Hyperspectral compression method based on reference band adjustment

Abstract

The invention relates to a hyperspectral compression method based on reference band adjustment, which comprises the following steps: firstly, calculating the correlation c (i, j) between a to-be-compressed band and a previous band, taking the band with the highest correlation with the to-be-compressed band as correlative band, and taking the correlation as a; the adjacent wave bands are marked as band-1, and the correlation degree of the adjacent wave bands is marked as b; determining whether reference wave band adjustment is needed or not according to the difference value of a and b; determining whether to use the reference band during compression according to the value of a; and performing lossless compression on the to-be-compressed wave band by using the new reference wave band. According to the method, by adjusting a reference wave band, the correlation between the reference wave band and a to-be-predicted wave band is improved, and the correlation between the wave bands is utilized more fully; the reference wave band with low correlation is eliminated, and the negative influence of the reference wave band on the compression effect is reduced; the reference wave band can be multiplexed, and a higher compression ratio can be obtained by combining the existing hyperspectral lossless compression algorithms such as CCSDS and 3D-CALIC.

Description

Technical Field

[0001] The invention relates to the technical field of image processing, and in particular to a hyperspectral compression method based on reference band adjustment. Background Art

[0002] At present, hyperspectral imaging technology has been widely used in agriculture, remote sensing and other fields. With the advancement of technology and social development, the spectral resolution and spatial resolution are constantly improving, the scope of application is expanding, and the amount of hyperspectral image data is growing sharply. Due to the limited storage capacity and data transmission capacity of sensors, the use of effective compression technology can reduce the transmission time and optimize the utilization of channel bandwidth. Unlike two-dimensional natural images, hyperspectral remote sensing images have inter-spectral correlation in addition to spatial correlation. Inter-spectral correlation refers to the redundancy between pixels with the same spatial position and distributed on each band image. Due to the improvement of spectral resolution, the texture of hyperspectral images is relatively fine, so its spatial correlation is poor, but the inter-spectral correlation of hyperspectral images is extremely strong, so the compression of hyperspectral images needs to make full use of inter-spectral correlation.

[0003] However, most compression technologies at this stage usually follow the natural band order of spectral images, that is, they are arranged in the order of imaging wavelengths. However, due to the presence of the same type of absorbing gas, the data features between non-adjacent bands may show stronger inter-spectral correlation than adjacent bands. When the natural band order is used directly for compression, the reference band is the adjacent band. At this time, the reference band is not necessarily the one with the highest correlation with the band to be compressed, so the inter-spectral correlation cannot be well utilized to improve the compression effect. Summary of the invention

[0004] The purpose of the present invention is to provide a hyperspectral compression method based on reference band adjustment, which solves the problem that the existing technology cannot make good use of the inter-spectral correlation to improve the compression effect. It adjusts the reference band to improve the correlation between the reference band and the band to be predicted, and makes more full use of the correlation between the bands; removes the reference band with too low correlation to reduce its negative impact on the compression effect; the reference band can be reused, and one band can be used as a reference band for several bands; combined with hyperspectral lossless compression algorithms such as CCSDS and 3D-CALIC, a higher compression rate can be obtained.

[0005] To achieve the above object, the technical solution of the present invention is:

[0006] A hyperspectral compression method based on reference band adjustment, which transmits the hyperspectral image data of a hyperspectral imaging analyzer to a storage medium, reads the hyperspectral image data in the storage medium, extracts and integrates the hyperspectral image data, and then outputs the hyperspectral image data to a compression system, compresses the hyperspectral image data, and finally stores and / or transmits the compressed hyperspectral image data; the specific steps of compressing the hyperspectral image data are as follows:

[0007] The first step is to calculate the correlation coefficient matrix to store the correlation between bands.

[0008] Calculate the correlation between all bands in the hyperspectral image data and store it as the correlation coefficient matrix CorrM. The calculation formula is as follows:

[0009]

[0010] Among them, p i (x,y) and p j (x, y) are the corresponding pixel values ​​of the i-th and j-th spectral segments, respectively. and Represent the pixel mean of the corresponding band, L and W represent the length and width of the image respectively. Set the diagonal of CorrM to negative infinity to avoid selecting the current band when adjusting the reference band;

[0011] The second step is to traverse CorrM and find the correlated_band with the highest correlation with the current band. The correlated_band is the index of the maximum value found in CorrM[band,:band+1].

