Video-level spectral imaging system and imaging method based on compressed sensing

Abstract

The invention relates to the technical field of images, and particularly provides a video-level spectral imaging system and imaging method based on compressed sensing, and the video-level spectral imaging system based on compressed sensing is used for shooting a target to obtain an original multi-spectral image. Each pixel in the original multispectral image comprises spectral information of one wave band, the spectral information corresponds to the spatial position of a spectral channel in the light splitting element, and the original multispectral image is subjected to convolution operation to obtain a preliminary pseudo panchromatic image; performing sharpening operation on each pixel of the preliminary pseudo panchromatic image to obtain a pseudo panchromatic image; obtaining a demosaicing residual error based on the original multispectral image and the pseudo panchromatic image; and finally, restoring the pseudo panchromatic image by using the demosaiced residual error to obtain a demosaiced single-spectrum image of each wave band. The spectrum channels are arranged on the target surface of the light splitting element in a pseudo-random mode, and all spectrum images are accurately and efficiently restored.

Description

Technical Field

[0001] The invention belongs to the field of image technology, and in particular relates to a video-level spectral imaging system and an imaging method based on compressed sensing. Background Art

[0002] Spectral imaging technology can obtain two-dimensional spatial information and one-dimensional spectral information of an object. Spatial information can determine the external features of an object, such as its shape and size, while spectral information can analyze the internal features of an object, such as its composition and content. The spectral characteristics of spectral imaging technology make it widely used in many fields. Traditional spectral imaging mostly requires scanning to obtain three-dimensional spatial and spectral data. The data acquisition time is long, and the hardware system usually has scanning moving parts. As a new type of spectral imaging implementation method, snapshot spectral imaging technology can simultaneously obtain the spatial information and spectral information of the target scene, obtain a spectral data cube, and is small in size and has low preparation and maintenance costs.

[0003] The core spectroscopic element of snapshot spectral imaging technology is a mosaic multi-spectral spectroscopic film. Each spectral channel is arranged periodically on the target surface of the spectroscopic element. This arrangement has high production efficiency and many methods for image restoration. However, due to the limitation of the Nyquist-Shanon sampling theorem, the original resolution is low, the measured samples have limitations in restoring the original image signal and most of the unimportant parameters are discarded. The image restoration results have visual ringing artifacts. At the same time, due to the periodicity of the spectral channel arrangement, it has poor robustness to system noise, environmental noise, and electronic noise. There are many problems in the restoration of the measured signal of the sparse solution reconstruction of the underdetermined linear system. Therefore, a new method is needed to accurately and efficiently restore multi-spectral images. Summary of the invention

[0004] To achieve the above object, the technical solution created by the present invention is implemented as follows:

[0005] The present invention discloses a video-level spectral imaging system based on compressed sensing, including an optical lens, a spectral element, an imaging chip and a signal processing module. The target surface of the spectral element is arrayed with a plurality of spectral units, each spectral unit includes a×a randomly arranged spectral channels, each spectral channel corresponds to a pixel on the surface of the imaging chip one by one, and the a×a spectral channels in each spectral unit are named in ascending order according to the working wavelength. in, The secondary peak wavelength of band1 is smaller than the main peak wavelength of band1.

[0006] The present invention also discloses a video-level spectral imaging method based on compressed sensing. After photographing a target using the above-mentioned imaging system, a high-resolution reconstruction of the multispectral image obtained by photographing is performed. The imaging method comprises the following steps:

[0007] S1: Using the video-level spectral imaging system based on compressed sensing to shoot the target, obtaining an original multispectral image, each pixel in the original multispectral image contains spectral information of a band, which respectively corresponds to the spatial position of the spectral channel in the spectroscopic element, and the original multispectral image is convolved to obtain a preliminary pseudo-full-color image;

[0008] S2: performing a sharpening operation on each pixel of the preliminary pseudo-full-color image to obtain a pseudo-full-color image;

[0009] S3: Based on the original multispectral image in step S1, the pseudo-panchromatic image in step S2 obtains a demosaiced residual;

[0010] S4: Use the demosaicing residual to restore the pseudo-panchromatic image and obtain the demosaiced single-spectrum image of each band.

[0011] Furthermore, the original multispectral image I in step S1 raw After the mean convolution kernel F M After the convolution operation, the preliminary pseudo full-color image I is obtained. PPI,blur :I PPI,blur =I raw *F M .

