe array of the focal plane readout circuit is obtained by subtracting the column / row average of the corresponding frame from all rows / columns of each frame of the original noise array. The specific formula is:

[0088]

[0089]

[0090] In step S4, the random noise array, spatial random noise and temporal-spatial random noise are used to eliminate the row and column stripe fixed noise, and the process of obtaining the denoised image includes:

[0091] The random noise array is obtained by subtracting the row / column average value of the corresponding frame from all the columns / rows of each frame of the row / column stripe array, and the specific formula is:

[0092]

[0093]

[0094] The total average value of all pixel outputs of random noise is as follows:

[0095]

[0096] The spatial random noise and the temporal-spatial random noise are calculated by the random noise array, and the specific formula is:

[0097]

[0098]

[0099] Through the random noise array, the output values ​​of the pixels in the array are mapped to the grayscale space of the corresponding coordinates, and the obtained image is a denoised image with the row and column stripe fixed noise eliminated.

[0100] The comparison between the denoised image and the original image is as follows: Figure 3 As shown, the acquisition process includes:

[0101] The output values ​​of the pixels in the original output data array are mapped to the grayscale space of the corresponding coordinates, which is the original image. The denoised image is similar, which is the output value of the pixels in the random noise array.

[0102] In step S5, the process of obtaining the row and column stripe noise array, the spatial row / column noise and the temporal row / column noise, and the row and column stripe noise image includes:

[0103] The row and column stripe noise array is obtained by subtracting the corresponding pixel data in the random noise array from the pixel data in the original noise array. The specific formula is:

[0104] N stripe (i,j,k)=N(i,j,k)-N random (i,j,k)

[0105] The formula for calculating the total average value of all pixel outputs of the row and column stripe noise array is:

[0106]

[0107] The spatial row / column noise and the temporal row / column noise are calculated by the row and column stripe noise array, and the specific formula is:

[0108]

[0109]

[0110]

[0111]

[0112] Through the row and column stripe noise array, the output values ​​of the pixels in the array are mapped to the grayscale space of the corresponding coordinates to obtain a noise image of the row and column stripes.

[0113] It should be further explained that in the specific implementation process, Figure 4 The calculated values ​​of each noise component of the focal plane readout circuit and its component weight are shown in LSB. Experiments show that the model proposed in the present invention can calculate various noise components, among which the spatial noise component accounts for a larger proportion.

[0114] It should be further explained that in the specific implementation process, Figure 5 The distribution diagram of the row / column spatial noise of the focal plane readout circuit is shown in Figure 1. Experiments show that the model proposed in the present invention can calculate the temporal noise of the unit rows and columns.

[0115] The method of the present invention adopts the ideas of matrix and mean subtraction, splits the original output data into different arrays, and then calculates the temporal and spatial non-uniformity of different noise components through different arrays. In the process of calculation, the influence of other noise components is eliminated, which is more accurate and has stronger isolation.

[0116] The method designed by the present invention calculates noise components through different arrays, and can analyze noise characteristics separately in units of frames, pixels, rows and columns, thus making up for the deficiency that traditional models cannot analyze row and column noise characteristics separately.

[0117] The method of the present invention is concise and easy to implement, can be designed in various programming languages, and based on the idea of ​​matrix, the converted data output can obtain the required noise image and noise reduction image. Compared with the traditional noise model, it is more intuitive and easy to analyze the noise characteristics of the focal plane readout circuit.

[0118] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A noise model for a focal plane readout circuit, characterized in that include: Array and array operation conversion mode; the array includes original output data array, frame array, original noise array, row / column mean array, row / column stripe array, random noise array, row and column stripe noise array; The array conversion method includes converting the original output data array into a frame mean array by averaging the original output data array in units of frames, converting the original noise array by subtracting the frame mean array, converting the original noise array into a row / column mean array by averaging the original noise array in units of rows / columns, converting the original noise array into a row / column mean array by subtracting the column / row mean from each row / column, converting the original noise array into a row / column stripe array, converting the row / column stripe array into a random noise array by subtracting the row / column mean array, and converting the original noise array into a row and column stripe noise array by subtracting the random noise array.

