De-noising system for remote images of ground buildings using spectrum constraints and de-noising method thereof

Patent metadata

Field	Value
Publication number	US-9563939-B2
Application number	US-201514964559-A
Country	US
Kind code	B2
Filing date	Dec 9, 2015
Priority date	Dec 30, 2014
Publication date	Feb 7, 2017
Grant date	Feb 7, 2017

Abstract

A de-noising method for remote images of ground buildings using spectrum constraints. The method includes: 1) obtaining a reference image of ground buildings from a remote image database of the ground buildings, performing a Fourier transformation on the reference image to obtain an amplitude spectrum, and performing a threshold segmentation, an erosion operation and a dilation operation successively on the amplitude spectrum to obtain a binary template of spectrum of the ground buildings; and 2) obtaining a real-time image of the ground buildings by a high-speed aircraft, performing a Fourier transformation on the real-time image to obtain a spectrum, filtering the spectrum of the real-time image in frequency domain by the binary template of spectrum of the ground buildings, and performing an inverse Fourier transformation thereon to generate a filtered real-time image of the ground buildings.

First claim

The invention claimed is:

1. A de-noising method for remote images of ground buildings using spectrum constraints, the method comprising:
1) obtaining a reference image of target ground buildings from a remote image database of ground buildings, performing a Fourier transformation on the reference image to obtain an amplitude spectrum, and performing a threshold segmentation, an erosion operation and a dilation operation successively on the amplitude spectrum to obtain a binary template of spectrum of the target ground buildings; and
2) obtaining a real-time image of the target ground buildings by a high-speed aircraft, performing a Fourier transformation on the real-time image to obtain a spectrum, filtering the spectrum of the real-time image in frequency domain by the binary template of spectrum of the target ground buildings to obtain a filtered spectrum, and performing an inverse Fourier transformation on the filtered spectrum to generate a denoised real-time image of the target ground buildings.

2. The method of claim 1 , wherein step 1) further comprises:
1.1) obtaining a reference image p of target ground buildings from a remote image database of ground buildings, and performing a two-dimensional fast Fourier transformation and a centralizing operation successively on the reference image p to obtain a centralized spectrum P thereof;
1.2) calculating an amplitude spectrum P according to the centralized spectrum P of the reference image p where P is a modulus of the centralized spectrum P and P =|P|;
1.3) generating a histogram distribution Hist p (x) of the amplitude spectrum P , and normalizing the histogram distribution Hist p (x) to obtain a normalized histogram distribution

Hist
p
′

⁡

(
x
)

=

Hist
p

⁡

(
x
)

∑
x

⁢

Hist
p

⁡

(
x
)

,

where an abscissa x is the amplitude of the amplitude spectrum P ;
1.4) calculating a segmentation threshold T according to the normalized histogram distribution Hist p ′(x), where

∑

i
=
0

T

⁢

⁢

Hist
p
′

⁡

(
x
)

=

1
-
γ

and γ is a reserved quantity of a target spectrum template;
1.5) performing a threshold segmentation on the amplitude spectrum P according to the segmentation threshold T to obtain a binary segmentation result BW 0 ; and
1.6) performing an erosion operation and a dilation operation successively on the binary segmentation result BW 0 to obtain a binary template BW of spectrum of the target ground buildings.

3. The method of claim 2 , wherein step 1.2) further comprises:
calculating each point of the amplitude spectrum P by an equation P(u,v) =|P(u,v)|=√{square root over ((a 2 +b 2 ))} to obtain the amplitude spectrum P , where P(u,v) is a point of the centralized spectrum P with a complex form of P(u,v)=a+bi, (u,v) is a coordinate of P(u,v), 1≦u≦256, 1≦v≦256, and a and b are constants.

4. The method of claim 3 , wherein step 1.5) further comprises:
determining BW 0 (u,v)=1 if P(u,v) ≧T, otherwise BW 0 (u,v)=0, where P(u,v) is a point of the amplitude spectrum P and BW 0 (u,v) is a corresponding point of the binary segmentation result BW 0 .

5. The method of claim 4 , wherein step 2) further comprises:
2.1) obtaining a real-time image f of the target ground buildings by a high-speed aircraft, and performing a two-dimensional fast Fourier transformation and a centralizing operation successively on the real-time image f to obtain a centralized spectrum F of the real-time image f;
2.2) generating a filter function H by the binary template BW of spectrum of the target ground buildings generated in step 1.6);
2.3) performing a dot product operation on each element of the centralized spectrum F and a corresponding element of the filter function H to obtain a filtered spectrum G of the real-time image f thereby filtering the real-time image f in frequency domain where G=F*H; and
2.4) performing a two-dimensional inverse fast Fourier transformation on the filtered spectrum G of the real-time image f and performing a modulus operation on a result thereof to obtain a denoised real-time image g.

6. The method of claim 5 , wherein step 2.2) further comprises:
generating a filter function H:

H
⁡

(

u
′

,

v
′

)

=

{

1
,

BW
⁡

(

u
,
v

)

=
1

λ
⁢

⁢
%

,

Assignees

• Univ Huazhong Science Tech

Inventors

• Zhang Tianxu
• Wang Zheng
• He Li
• Liu Li
• Hou Xuan
• Zhang Chuan
• Fan Sufei
• Chen Yimeng

Classifications

• G06T2207/30184

  Infrastructure · CPC title

• G06T5/10

  using non-spatial domain filtering · CPC title

• G06T2207/20148

  Physics · mapped topic

• G06T7/0091

  Physics · mapped topic

• G06T2207/20056

  Discrete and fast Fourier transform, [DFT, FFT] · CPC title