CN111076815B - A method for correcting non-uniformity of hyperspectral images - Google Patents

Abstract

The invention discloses a method for non-uniformity correction of hyperspectral images. The invention first acquires imaging spectrometer data and performs data preprocessing, then performs distribution statistics on the data, uses data at different distribution positions to generate radiation correction coefficients, and finally completes high-resolution imaging. Non-uniformity correction of spectral images. Aiming at the problem of image non-uniformity correction caused by lack of non-uniformity correction coefficient or coefficient failure, the method basically does not cause loss of image information, and plays an important role in imager calibration and data preprocessing.

Description

A method for correcting non-uniformity of hyperspectral images

technical field

The invention belongs to the field of remote sensing detection and imaging spectrometer data processing, in particular to a method for correcting the non-uniformity of hyperspectral images.

Background technique

Due to the material, process and environmental changes of the detector itself, there are differences in the response value of the detector for uniform radiation or reflection signals, and the non-uniformity seriously affects the image quality and application effect of the imaging spectrometer. Therefore, non-uniformity is a necessary part of image processing.

The non-uniform correction methods include laboratory calibration method, image filtering and image-based statistical matching method. The laboratory calibration method can achieve high accuracy, but due to insufficient laboratory calibration, the detector is affected by the environment and other changes. The calibration coefficient is invalid, and it is difficult to correct the non-uniformity of the data; image filtering will cause the loss of image information, and the algorithm is complex and the operation speed is slow; the image-based statistical matching method is mainly based on moment matching. Different from the statistical type of the present invention, most of them are the mean value, standard deviation, correlation coefficient, etc. of the image, and this method has higher requirements on the scene distribution of the image. Therefore, there is a need for a non-uniformity correction method to realize the problem of non-uniformity correction caused by no calibration coefficient or failure of the calibration coefficient.

SUMMARY OF THE INVENTION

The problem to be solved by the present invention is to provide a non-uniformity correction method, which solves the problem of image non-uniformity correction caused by invalid calibration coefficients at present.

The present invention includes the following steps:

(1) Acquire the flight image data of a single imaging spectrometer, perform data preprocessing, subtract dark current from all image data, and obtain an image data set.

(2) Perform distribution statistics on the image data collected by each detector, the steps are as follows:

(2-1) Count the maximum value, minimum value and dynamic range of all image gray values of each detector;

(2-2) Count the cumulative histogram of each detector, and normalize the cumulative histogram to generate a normalized cumulative histogram.

(3) The radiation correction coefficient is generated, and the steps are as follows;

(3-1) Extract the gray value G corresponding to the specified normalized value in T groups (T≥2) from the normalized cumulative histogram, and obtain the mean value H of the gray value G in each row;

(3-2) Calculate the correction coefficients A and B for non-uniformity correction using the following equations

为T组灰度值G的均值，

为T组均值H的均值，T为步骤(3-1)提取的组数。in:

is the mean value of the gray value G of the T group,

is the mean value of the mean value H of T groups, and T is the number of groups extracted in step (3-1).

(4) Non-uniformity correction, all pixels are corrected by the non-uniformity correction coefficient, and the non-uniformity correction result is obtained.

Through the above method, the system of the present invention can realize the situation of lack of laboratory calibration and the failure of the calibration coefficient due to the change of the imager due to the influence of the environment, etc. Compared with the denoising method, the information of the image is better preserved.

Description of drawings

The present invention will be further described below in conjunction with the accompanying drawings and embodiments, in which:

Fig. 1 is the flow chart of the non-uniformity correction method;

FIG. 2 is a comparison diagram of the effect of the non-uniformity correction performed by the present invention, wherein the figure (a) is the original image, and the figure (b) is the non-uniformity correction result diagram.

Detailed ways

The non-uniformity correction method will be described in detail below with reference to FIGS. 1 to 2 .

(1) Acquire the flight data of a single imaging spectrometer. The size of the imaging spectrometer detector is 1024 × 256, the spatial dimension is 1024, and the spectral dimension is 256. The image size of a single flight route is about 100,000 frames, and data preprocessing is performed to convert all data Subtract the dark current to obtain the image data set, denoted as D _n (i, j), D _n (i, j) represents the image gray value of the i-th column, j-th row, and n-th frame of the detector, 0≤i＜ 1024, 0≤j<256, 0≤n<100000; part of the original image is shown in Figure 2(a);

(2) Carry out distribution statistics on the gray value of the image collected by each detector;

(2-1) Count the maximum value Max(i,j), the minimum value Min(i,j) and the dynamic range Rag(i,j) of all collected grayscale values of each detector, Rag(i,j)= Max(i,j)-Min(i,j)+1;

(2-2) Count the cumulative histogram of each detector and normalize it to obtain a normalized histogram H _r (i, j), where H _r (i, j) represents the i-th column of the detector, the Line j, the ratio of the number of pixels whose gray value is less than or equal to r to the total number of frames.

(3) Generation of radiation correction coefficients;

(3-1) Extract 9 groups of characteristic distribution values, and obtain the gray values corresponding to H _r (i, j) equal to 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, and 0.9, corresponding to G _t ( i,j),t=1,2,...,9; find the mean value H _t (j) of the gray value of each row, and the calculation formula is

(3-3) Calculate the correction coefficients A(i,j) and B(i,j) for the non-uniformity correction of each detector using the following equations

in:

(4) Non-uniformity correction, use the non-uniformity correction coefficient to correct all pixels, and obtain the non-uniformity correction result:

_Rn (i,j)=A(i,j)× _Dn (i,j)+B(i,j)

Among them, R _n (i, j) represents the image uniformity correction result of the i-th column, j-th row, and n-th frame of the detector. Part of the correction result is shown in Figure 2(b).

Claims (1)

1. a hyperspectral image non-uniformity correction method is characterized in that comprising the following steps: (1) Acquire the flight image data of a single imaging spectrometer, perform data preprocessing, subtract dark current from all image data, and obtain an image data set; (2) Perform distribution statistics on the image data collected by each detector, and the specific steps are as follows: (2-1) Count the maximum value, minimum value and dynamic range of all image gray values of each detector; (2-2) Count the cumulative histogram of each detector element, and normalize the cumulative histogram to generate a normalized cumulative histogram; (3) The radiation correction coefficient is generated, and the specific steps are as follows; (3-1) Extract T groups from the normalized cumulative histogram, T≥2, specify the gray value G corresponding to the normalized value, and obtain the mean value H of the gray value G in each row; (3-2) Calculate the correction coefficients A and B for non-uniformity correction using the following equations

in:

is the mean value of the gray value G of the T group,

is the mean value of the mean value H of T groups, and T is the number of groups extracted in step (3-1); (4) Non-uniformity correction, use the non-uniformity correction coefficient to correct all pixels, and obtain the non-uniformity correction result: _Rn (i,j)=A(i,j)× _Dn (i,j)+B(i,j) where R _n (i,j) represents the image uniformity correction result of the i-th column, j-th row, and n-th frame of the detector, and D _n (i, j) represents the i-th column, j-th row, and n-th frame of the detector The gray value of the image, A(i,j), B(i,j) represent the detector i-th column, j-th row correction coefficient.

Images