2. An image noise reduction method for a focal plane readout circuit using the noise model of claim 1, characterized in that: The following steps are involved: Step S1: shading the pixels of the focal plane readout circuit, collecting the original output data of F frames, where F≥100; Step S2: converting the multi-frame acquisition data into a frame mean array and an original noise array, calculating the frame-to-frame noise, and outputting the original noise image; Step S3: the original noise array is converted into a column mean array and a row mean array to obtain a row fringe array and a column fringe array; Step S4: The row / column stripe array and the row / column mean array are operated to obtain a random noise array, spatial random noise and time-space random noise are calculated, and a denoised image with row and column stripe fixed noise eliminated is output; Step S5: The original noise array and the random noise array are operated to obtain a row and column stripe noise array, the spatial row / column noise and the temporal row / column noise are calculated, and a row and column stripe noise image is output.

3. The image denoising method according to claim 2, characterized in that: In step S1, the process of obtaining the original output data is as follows: After the focal plane readout circuit works, F frames of digital output data are collected through the host computer and the acquisition card. The digital output data is the original output data. Each frame of the collected original output data contains M×N pixels. The output value of each pixel in the data array of the collected multiple frames of original output data is expressed as g(i,j,k), where i=1~M, j=1~N, and k=1~F.

4. The image denoising method according to claim 3, characterized in that: In step S1, the calculation formula for the total average value S of all pixel outputs is:

5. The image denoising method according to claim 4, characterized in that: In step S2, the acquisition of the frame mean array and the original noise array includes: The frame mean array is obtained by averaging all pixel outputs of each frame of the collected multi-frame image data, and the formula is: The original noise array of the detector is obtained by subtracting the frame average of the corresponding frame from all pixel data of each frame of the collected multi-frame image array. The formula is: The frame-to-frame noise is calculated by the mean array, and the formula is:

6. The image denoising method according to claim 5, characterized in that: In step S3, the process of obtaining the column / row mean array and the row / column stripe array includes: All columns / rows of each frame of the original noise array are averaged to obtain the column / row average array, and the formula is: The row / column stripe array of the focal plane readout circuit is obtained by subtracting the column / row average of the corresponding frame from all rows / columns of each frame of the original noise array, and the formula is:

7. The image denoising method according to claim 6, characterized in that: In step S3 and step S4, the random noise array, spatial random noise and temporal-spatial random noise are used to eliminate row and column stripe fixed noise to obtain a denoised image, which includes: The random noise array is obtained by subtracting the row / column average of the corresponding frame from all the columns / rows of each frame of the row / column stripe array, and the formula is: The total average value of all pixel outputs of random noise is given by: The spatial random noise and the temporal-spatial random noise are calculated by the random noise array, and the formula is: Through the random noise array, the output values ​​of the pixels in the array are mapped to the grayscale space of the corresponding coordinates, and the obtained image is the denoised image that eliminates the fixed noise of row and column stripes.

8. The image denoising method according to claim 7, characterized in that: The comparison process between the denoised image and the original image includes: the output values ​​of the pixels in the original output data array are mapped to the grayscale space of the corresponding coordinates, which is the original image; the denoised image is the output values ​​of the pixels in the random noise array.

9. The image denoising method according to claim 8, characterized in that: In step S5, the process of obtaining the row and column stripe noise array, the spatial row / column noise and the temporal row / column noise, and the row and column stripe noise image includes: The row and column stripe noise array is obtained by subtracting the corresponding pixel data in the random noise array from the pixel data in the original noise array. The formula is: N stripe (i,j,k)=N(i,j,k)-N random (i,j,k) The formula for calculating the total average value of all pixel outputs of the row and column stripe noise array is: The spatial row / column noise and the temporal row / column noise are calculated by the row and column stripe noise array, and the formula is: Through the row and column stripe noise array, the output values ​​of the pixels in the array are mapped to the grayscale space of the corresponding coordinates to obtain a noise image of the row and column stripes.

10. An application of the image noise reduction method according to any one of claims 2 to 9 in the technical fields of focal plane circuit testing and image non-uniformity correction.

Images