CN103996178A - Sand and dust weather color image enhancing method - Google Patents

Abstract

The invention relates to a sand and dust weather color image enhancing method which comprises the steps that color space is converted, and three components of a converted CIE LAB space are represented through L*, a* and b*; L* channel denoising is carried out; L* channel contrast ratio stretching is carried out; L* channel sharpening processing is carried out; a gray world assumption is used for carrying out color correcting processing on an a* channel and a b* channel; and color images after processing are reestablished. Color cast and noise influence existing in the images can be removed, and the image contrast ratio is improved.

Description

A kind of dust and sand weather colour-image reinforcing method

Affiliated technical field

The present invention relates to the image enhancement technique under severe weather conditions, especially for the enhancing of dust and sand weather coloured image.

Background technology

Dust and sand weather is a kind of inclement weather that often occurs NW China and North China in spring.According to incompletely statistics, only, during 2000～2003 years, the dust and sand weather of Kashi Area of Xingjiang Autonomous Region is on average about 103 days/year [1].According to the height difference of horizontal visibility, dust and sand weather grows from weak to strong and is divided into successively floating dust (visibility is greater than 10 kilometers), sand (visibility is less than 10 kilometers), sandstorm (visibility is less than 1 kilometer) and strong sandstorm (visibility is less than 50 meters) four classes.Wherein more common with sand and sandstorm, when dust and sand weather occurs, can make visibility significantly reduce, all trades and professions are all had to harmful effect in various degree, particularly serious on the impact of communications and transportation.In addition, image and video visual quality that under dust and sand weather condition, monitoring camera is taken significantly decline, and outdoor monitoring system cannot the work of normal reliable ground.So, the image obtaining under dust and sand weather condition is strengthened to processing, improve contrast and the visual quality of image, there is important using value and social reality meaning.

Dust and sand weather figure image intensifying belongs to inclement weather figure image intensifying field, but current research mainly concentrates on Misty Image enhancing aspect, and existing method can be divided into two classes: the method based on model and the method based on image processing.The middle employing of document [2] is set up Atmospheric models and is restored the method for degraded image under greasy weather condition, but needs multiple parameters; Underwater picture and sand and dust image have similarity aspect a lot, and such as image inside all exists obvious colour cast and noise effect, the contrast of image is all lower.Document [3] has proposed a kind of image co-registration scheme for strengthening underwater picture.First input picture is carried out to color correction, and use histogram equalization technology to carry out denoising and contrast stretching to color correction result, and then extract several features for determining image co-registration matrix of coefficients used.

List of references:

[1] Jiang Yuanan, Zhang Yunhui, etc. Analysis of Sand-dust Weather in Kashi Area [J], Xinjiang meteorology, 2005, (28).

[2] Wang Zhijian, the image restoration based on Atmospheric models improves algorithm and application [J], computer engineering and application, 2007,43 (3): 239-248.

[3]C.Ancuti,C.O.Ancuti,T.Haber,P.Bekaert,Automatic red-eye detection and removal[C],in Proceeding of CVPR,81-88,2012.

Summary of the invention

The object of the invention is to propose a kind of Enhancement Method for dust and sand weather coloured image, the colour cast existing in removal of images and noise effect, the contrast of raising image.Technical scheme of the present invention is as follows:

A kind of dust and sand weather colour-image reinforcing method, comprises the following steps:

1. color space conversion: a width dust and sand weather coloured image that represents input with I, I is transformed into CIE LAB color space by rgb space, I represents with L*, a* and b* respectively at the three-component in CIELAB space, L*, a* and b* triple channel are normalized, triple channel value is all adjusted to interval [0,1], use I _l, I _aand I _brepresent respectively normalized each channel image afterwards;

2. L* passage denoising: use Gauss's low pass template to I _lcarry out filtering processing, remove I _lin noise, result I _l1represent;

3. L* passage contrast stretching: adopt Sigmoid function to I _l1carry out gray level stretch processing, compress low gray level and high grade grey level part, to improve contrast, contrast stretching result I _l2represent;