[0012] Furthermore, in step S2, sharpening coefficients in eight directions around each target pixel in the preliminary pseudo-full-color image are calculated respectively to form a sharpening ratio, and the sharpening ratio is a 3×3 matrix with a central value of 0.

[0013] Furthermore, the sharpening weight of the target pixel in the xth column and yth row in the preliminary pseudo-full-color image is F sharpen,x,y , F sharpen,x,y The sharpening coefficient in the mth column and nth row of the matrix The values ​​are:

[0014]

[0015] in, Represents the original multispectral image I raw The pixel value of the target pixel at the coordinate (x, y), Represents the original multispectral image I raw In the adjacent band period of the target pixel, the pixel value of the pixel at the coordinate (x', y') in the same band as the target pixel.

[0016] Further, in step S3, the preliminary pseudo full-color image I is sharpened by the sharpening ratio of each target pixel. PPI,blur Perform pixel-by-pixel sharpening to obtain the pixel value of the target pixel at the coordinate (x, y)

[0017] Further, The calculation formula is as follows:

[0018]

[0019] in, Represents the original multispectral image I raw The pixel value at the coordinates (x+m-2,y+n-2), Represents the preliminary pseudo-full-color image I PPI,blur The pixel value at coordinates (x+m-2,y+n-2).

[0020] Further, it is characterized in that: by copying the sharpening weight F sharpen Get the demosaicing weight F sharpen ', using the demosaicing weight F sharpen ' Calculate the demosaicing residual for pixels corresponding to each band separately.

[0021] Furthermore, the demosaicing residual of the pixel at coordinate (x, y) in band Z is Calculated by the following formula:

[0022]

[0023] in, Represents the original multispectral image I raw The pixel value at the coordinates (x+j-3,y+k-3), Represents the pseudo-full-color image I PPI The pixel value at the coordinates (x+j-3,y+k-3), Indicates whether the band corresponding to the pixel at the coordinate (x, y) in the processed image belongs to the Z band. H is the interpolation weight factor, Represents the interpolation weight factor F H The value of the jth column and the kth row in F H is the interpolation weight factor, expressed as:

[0024]

[0025] Furthermore, step S4 is specifically as follows: adding the demosaiced residual of the pixels of band Z obtained in S3 to the pixel values ​​of band Z in the pseudo-full-color image to obtain the demosaiced pixel values ​​of band Z to form a single-spectrum image of band Z.

[0026] Compared with the prior art, the invention can achieve the following beneficial effects:

[0027] The present invention creates a video-level spectral imaging system and imaging method based on compressed sensing, which breaks the periodicity of the arrangement of mosaic multi-spectral channels on the target surface. The spectral channels are arranged pseudo-randomly on the target surface, avoiding the possibility that sufficient details may not be captured during the later sampling process. The randomly arranged spectral channels can ensure that as many details of the original image as possible can be retained during sampling. At the same time, the original image is restored using the method of the present invention, which can improve the spatial information resolution of the restored image and improve the image restoration degree. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawings constituting part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments and descriptions of the present invention are used to explain the present invention and do not constitute an improper limitation on the present invention. In the drawings:

[0029] Figure 1 It is a flow chart of the video-level spectral imaging method based on compressed sensing of the present invention;

[0030] Figure 2 It is a schematic diagram of the arrangement of spectral channels of the target surface of the spectroscopic element of the present invention;

[0031] Figure 3 Schematic diagram of the band distribution of the original multispectral image of the present invention;

[0032] Figure 4 (a) is the weight matrix of the sharpening coefficient just above the target pixel of the present invention, Figure 4 (b) is the weight matrix of pixels in the same band as the target pixel in adjacent band periods;

[0033] Figure 5 (a) is the weight matrix of the sharpening coefficient of the upper right corner of the target pixel of the present invention, Figure 5 (b) is the weight matrix of pixels in the same band as the target pixel in adjacent band periods;

[0034] Figure 6 The figure is a schematic diagram of the present invention performing binary mask processing on the original multispectral image. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the invention more clear, the invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described here are only used to explain the invention and do not constitute a limitation of the invention.

[0036] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", etc. may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0037] In the description of the invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installation", "connection" and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the invention can be understood according to specific circumstances.

