A multispectral remote sensing image restoration method

Abstract

A multispectral remote sensing image restoration method relates to the technical field of remote sensing image processing. In order to solve the problems that the existing technology causes blurring of images when a strip filter is abnormal, and the existing restoration method causes different degrees of loss of original information of the remote sensing multispectral image, the energy attenuation coefficient of light after multiple reflections is set to d, the offset of the position where the energy is superimposed after reflections relative to the original normal light imaging position is a, the clear image of normal imaging is J, and the blurred image obtained after superposition after reflections is Z; when the pixel position is set to k, the code value of the clear image is J[k], and the code value of the blurred image after the energy of the position is superimposed is Z[k], then the code value of the clear image at the position k+a is J[k+a], and the code value of the corresponding blurred image is Z[k+a], and so on, to obtain a restored image of the clear image. The method of the invention avoids the complex calculation of other general restoration algorithms and the loss of image detail information caused by restoration.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image processing, and in particular to a multi-spectral remote sensing image restoration method. Background Art

[0002] In recent years, with the rapid development of remote sensing technologies such as aviation and aerospace, remote sensing imaging technology has become increasingly mature. As an important imaging technology, multispectral imaging technology is widely used in agriculture, military defense, water resource control, and mineralogy mapping of the earth's surface. Compared with traditional color or full-color images, multispectral images can provide a more realistic description of real scenes and effectively enhance the efficiency of computer vision tasks.

[0003] However, the remote sensing imaging link is very complex, including information transmission, collection, conversion and processing, which are inevitably affected by various factors, such as atmospheric absorption and scattering, non-uniform response of sensor devices, filter coating non-uniformity, satellite platform vibration, etc., which will lead to a decrease in image quality. Although researchers have been committed to improving image quality by optimizing system design, the impact of degradation factors cannot be completely resolved. With the development of applications such as fine target classification, micro-target detection, and ground component analysis, the requirements for multispectral quality are also constantly increasing.

[0004] The use of image processing technology to restore the obtained multispectral remote sensing images has attracted extensive attention from scholars and researchers at home and abroad, such as the proposed image restoration method based on regularization, the image restoration method based on low-rank matrix decomposition, the image restoration method based on filtering, etc. However, the above methods do not consider the prior information of the imaging system itself, and perform image restoration processing, resulting in varying degrees of loss of original image detail information while restoring the image. Summary of the invention

[0005] The present invention provides a multispectral remote sensing image restoration method to solve the problems that when a strip filter is abnormal in the prior art, the image becomes blurred, and the existing restoration method causes different degrees of loss of original information of the remote sensing multispectral image.

[0006] A multispectral remote sensing image restoration method is implemented by the following steps:

[0007] Step 1: Set the energy attenuation coefficient of the light after multiple reflections to d, d<1; the offset of the position where the energy is superimposed after reflection relative to the original normal light imaging position is a, the clear image of the normal imaging is J, and the blurred image obtained after superposition after reflection is Z;

[0008] Step 2: When the pixel position is set to k, the code value of the clear image is J[k], and the code value of the blurred image after the reflected energy at this position is Z[k]. Then the code value of the clear image at position k+a is J[k+a], and the code value of the corresponding blurred image is Z[k+a]. Similarly, the following formula is obtained:

[0009] Z[k]=J[k]+J[k+a]*d

[0010] Z[k+a]=J[k+a]+J[k+a*2]*d

[0011] Z[k+a*2]=J[k+a*2]+J[k+a*3]*d

[0012] …

[0013] Z[k+a*n]=J[k+a*n]+J[k+a*(n+1)]*d

[0014] In the formula, n represents n times the offset, which can be derived from the above formula:

[0015]

[0016] Thus, the code value J[k] of the clear image at position k is obtained as follows:

[0017]

[0018] Where N is a positive integer, indicating that the final translation is N times a. When N takes a certain value, d<1, then (-1) N+1 *J[k+a*(N+1)]*d N+1 is a high-order infinitesimal quantity, that is:

[0019] (-1) N+1 *J[k+a*(N+1)]*d N+1 →0

[0020] Finally, the code value J[k] of the clear image at position k is:

[0021]

[0022] Beneficial effects of the present invention:

[0023] The image restoration method of the present invention uses the principle of optical analysis, that is, after multiple reflections of light, the offset of the energy attenuation superposition position relative to the original normal light imaging position is constant. Starting from the prior information of the imaging system itself, a simple and effective model for image restoration is obtained, avoiding the complex calculation of other general restoration algorithms and the loss of image detail information after restoration. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 An imaging optical path diagram of an optical system in a multi-spectral remote sensing image restoration method according to the present invention;

[0025] Figure 2 The figures are the effect diagrams before and after image restoration using the method of the present invention; (a) is the original blurred image, and (b) is the restored image. DETAILED DESCRIPTION

[0026] Combination Figure 1 and Figure 2 The present embodiment describes a multi-spectral remote sensing image restoration method.