[0012] The third step is to get the correlation coefficient a between band and correlated_band, and the correlation coefficient b between the current band band and its adjacent band band-1:

[0013] a=CorrM[band,correlated_band]

[0014] b = CorrM[band,band-1];

[0015] Step 4: Input the continuity breakdown threshold th and the reference spectrum usage threshold t;

[0016] Step 5: Data combination after threshold judgment;

[0017] The sixth step is to use the new reference band to compress the hyperspectral image data.

[0018] In the second step, only the band with the highest correlation in the band before the band to be compressed is found as correlated_band, ensuring that the reference band has been decompressed when the current band is decompressed.

[0019] The fifth step of threshold judgment is to add a reference band using threshold t judgment link. When the correlation a of the highest correlation band is lower than the reference spectrum using threshold t, the 0 array is used as the reference band.

[0020] Added the continuity breakdown threshold th judgment link. When the correlation ab is lower than the continuity breakdown threshold th, the reference band adjustment is not performed.

[0021] When the correlation ab is greater than the continuity breakdown threshold th, the correlated_band is adjusted to a new reference band.

[0022] The advantages of the present invention are: 1. While ensuring smooth decompression, the present invention adjusts the reference band to the band with the highest correlation with the band to be predicted, so that the hyperspectral image data can make more full use of the inter-spectral correlation during compression; 2. The present invention solves the problem that one band can only be used as a reference band for a certain band, and realizes the reuse of reference bands; 3. The method of the present invention is scalable and can be combined with a variety of hyperspectral image data compression methods to improve compression efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of the processing method of the present invention;

[0024] Figure 2 It is a schematic diagram of an implementation device of the present invention;

[0025] Figure 3 It is a reference band adjustment structure diagram of the present invention. DETAILED DESCRIPTION

[0026] The present invention is further described below in conjunction with the accompanying drawings, which are only used for exemplary description and cannot be understood as limiting the present invention.

[0027] In order to more concisely describe the present embodiment, some parts known to those skilled in the art but not related to the main content of the present invention are omitted in the drawings or descriptions. In addition, for the convenience of description, some parts in the drawings are omitted, enlarged or reduced, but they do not represent the size or entire structure of the actual product.

[0028] The invention discloses a hyperspectral compression method based on reference band adjustment, which transmits the hyperspectral image data of a hyperspectral imaging analyzer to a storage medium, and transmits the hyperspectral image data of the hyperspectral imager to a storage medium SDRAM through a high-speed interface TLK2711. The hyperspectral imaging analyzer is intended to generate spectral imaging data of multiple bands, including a visible band and an infrared shortwave band. SDRAM, as a storage medium for data, receives and stores the data.

[0029] The hyperspectral image data in the storage medium is read from the SDRAM, extracted and integrated, and then output to the compression system to compress the hyperspectral image data, and finally the compressed hyperspectral image data is stored and / or transmitted.

[0030] The specific steps for compressing hyperspectral image data are as follows:

[0031] The first step is to calculate the correlation coefficient matrix to store the correlation between bands.

[0032] Calculate the correlation between all bands in the hyperspectral image data and store it as the correlation coefficient matrix CorrM. The calculation formula is as follows:

[0033]

[0034] Among them, p i (x,y) and p j (x, y) are the corresponding pixel values ​​of the i-th and j-th spectral segments, respectively. and Represent the pixel mean of the corresponding band, L and W represent the length and width of the image respectively. Set the diagonal of CorrM to negative infinity to avoid selecting the current band when adjusting the reference band;

[0035] The second step is to traverse CorrM and find the correlated_band with the highest correlation with the current band. The correlated_band is the index of the maximum value found in CorrM[band,:band+1].

[0036] In the second step, only the band with the highest correlation in the band before the band to be compressed is found as correlated_band, ensuring that the reference band has been decompressed when the current band is decompressed.

[0037] The third step is to get the correlation coefficient a between band and correlated_band, and the correlation coefficient b between the current band band and its adjacent band band-1:

[0038] a=CorrM[band,correlated_band]

[0039] b = CorrM[band,band-1];

[0040] Step 4: Input the continuity breakdown threshold th and the reference spectrum usage threshold t;

[0041] Step 5: Data combination after threshold judgment;

[0042] A reference band threshold t judgment link is added. When the correlation a of the highest correlation band is lower than the reference spectrum threshold t, the 0 array is used as the reference band, that is, the reference band is not used. At this time, the reference band will have a counter-effect on compression.

[0043] A continuity breakdown threshold th judgment link is added. When the correlation ab is lower than the continuity breakdown threshold th, the reference band adjustment is not performed, because the reference band correlation is already very high at this time, and the gain brought by adjusting the reference band is very small.