[0038] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0039] The present invention provides a video-level spectral imaging system based on compressed sensing. The imaging system comprises an optical lens, a spectroscopic element, a bandpass filter, an imaging chip and a signal processing module. The spectral information of a measured object is obtained through the optical lens, and the spectrum of the measured object is split by the spectroscopic element. A plurality of array-arranged spectroscopic units are arranged on the target surface of the spectroscopic element. Each spectroscopic unit comprises a×a spectral channels, each spectral channel corresponds to a pixel on the imaging chip one by one, and the positions of the spectral channels in each spectroscopic unit are randomly arranged. After the imaging chip receives the spectral information of the measured object, the signal processing module is used to process the information. Compared with the periodic arrangement of the spectral channels, the system can avoid the loss of a large amount of information when restoring the resolution due to overlap or omission between pixels after the multi-spectral image is captured.

[0040] Furthermore, the a×a spectral channels in each spectroscopic unit are named band1, band2, ..., band a2 , where band a2The secondary peak wavelength of band1 is smaller than the primary peak wavelength of band2. The spectral response curve of each spectral channel has a primary peak and a secondary peak. The response of the secondary peak is filtered out by a bandpass filter. Therefore, the spectroscopic element designed in the present invention has a primary peak and a secondary peak. a2 The secondary peak wavelength of band1 is smaller than the main peak wavelength of band1, so as to avoid filtering out the light of band1 channel when the bandpass filter is filtering, resulting in the lack of response of band1 channel in the imaging system. 2 The spectral channels correspond to a 2 bands, namely z1, z2, ..., za 2 , z1, z2, ..., za 2 Represents a band respectively, z1, z2, ..., za 2 Any band in the multispectral image can be represented by Z. Each pixel in the multispectral image contains the spectral information of a band Z. 2 Pixels constitute a band period, which is an a×a matrix.

[0041] Taking a=3 as an example, each spectroscopic unit on the spectroscopic element includes 3×3 spectral channels. These 9 spectral channels are named band1, band2, ..., band9 in ascending order according to the working wavelength. The spectral channels of each spectroscopic unit on the target surface of the spectroscopic element are randomly arranged, and the arrangement is as follows: Figure 2 In the multispectral image obtained by the imaging system, the 9 bands Z corresponding to the 9 spectral channels are represented as z1, z2, ..., z9, respectively. The 9 pixels corresponding to the 9 bands Z are a band period, which is a 3×3 matrix. The pixel arrangement in the 9-band multispectral image is as follows: Figure 3 As shown, pixels corresponding to 9 bands constitute a band period, and pixels corresponding to different bands in each band period are arranged differently.

[0042] like Figure 1 As shown, the video-level spectral imaging method based on compressed sensing of the present invention specifically includes the following steps:

[0043] S1: Use the video-level spectral imaging system based on compressed sensing to take a picture of the object to be measured and obtain the original multispectral image I raw , the original multispectral image I raw After the mean convolution kernel F M After the convolution operation, the preliminary pseudo full-color image I is obtained. PPI,blur , preliminary pseudo-full-color image I PPI,blur Calculated by formula (1):

[0044] I PPI,blur =I raw *FM (1)

[0045] Where a can be obtained by the number of spectral channels in each spectroscopic unit, and the number of spectral channels in each spectroscopic unit is a 2 ,*denotes the convolution operation, F M is a matrix of size a filled with all ones:

[0046]

[0047] When a=3,

[0048] S2: Preliminary pseudo full-color image I obtained in step S1 PPI,blur , sharpen each pixel to obtain the pseudo full-color image I PPI The specific steps are as follows: First, predict the preliminary pseudo-full-color image I PPI,blur The sharpening ratio of each pixel in the image is: for the target pixel at the coordinate (x, y) to be processed, the value at the target pixel is set to 0, and the sharpening coefficients in the eight directions around it together constitute the sharpening ratio F sharpen,x,y , represented as a 3×3 matrix:

[0049]

[0050] in, Indicates the sharpening ratio F sharpen,x,y The sharpening coefficients in the matrix of m columns and n rows, m = 1, 2, 3, n = 1, 2, 3, for example, Represents the sharpening factor just above the target pixel, that is, the 2nd column and the 1st row.