[0027] First, the mechanism of image blur is analyzed in principle to provide prior knowledge for image restoration. The imaging optical path of the optical system is as follows: Figure 1 As shown. Under normal imaging conditions, the imaging beam reaches the filter, is split by the strip filter, and the energy within the effective bandwidth passes through the strip filter to reach the CMOS detector. Under abnormal imaging conditions, that is, when there is an abnormality in the strip filter, part of the light passing through the front surface of the filter may be reflected back to the front surface of the filter at the rear surface of the filter, and then pass through the rear surface of the filter to enter the CMOS photosensitive surface after being reflected again on the front surface. In this way, the normal imaging beam and the abnormal imaging beam are superimposed at the same position on the CMOS photosensitive surface, resulting in a blurred image.

[0028] After analysis, it is found that the blur of the image is caused by the energy of the abnormal imaging being superimposed on the normal image after multiple reflection attenuation. This provides an idea for image restoration, that is, we need to separate the energy of the abnormal imaging from the blurred image to obtain a clear image after restoration.

[0029] We assume that the energy attenuation coefficient after multiple reflections is d (d<1). Since the multispectral remote sensing linear array camera is a push-scan imaging, it is assumed that the offset of the position of energy superposition after reflection relative to the original normal light imaging position is a, the clear image of normal imaging is J, and the blurred image obtained after superposition after reflection is Z. We assume that when the pixel position is k, the code value of the clear image is J[k], and the code value of the blurred image after the energy superposition after reflection at this position is Z[k]. Then the code value of the clear image at position k+a is J[k+a], and the code value of the corresponding blurred image is Z[k+a]. By analogy, the following expression can be obtained:

[0030] Z[k]=J[k]+J[k+a]*d

[0031] Z[k+a]=J[k+a]+J[k+a*2]*d

[0032] Z[k+a*2]=J[k+a*2]+J[k+a*3]*d

[0033] …

[0034] Z[k+a*n]=J[k+a*n]+J[k+a*(n+1)]*d

[0035] From the above, it can be deduced that:

[0036]

[0037] It can be obtained that the code value of the clear image at position k is:

[0038]

[0039] Where N is a positive integer, indicating that the final translation amount is N times a. When N reaches a certain value, because d<1, (-1) N+1 *J[k+a*(N+1)]*d N+1 is a high-order infinitesimal quantity, that is:

[0040] (-1) N+1 *J[k+a*(N+1)]*d N+1 →0

[0041] Therefore, the restored image J[k] is:

[0042]

[0043] In actual engineering applications, it is found that the energy attenuation coefficient d is often not a constant, but may be a linear or nonlinear function d(k) of the change in pixel position k. At the same time, the position offset may also be a piecewise function a(k) of the change in pixel position k. This does not affect the overall restoration method above, so a more general restoration method is obtained as follows:

[0044]

[0045] After the above restoration method, the restored image can be obtained as follows Figure 2 As shown, it can be seen that the algorithm of the present invention effectively eliminates the blurring phenomenon of the image and effectively retains the detail information of the original image.

[0046] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0047] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A multispectral remote sensing image restoration method, characterized by: The method is implemented by the following steps: Step 1: Set the energy attenuation coefficient of the light after multiple reflections to d, d < 1; the offset of the position where the energy is superimposed after reflection relative to the original normal light imaging position is a, the clear image of the normal imaging is J, and the blurred image obtained after superposition after reflection is Z; Step 2: When the pixel position is set to k, the code value of the clear image is J[k], and the code value of the blurred image after the reflected energy at this position is Z[k]. Then the code value of the clear image at position k+a is J[k+a], and the code value of the corresponding blurred image is Z[k+a]. Similarly, the following formula is obtained: Z[k]=J[k]+J[k+a]*d Z[k+a]=J[k+a]+J[k+a*2]*d Z[k+a*2]=J[k+a*2]+J[k+a*3]*d …… Z[k+a*n]=J[k+a*n]+J[k+a*(n+1)]*d In the formula, n is n times the offset; From the above formula, we can deduce: Thus, the code value J[k] of the clear image at position k is obtained as follows: Where N is a positive integer, indicating that the final translation is N times a. When N takes a certain value, d < 1, then, (-1) N+1 *J[k+a*(N+1)]*d N+1 is a high-order infinitesimal quantity, that is: (-1) N+1 *J[k+a*(N+1)]*d N+1 →0 Finally, the code value J[k] of the clear image at position k is:

2. A multispectral remote sensing image restoration method according to claim 1, characterized in that: Assuming the energy attenuation coefficient d is a linear or nonlinear function d(k) about the pixel position k, and the position offset is a piecewise function a(k) about the pixel position k, the final restoration method formula is as follows:

Images