4. L* passage sharpening processing: adopt unsharp template to I _l2carry out sharpening processing, Gauss's low pass template is processed I _l2, obtain low-pass pictures, with representing adopt following formula to calculate sharpening image, use I _l3represent:

5. adopt gray scale world hypothesis to carry out color correction process to the a* passage after normalization and b* passage, step is as follows:

1) calculate I _aand I _bmean value, use respectively μ _aand μ _brepresent;

2) the adjustment coefficient of calculating a* passage and b* passage, uses respectively λ _aand λ _brepresent Qi Zhongyou

λ _A＝0.5/μ _A，λ _A＝0.5/μ _B；

3) use I _a1and I _b1expression color correction after image, they meet I _a1=I _a× λ _a, I _b1=I _b× λ _b;

6. the coloured image after reconstruction processing.

As preferred implementation, in step 3, the expression formula of S function used is: in formula, x represents input gray grade, and y represents output gray level, and their span is all between [0,1]; In step 6, the step of employing is:

1) by I after treatment _l3in the value of each pixel adjust between [0,100], result I ' _lrepresent;

2) by I _a1and I _b1in the value of each pixel adjust to respectively between [127,128], result I ' _aand I ' _brepresent;

3) use the inverse transformation formula of CIE LAB color space to RGB color space, the coloured image after reconstruction processing.

The Enhancement Method for dust and sand weather coloured image the present invention proposes, is first luminance channel and chrominance channel by separation of images, and luminance channel image is carried out to denoising, contrast stretching and sharpening processing, and chrominance channel is carried out to color correction.Then two passages after treatment are reassembled into coloured image after treatment.Computer artificial result shows, the method proposing can significantly improve the visual quality of dust and sand weather image, and the computation complexity of institute's extracting method is not high in addition, can meet the requirement of real-time processing.

Brief description of the drawings

Fig. 1 is the process flow diagram of institute's extracting method.

Fig. 2 is Gauss's low pass template (variance is 0.5) of image denoising process 5 × 5 sizes used.

Fig. 3 is contrast stretching process S function schematic diagram used.

Fig. 4 sand and dust image enhancement processing process example, (a) input dust and sand weather image, (b) L channel image, (c) strengthens L channel image after treatment, (d) image after treatment.

Fig. 5 part of test results, a row figure in left side is that row on former figure right side are schemed corresponding figure's) result figure.

Embodiment

Institute of the present invention extracting method comprises three parts: color space conversion, luminance channel processing, chrominance channel processing.Fig. 1 has provided the block diagram of institute's extracting method.

1, color space conversion

Represent a width dust and sand weather coloured image of input with I.I is transformed into CIELAB color space by rgb space, and I represents with L*, a* and b* respectively at the three-component in CIE LAB space, and wherein the value of L* passage is between [0,100], and the value of a* component and b* component is all between [127,128].

Process conveniently for later step, L*, a* and b* triple channel are normalized, all adjust to interval [0,1] by triple channel value.Use I _l, I _aand I _brepresent respectively normalized each channel image afterwards.

2, L* passage denoising

Use Gauss's low pass template of 5 × 5 sizes to I _lcarry out filtering processing, remove I _lin noise.As shown in Figure 2, wherein Gaussian function variance is in the horizontal and vertical directions all 0.5 to template used coefficient.Result I _l1represent.

3, L* passage contrast stretching

Adopt S type function (Sigmoid function) to I _l1carry out gray level stretch processing, compress low gray level and high grade grey level part, reach the object that increases contrast.The expression formula of S function used is:

$y=\frac{1}{{1+e}^{-10(x-0.5)}}---\left(1\right)$

In formula, x represents input gray grade, and y represents output gray level, and their span is all between [0,1].Fig. 3 has provided the schematic diagram of S function used.Contrast stretching result I _l2represent.

4, L* passage sharpening processing

Adopt unsharp template to I _l2carry out sharpening processing.Concrete way is to use 5 × 5 Gauss's low pass templates shown in Fig. 2 to process I _l2, obtain low-pass pictures, with representing adopt following formula to calculate sharpening image, use I _l3represent.