[0051] In the present invention, the weight matrix of the sharpening coefficients directly above and below the target pixel at (x, y) is designed as a 2×3 matrix, and the weight matrix of the sharpening coefficients directly to the left and right is designed as a 3×2 matrix. For the weight matrix of the sharpening coefficients directly above the target pixel, the target pixel position is in the 2nd row and 2nd column of the matrix, for the weight matrix of the sharpening coefficients directly below the target pixel, the target pixel position is in the 1st row and 2nd column of the matrix, for the weight matrix of the sharpening coefficients directly to the left of the target pixel, the target pixel position is in the 2nd row and 2nd column of the matrix, and for the weight matrix of the sharpening coefficients directly to the right of the target pixel, the target pixel position is in the 2nd row and 1st column of the matrix. The weight value of the target pixel position in the weight matrix is ​​set to 4, the weight values ​​of the three positions slightly farther from the target pixel position are set to 2, and the weight values ​​of the two positions farthest from the target pixel position are set to 1. In the adjacent band period, the pixel in the same band as the target pixel has its coordinates represented as (x', y'). For the pixel at (x', y'), its weight matrix is ​​arranged in the same way as the target pixel, but the weight values ​​are all set to the opposite of the weight value of the target pixel, that is, -4, three -2s and two -1s respectively. It should be noted that when calculating the sharpening coefficient directly above the target pixel, the adjacent band period refers to the band period directly above the band period where the target pixel is located, when calculating the sharpening coefficient directly to the left of the target pixel, the adjacent band period refers to the band period directly to the left of the band period where the target pixel is located, when calculating the sharpening coefficient directly to the right of the target pixel, the adjacent band period refers to the band period directly to the right of the band period where the target pixel is located, and when calculating the sharpening coefficient directly below the target pixel, the adjacent band period refers to the band period directly below the band period where the target pixel is located. This design method can enhance the weight value of the target pixel and weaken the surrounding weight values. This method has been fully verified and has effectiveness and reliability. The sharpening factor is directly above the target pixel at (x, y). As an example, the weight matrix Figure 4 As shown in (a), the target pixel position is in the 2nd row and 2nd column of the weight matrix. The weight value of the target pixel position is set to 4, the weight values ​​of the left, top and right of the target pixel position are set to 2, and the weight values ​​of the top left and top right farthest from the target pixel position are set to 1. Then, the weight matrix of the pixel at (x', y') corresponding to it is as follows: Figure 4 (b) as shown.

[0052] For the sharpening coefficients in the four diagonal directions of the target pixel at (x, y), i.e., the sharpening coefficients of the upper left, upper right, lower left, and lower right of the target pixel, the weight matrix is ​​designed as a 3×3 matrix, but only 6 valid weight values ​​are retained, and the remaining weight values ​​are set to 0. For the weight matrix of the sharpening coefficient of the upper left of the target pixel, the target pixel position is in the 3rd row and 3rd column of the weight matrix, for the weight matrix of the sharpening coefficient of the upper right of the target pixel, the target pixel position is in the 3rd row and 1st column of the weight matrix, for the weight matrix of the sharpening coefficient of the lower left of the target pixel, the target pixel position is in the 3rd column and 1st row of the weight matrix, and for the weight matrix of the sharpening coefficient of the lower right of the target pixel, the target pixel position is in the 1st row and 1st column of the weight matrix. The weight value of the target pixel position in the weight matrix is ​​set to 4, the weight values ​​of the three positions adjacent to the target pixel are set to 2, the weight values ​​of the two non-corner positions of the five positions slightly away from the target pixel are set to 1, and the weight values ​​of the remaining three positions are set to 0. In the adjacent band period, the pixel in the same band as the target pixel has coordinates expressed as (x', y'), and its weight matrix is ​​arranged in the same way as the target pixel at (x, y). However, when designing its weight matrix, the weight values ​​are all set to the opposite of the weight values ​​of the target pixels, that is, -4, three -2s and two -1s, and the remaining weight values ​​are 0. It should be noted that when calculating the sharpening coefficient at the upper right of the target pixel, the adjacent band period refers to the band period at the upper right of the band period where the target pixel is located. When calculating the sharpening coefficient at the upper left of the target pixel, the adjacent band period refers to the band period at the upper left of the band period where the target pixel is located. When calculating the sharpening coefficient at the lower left of the target pixel, the adjacent band period refers to the band period at the lower left of the band period where the target pixel is located. When calculating the sharpening coefficient at the lower right of the target pixel, the adjacent band period refers to the band period at the lower right of the band period where the target pixel is located. Take the sharpening coefficient at the upper right of the target pixel as the example. As an example, the weight matrix Figure 5 As shown in (a), the target pixel position is in the 3rd row and 1st column of the weight matrix. The weight value of the target pixel position is set to 4, the weight values ​​of the pixel just above, to the right and to the right of the target pixel position are set to 2, the weight values ​​of the two non-corner positions slightly away from the target pixel position are set to 1, and the weight values ​​of the remaining three positions are set to 0. Then, the weight matrix of the pixel at (x', y') is as follows: Figure 5 (b) as shown.