[0044] When the correlation ab is greater than the continuity breakdown threshold th, the correlated_band is adjusted to a new reference band.

[0045] The band adjustment structure shown in this step is as follows Figure 3 As shown, the updated reference bands and the bands to be predicted are combined in pairs, two bands are compressed each time, and the compressed code stream of the latter band is stored and spliced ​​again during decompression.

[0046] The sixth step is to use the new reference band to compress the hyperspectral image data.

[0047] Finally, the binary code stream obtained after compression in the sixth step is stored in SDRAM, waiting to be sent to the host computer. Figure 2 The implementation structure diagram of the device of the invention is given.

[0048] The hyperspectral image data is firstly used to calculate the correlation between bands in pairs, and the reference band is adjusted according to the correlation. Then the data enters the predictor and encoder, and finally a binary code stream is output.

[0049] The specific embodiments are as follows:

[0050] By applying the method provided by the present invention, the hyperspectral data sets generated by AVIRIS (Airborne Visible Infrared Imaging Spectrometer) sensor: Indian Pines, Yellowstone Un_Sc0, Yellowstone Un_Sc3 and Hawaii_Sc01 are taken as example data, and the experimental device is referred to. The bit rate bpp (bit per pixel) is used as the measurement standard, which is the compressed code stream length / number of pixels; the difference in compression effect before and after using the method is calculated, as shown in Table 1.

[0051] Table 1 Difference in compression effect before and after using the method of the present invention

[0052] Dataset name Natural sort compression The method of the present invention compresses Improvement rate (%) Indian_pines 6.9651 6.7728 2.8 Yellowstone_sc00 7.6107 7.5423 0.9 Hawaii_sc01 3.1467 3.0862 2.0 Maine_sc10 3.2667 3.2303 1.1

[0053] As shown in Table 1, for hyperspectral image data, the method proposed in this paper improves the compression performance of natural bands by 0.9%-2.8%, which has a more obvious improvement effect, and the data has a higher compression ratio after compression.

[0054] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of the present invention. That is, any equivalent changes and modifications made according to the content of the patent application scope of the present invention should be within the technical scope of the present invention.

Claims

1. A hyperspectral compression method based on reference band adjustment, which transmits the hyperspectral image data of a hyperspectral imaging analyzer to a storage medium, reads the hyperspectral image data in the storage medium, extracts and integrates the data, and then outputs the data to a compression system to compress the hyperspectral image data, and finally stores and / or transmits the compressed hyperspectral image data; Features: The specific steps for compressing hyperspectral image data are as follows: The first step is to calculate the correlation coefficient matrix to store the correlation between the bands; Calculate the correlation between all bands in the hyperspectral image data and store it as the correlation coefficient matrix CorrM. The calculation formula is as follows: Among them, p i (x,y) and p j (x, y) are the corresponding pixel values ​​of the i-th and j-th spectral segments respectively. and Represent the pixel mean of the corresponding band, L and W represent the length and width of the image respectively. Set the diagonal of CorrM to negative infinity to avoid selecting the current band when adjusting the reference band; The second step is to traverse CorrM and find the correlated_band with the highest correlation with the current band. The correlated_band is the index of the maximum value found in CorrM[band,:band+1]. The third step is to get the correlation coefficient a between band and correlated_band, and the correlation coefficient b between the current band band and its adjacent band band-1: a=CorrM[band,correlated_band] b = CorrM[band,band-1]; Step 4: Input the continuity breakdown threshold th and the reference spectrum usage threshold t; Step 5: Data combination after threshold judgment; The sixth step is to use the new reference band to compress the hyperspectral image data.

2. The hyperspectral compression method based on reference band adjustment according to claim 1, characterized in that: In the second step, only the band with the highest correlation in the band before the band to be compressed is found as correlated_band, ensuring that the reference band has been decompressed when the current band is decompressed.

3. The hyperspectral compression method based on reference band adjustment according to claim 1, characterized in that: The fifth step of threshold judgment is to add a reference band using threshold t judgment link. When the correlation a of the highest correlation band is lower than the reference spectrum using threshold t, the 0 array is used as the reference band. Added the continuity breakdown threshold th judgment link. When the correlation ab is lower than the continuity breakdown threshold th, the reference band adjustment is not performed. When the correlation ab is greater than the continuity breakdown threshold th, the correlated_band is adjusted to a new reference band.

Images