$I_{L3}=2\·I_{L2}-{\tilde{I}}_{L2}---\left(2\right)$

5, the color correction of a* passage and b* passage

Adopt gray scale world hypothesis (Gray World) to carry out color correction process to the a* passage after normalization and b* passage.Specific practice is as follows:

4) calculate I _aand I _bmean value, use μ _aand μ _brepresent.

5) the adjustment coefficient of calculating a* passage and b* passage, uses respectively λ _aand λ _brepresent Qi Zhongyou

λ _A＝0.5/μ _A，λ _A＝0.5/μ _B (3)

6) use I _a1and I _b1expression color correction after image, they meet

I _A1＝I _A×λ _A，I _B1＝I _B×λ _B (4)

6, Image Reconstruction

Use following formula by I after treatment _l3in the value of each pixel adjust between [0,100], result I ' _lrepresent

I’ _L＝I _L3×100 (5)

Use following formula by I _a1and I _b1in the value of each pixel adjust between [127,128], result I ' _aand I ' _brepresent

I’ _A＝I _A1×255-127 (6)

I’ _B＝I _B1×255-127 (7)

Use the inverse transformation formula of CIELAB color space to RGB color space, the coloured image after reconstruction processing.

Matlab2014a under employing Windows7 SP1 system is as experiment simulation platform.Select patent applicant from Internet download 78 width sand and dust images as test set.The method that adopts this patent to propose is processed test pattern, has obtained good treatment effect.For the image of 1024768 sizes, adopt the processing speed average out to 18ms of institute's extracting method, processing speed is very fast.Fig. 4 has provided employing institute extracting method processing procedure example.Fig. 5 has provided more result, and wherein left side is input picture, and right side is result.

Adopt the method for the invention, compared with prior art, corrected the colour cast situation that original image exists, obviously improved the contrast of image, improved the visual quality of image, saved the time of computing machine processing, reached the effect of more approaching practicality.

Claims (3)

1. a dust and sand weather colour-image reinforcing method, comprises the following steps:

1. color space conversion: a width dust and sand weather coloured image that represents input with I, I is transformed into CIELAB color space by rgb space, I represents with L*, a* and b* respectively at the three-component in CIELAB space, L*, a* and b* triple channel are normalized, triple channel value is all adjusted to interval [0,1], use I _l, I _aand I _brepresent respectively normalized each channel image afterwards;

2. L* passage denoising: use Gauss's low pass template to I _lcarry out filtering processing, remove I _lin noise, result I _l1represent;

3. L* passage contrast stretching: adopt Sigmoid function to I _l1carry out gray level stretch processing, compress low gray level and high grade grey level part, to improve contrast, contrast stretching result I _l2represent;

4. L* passage sharpening processing: adopt unsharp template to I _l2carry out sharpening processing, Gauss's low pass template is processed I _l2, obtain low-pass pictures, with representing adopt following formula to calculate sharpening image, use I _l3represent:

5. adopt gray scale world hypothesis to carry out color correction process to the a* passage after normalization and b* passage, step is as follows:

1) calculate I _aand I _bmean value, use respectively μ _aand μ _brepresent;

2) the adjustment coefficient of calculating a* passage and b* passage, uses respectively λ _aand λ _brepresent Qi Zhongyou

λ _A＝0.5/μ _A，λ _A＝0.5/μ _B；

3) use I _a1and I _b1expression color correction after image, they meet I _a1=I _a× λ _a, I _b1=I _b× λ _b;

6. the coloured image after reconstruction processing.

2. dust and sand weather colour-image reinforcing method according to claim 1, is characterized in that, in step 3, the expression formula of S function used is: in formula, x represents input gray grade, and y represents output gray level, and their span is all between [0,1].

3. dust and sand weather colour-image reinforcing method according to claim 1, is characterized in that, in step 6, the step of employing is:

1) by I after treatment _l3in the value of each pixel adjust between [0,100], result I ' _lrepresent;

2) by I _a1and I _b1in the value of each pixel adjust to respectively between [127,128], result I ' _aand I ' _brepresent;

3) use the inverse transformation formula of CIE LAB color space to RGB color space, the coloured image after reconstruction processing.