[0053] The 6 valid weight values ​​in the weight matrix can be used to calculate the sharpening coefficient of the mth column and the nth row in the sharpening ratio. Values:

[0054]

[0055] in, Represents the original multispectral image I raw The pixel value of the pixel at the coordinate (x, y) (i.e. the target pixel), Represents the original multispectral image I raw The pixel value of the pixel at coordinates (x',y').

[0056] According to the sharpening coefficients calculated in 8 directions, the sharpening ratio F can be obtained. sharpen,x,y The matrix, using the sharpening weight F sharpen,x,y For the preliminary pseudo-panchromatic image I PPI,blur Sharpening is performed. The sharpening process is as follows: the preliminary pseudo full-color image I PPI,blur The pixel value of the pixel with coordinates (x,y) After calculation by formula (3), the pseudo full-color image I can be obtained: PPI All pixel values, pseudo full-color image I PPI The pixel value at coordinate (x,y) The calculation formula is expressed as follows:

[0057]

[0058] in, Represents the original multispectral image I raw The pixel value at the coordinates (x+m-2,y+n-2), Represents the preliminary pseudo-full-color image I PPI,blur The pixel value at coordinates (x+m-2,y+n-2).

[0059] S3: First, by sharpening the weight F in step S2 sharpen,x,y The sharpening coefficient in each direction is copied to the outer layer to obtain the demosaicing ratio F sharpen ', the demosaicing weight F of the pixel at position (x, y) sharpen ' ,x,y for:

[0060]

[0061] in, F is the de-mosaicing weight sharpen ' ,x,y The value in column j and row k can be understood as the value of the demosaicing weight of the target pixel at position (x, y) in the first column and the first row. for The value in the second column, first row for The value in the third column, first row for And so on.

[0062] Then, design a binary mask B for band Z Z , To process whether the band corresponding to the pixel at the coordinate (x, y) in the processed image belongs to the Z band, Defined by:

[0063]

[0064] That is, when the pixel at coordinates (x, y) belongs to band Z, The value is 1, otherwise it is 0. Take the binary mask processing of the original multispectral image as an example, Figure 6 As shown in the figure, after an original multispectral image containing 9 bands (z1, z2, ..., z9) is processed by binary mask, 9 single-spectral images containing only one band can be obtained. Figure 6 The single spectral images of bands z1, z5 and z9 are shown in FIG. The processing of pseudo-panchromatic images is the same as that of the pseudo-panchromatic images, which will not be repeated here.

[0065] Finally, using the binary mask B Z And demosaicing weight F sharpen ' ,x,y , respectively calculate the demosaicing residual of the pixel at coordinate (x, y) corresponding to each band Z in the pseudo-full-color image The calculation formula is as follows:

[0066]

[0067] in, Represents the original multispectral image I raw The pixel value at the coordinates (x+j-3,y+k-3), Represents the pseudo-full-color image I PPI The pixel value at the coordinates (x+j-3,y+k-3), It represents the screening of whether the band corresponding to the pixel at the coordinate (x, y) in the processed image belongs to band Z. It can be understood as: In Represents the original multispectral image I raw The pixel at the coordinate (x+j-3, y+k-3) is screened to see if the band corresponding to the pixel belongs to band Z. In Represents the pseudo-full-color image I PPI The pixel at the coordinate (x+j-3, y+k-3) is screened to see whether the band corresponding to the pixel belongs to band Z; F H is the interpolation weight factor, Represents the interpolation weight factor F H The value of the jth column and the kth row in FH It is represented by the following matrix:

[0068]

[0069] Where a can be obtained by the number of spectral channels in each spectroscopic unit, and the number of spectral channels in each spectroscopic unit is a 2 indivual.

[0070] When a=3, F H for:

[0071]

[0072] Formula (5) is used to calculate the demosaicing residual of the pixel at the coordinate (x, y) corresponding to each band Z: The demosaicing residual I of the pixel corresponding to each band Z in the pseudo-full-color image can be obtained respectively residual,z .

[0073] S4: De-mosaicing residual I obtained from S3 residual,z , the pseudo full-color image is restored by the following formula:

[0074] I demosaiced,Z =I PPI +I residual,Z (6)

[0075] I demosaiced,Z is the single spectrum image after demosaicing of band Z. It can be understood that the demosaicing residual of each band image is combined with the pseudo full-color image I PPI The pixel values ​​of the corresponding bands in the image are added to obtain the single-spectrum image after demosaicing of each band. Repeat a 2 In step S4, we can get a 2 The demosaiced single-spectrum image is used to achieve high-resolution reconstruction of the original multispectral image.

