CN112258430B

CN112258430B - Universal correction method for remote sensing image radiation nonuniformity

Info

Publication number: CN112258430B
Application number: CN202011187801.5A
Authority: CN
Inventors: 范泽琳; 高放; 张鹏; 安源
Original assignee: Chang Guang Satellite Technology Co Ltd
Current assignee: Chang Guang Satellite Technology Co Ltd
Priority date: 2020-10-30
Filing date: 2020-10-30
Publication date: 2022-06-21
Anticipated expiration: 2040-10-30
Also published as: CN112258430A

Abstract

A universal correction method for remote sensing image radiation unevenness relates to the technical field of image processing, solves the problem of radiation unevenness caused by factors such as illumination conditions, atmospheric conditions, sensor equipment, imaging conditions and the like of the conventional remote sensing image, and comprises the following steps of firstly, carrying out multi-scale decomposition on an original image by adopting an NSCT (non-subsampled Contourlet transform) method; step two, carrying out NSCT on the obtained low-frequency subband image_LowpassPerforming Gaussian filtering to obtain an image background image, and removing the image background image; thirdly, performing linear restoration enhancement processing on the image radiation brightness value based on the local interested area radiation brightness information; replacing the low-pass sub-band image, and performing NSCT reconstruction to obtain a corrected band image; the invention can well inhibit the phenomenon of nonuniform radiation in the image by combining non-subsampled contourlet transformation, Gaussian filtering and a linear enhancement method based on local characteristics, and simultaneously, the change of radiation correction on the real radiation information of the ground object is greatly reduced.

Description

Universal correction method for remote sensing image radiation nonuniformity

Technical Field

The invention relates to the technical field of image processing, in particular to a universal correction method for a remote sensing image with non-uniform radiation.

Background

At present, most of correction methods for remote sensing image radiation unevenness perform corresponding compensation correction based on a distribution mode of radiation brightness, and have certain applicability. However, in practical situations, the radiation unevenness of the image has diversity, and there is a phenomenon that the radiation brightness changes irregularly and irregularly, and it is difficult to fit with an appropriate mathematical model.

In summary, the conventional radiation correction method for the radiation nonuniformity of the remote sensing image generally has certain limitations, and is difficult to adapt to the radiation nonuniformity condition which is diversified in reality.

Disclosure of Invention

The invention provides a universal correction method for remote sensing image radiation nonuniformity, aiming at solving the problem of radiation nonuniformity caused by factors such as illumination conditions, atmospheric conditions, sensor equipment, imaging conditions and the like of the conventional remote sensing image.

A universal correction method for remote sensing image radiation nonuniformity is realized by the following steps:

firstly, carrying out multi-scale decomposition on an original image by adopting an NSCT (non-subsampled Contourlet transform) method;

dividing an original image according to wave bands, and performing multi-scale decomposition on each wave band of the original image by adopting an NSCT (non-subsampled Contourlet transform) method; obtaining a low-frequency sub-band image NSCT of each wave band_LowpassAnd a plurality of directional band-pass sub-band images NSCT_BandpassSte；

Step two, performing NSCT on the low-frequency subband image obtained in the step one_LowpassPerforming Gaussian filtering to obtain an image background image, and removing the image background image; the method specifically comprises the following steps:

NSCT of the low-frequency subband image by adopting Gaussian filtering_LowpassProcessing to obtain non-uniform background image Info_BackgroundAnd the non-uniform background image information Info_BackgroundEliminating to obtain ground object radiation low-frequency image Info_FeatureLow；Info_FeatureLow＝NSCT_Lowpass-Info_Background；

Performing linear restoration enhancement processing on the image radiance value based on the local interesting area radiance information; the method specifically comprises the following steps:

the ground object radiation low-frequency image Info obtained in the step two_FeatureLowSelecting a local region of interest, and fitting low-frequency photons in the range of the local region of interest by a least square methodWith image NSCT_LowpassAnd ground object radiation low-frequency image Info_FeatureLowAnd the linear coefficient obtained by fitting is used for radiating the low-frequency image Info of the ground object_FeatureLowCorrecting to obtain corrected low-frequency image Info_{FeatureLowCor}；

Replacing the low-pass sub-band image, and performing NSCT reconstruction to obtain a corrected band image;

adopting the corrected low-frequency image Info obtained in the third step_{FeatureLowCor}Replacement of low frequency subband images NSCT_LowpassCombining the multi-direction band-pass sub-band image NSCT of the step one_BandpassSteNSCT reconstruction is carried out to obtain each wave band of the corrected image, and each wave band is combined to realize general correction of the original image.

The invention has the beneficial effects that: the general correction method can well inhibit the phenomenon of radiation nonuniformity in the image and simultaneously reduce the change of radiation correction on the real radiation information of the ground object to a greater extent by combining non-subsampled Contourlet (NSCT), Gaussian filtering and a local feature-based linear enhancement method. In addition, the method is not required to distinguish the type of the radiation unevenness of the image, and is applicable to images with various radiation unevenness conditions.

The universal correction method can effectively inhibit the phenomenon of uneven image radiation, simultaneously inhibits the change of radiation correction to the real radiation information of the ground object, has good universality to various types of uneven radiation, and provides a universal solution for the problem of uneven image radiation.

Drawings

FIG. 1 is a flow chart of a general correction method for remote sensing image radiation non-uniformity according to the present invention;

FIG. 2 is a graph of the luminance effect of an original remote sensing image;

fig. 3 is a graph of luminance effect after radiation homogenization correction.