[0076] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A video-level spectral imaging system based on compressed sensing, comprising an optical lens, a spectroscopic element, an imaging chip and a signal processing module, characterized in that: The target surface of the spectroscopic element is arrayed with a plurality of spectroscopic units, each of which includes a×a randomly arranged spectral channels, each spectral channel corresponds one-to-one to a pixel on the surface of the imaging chip, and the a×a spectral channels in each spectroscopic unit are named band1, band2, ..., band1 in ascending order according to the working wavelength. a2 , where band a2 The secondary peak wavelength of band1 is smaller than the main peak wavelength of band1.

2. A video-level spectral imaging method based on compressed sensing using the imaging system of claim 1, characterized in that: The following steps are involved: S1: photographing a target using the imaging system to obtain an original multispectral image, wherein each pixel in the original multispectral image contains spectral information of a band, which respectively corresponds to the spatial position of the spectral channel in the spectroscopic element, and the original multispectral image is subjected to a convolution operation to obtain a preliminary pseudo-full-color image; S2: performing a sharpening operation on each pixel of the preliminary pseudo-full-color image to obtain a pseudo-full-color image; S3: Based on the original multispectral image in step S1, the pseudo-panchromatic image in step S2 obtains a demosaiced residual; S4: Use the demosaicing residual to restore the pseudo-panchromatic image and obtain the demosaiced single-spectrum image of each band.

3. The video-level spectral imaging method based on compressed sensing according to claim 2, characterized in that: The original multispectral image I in step S1 raw After the mean convolution kernel F M After the convolution operation, the initial pseudo full-color image I is obtained. PPI,blur :I PPI ,blur =I raw *F M .

4. The video-level spectral imaging method based on compressed sensing according to claim 2, characterized in that: In step S2, sharpening coefficients in eight directions around each target pixel in the preliminary pseudo-full-color image are calculated respectively to form a sharpening ratio, which is a 3×3 matrix with a central value of 0.

5. The video-level spectral imaging method based on compressed sensing according to claim 4, characterized in that: The sharpening weight of the target pixel in the xth column and yth row in the preliminary pseudo-full-color image is F sharpen,x,y , F sharpen,x,y The sharpening coefficient in the mth column and nth row of the matrix The values ​​are: in, Represents the original multispectral image I raw The pixel value of the target pixel at the coordinate (x, y), Represents the original multispectral image I raw In the adjacent band period of the target pixel, the pixel value of the pixel at the coordinate (x', y') in the same band as the target pixel.

6. The video-level spectral imaging method based on compressed sensing according to claim 5, characterized in that: In step S3, the preliminary pseudo full-color image I is calibrated by the sharpening ratio of each target pixel. PPI,blur Perform pixel-by-pixel sharpening to obtain the pixel value of the target pixel at the coordinate (x, y) 7. The video-level spectral imaging method based on compressed sensing according to claim 6, characterized in that: The calculation formula is as follows: in, Represents the original multispectral image I raw The pixel value at the coordinates (x+m-2,y+n-2), Represents the preliminary pseudo-full-color image I PPI,blur The pixel value at coordinates (x+m-2,y+n-2).

8. The video-level spectral imaging method based on compressed sensing according to claim 7, characterized in that: By copying the sharpening weight F sharpen Get the demosaicing weight F sharpen ', using the demosaicing weight F sharpen ' Calculate the demosaicing residual for pixels corresponding to each band separately.

9. The video-level spectral imaging method based on compressed sensing according to claim 8, characterized in that: The demosaiced residual of the pixel at coordinate (x,y) in band Z Calculated by the following formula: in, Represents the original multispectral image I raw The pixel value at the coordinates (x+j-3,y+k-3), Represents the pseudo-full-color image I PPI The pixel value at the coordinates (x+j-3,y+k-3), Indicates whether the band corresponding to the pixel at the coordinate (x, y) in the processed image belongs to the Z band. H is the interpolation weight factor, Represents the interpolation weight factor F H The value of the jth column and the kth row in F H is the interpolation weight factor, expressed as:

10. The video-level spectral imaging method based on compressed sensing according to claim 9, characterized in that: Step S4 is specifically as follows: adding the demosaiced residual of the pixel of band Z obtained in S3 to the pixel value of band Z in the pseudo full-color image to obtain the demosaiced pixel value of band Z to form a single-spectrum image of band Z.

Images