Detailed Description

The invention is further illustrated with reference to the accompanying drawings and specific examples.

The main process of the invention comprises four parts of NSCT multi-scale decomposition, background information estimation and elimination by means of Gaussian filtering, ROI characteristic-based linear recovery enhancement and NSCT image reconstruction, and the detailed process is shown in figure 1.

Firstly, carrying out multi-scale decomposition on an original image wave band 1 by adopting NSCT;

NSCT is a two-dimensional image representation method combining non-downsampling pyramid decomposition and a non-downsampling direction filter bank, has the characteristics of multi-resolution, locality, multi-directionality, anisotropy, translation invariance and the like, and can effectively decompose an original two-dimensional image into a series of multi-scale band-pass sub-band images and a low-pass sub-band image. Here, NSCT is used to perform multi-scale decomposition on the original image band 1 to obtain 1 low-frequency subband image and several directional band-pass subband images, which are respectively denoted as NSCT_LowpassAnd NSCT_BandpassSte。

Secondly, based on the low-pass sub-band image obtained in the previous step, obtaining an image background image by means of Gaussian filtering, and removing the low-frequency sub-band image NSCT obtained in the previous step_LowpassThe method comprises the background information of the image and part of the low-frequency radiation information of the ground objects, wherein the image is processed by adopting Gaussian low-pass filtering, and then the non-uniform background image Info in the original image can be extracted_BackgroundRemoving the uneven background image from the low-frequency sub-band image, and leaving the ground object radiation low-frequency image Info_FeatureLowThe following formula:

Info_FeatureLow＝NSCT_Lowpass-Info_Background

and thirdly, based on a local Region of Interest (ROI) radiance image, after removing the background image in the previous step of linear restoration and enhancement of the radiance value of the image, the radiance information of the image is lost, so that linear restoration and enhancement processing of radiance needs to be performed on the low-frequency image of the ground feature obtained after processing. Because the radiation of the whole image is not uniform, a ROI with relatively uniform radiation and real color is selected in the image and is simulated by a least square methodNSCT in synthetic ROI_LowpassAnd Info_FeatureLowAnd the linear coefficient obtained by fitting is used for the whole Info_FeatureLowCorrecting to obtain Info_{FeatureLowCor}。

Fourthly, replacing the low-pass sub-band image, carrying out NSCT reconstruction to obtain a corrected band image, and converting the Info into a new entry_{FeatureLowCor}Replacement NSCT_LowpassCombined NSCT_BandpassSteNSCT reconstruction is performed to obtain an image band 1 in which the radiation non-uniformity phenomenon is suppressed by correction.

And (3) performing similar processing on other residual wave bands, repeating the steps 1-4 on the wave band 2 and the wave band 3 of the original image, and finally combining the three wave bands obtained by correction processing to finish the processing of the original remote sensing image.

The image used in the embodiment is a satellite image in a certain area of a scientific and technological development area in north lake of Changchun city, Jilin province, the coverage range is 125.377-125.432 degrees from east longitude, 43.972-43.992 degrees from north latitude, the image size is 3584 pixels 10240 pixels, and the image has an obvious radiation non-uniform phenomenon due to the influence of one or more reasons such as illumination conditions, atmospheric conditions, sensor equipment, imaging conditions and the like, and a high brightness band exists in the middle area of the image. Fig. 2 is a luminance graph of an original remote sensing image (after stretching with standard deviation of n equal to 1), wherein the upper left black thick frame selected region is the ROI region of the present embodiment, and fig. 3 is a luminance graph after radiation correction by this method (after stretching with standard deviation of n equal to 1). As can be seen from the observation, the present embodiment can effectively suppress the phenomenon of radiation unevenness in the image.

The foregoing is a more detailed description of the invention in connection with specific preferred embodiments and it is not intended that the invention be limited to these specific details. For those skilled in the art to which the invention pertains, several simple deductions or substitutions can be made without departing from the spirit of the invention, and all shall be considered as belonging to the protection scope of the invention.

Claims

1. A universal correction method for remote sensing image radiation nonuniformity is characterized in that: the method is realized by the following steps:

Step two, performing NSCT on the low-frequency subband image obtained in the step one_LowpassPerforming Gaussian filtering to obtain an image background image, and removing the image background image;

thirdly, based on the local interesting area radiance information, carrying out linear restoration enhancement processing on the image radiance value to obtain corrected low-frequency image Info_{FeatureLowCor}(ii) a The specific process is as follows:

the ground object radiation low-frequency image Info obtained in the step two_FeatureLowSelecting a local region of interest, and fitting a low-frequency sub-band image NSCT in the local region of interest by a least square method_LowpassAnd ground object radiation low-frequency image Info_FeatureLowAnd the linear coefficient obtained by fitting is used for radiating the low-frequency image Info of the ground object_FeatureLowCorrecting to obtain corrected low-frequency image Info_{FeatureLowCor}；

2. The universal correction method for remote sensing image radiation inhomogeneity according to claim 1, characterized in that: the specific process of the second step is as follows:

NSCT of the low-frequency subband image by adopting Gaussian filtering_LowpassProcessing to obtain non-uniform background image Info_BackgroundAnd the non-uniform background image information Info_BackgroundEliminating to obtain ground object radiation low-frequency image Info_FeatureLow；Info_FeatureLow＝NSCT_Lowpass-Info_Background。