CN105957030A - Infrared thermal imaging system image detail enhancing and noise inhibiting method - Google Patents

Abstract

The invention discloses an infrared thermal imaging system image detail enhancing and noise inhibiting method comprising the following steps: S1) extracting an edge image; S2) extracting a detail image; S3) extracting a basic image; S4) conducting detail enhancement to the extracted edge image, detail image and basic image respectively; and S5) integrating the enhanced edge image, the detail image and the basic image to obtain a final enhanced image. The invention provides an infrared thermal imaging system image detail enhancing and noise inhibiting method, which extracts and processes an edge image, a detail image and a base layer image separately, and integrates the processed images to obtain a final image whose contrast and detail are significantly enhanced. At the same time, obvious edges are removed from the detail layer corresponding to the second smooth image so as to completely rule out the phenomenon of edge gray-scale reversal.

Description

One is applied to thermal infrared imager image detail and strengthens and noise suppressing method

Technical field

The present invention relates to one and be applied to the enhancing of thermal infrared imager image detail and noise suppressing method.

Background technology

Modern high performance thermal infrared imager is obtained in that the raw image data that dynamic range is the biggest, bit wide general 12-16 position, but It is that the dynamic range bit wide of display device is generally 8, so must be to compression raw image data so on general display device Only can show big profile；Basic background information；Have lost the substantial amounts of minutia of scene, thus cause important detailed information to be lost Lose.

The image de-noising method of prior art has: color histogram equalization, utilizes the probability density function meter of entire image Calculating integral density distribution function, crossing grey-scale map realization process picture superposition the method can be to rectangular histogram probability density High moral gray level realizes obvious reinforced effects, and is difficult to low probability density gray level effectively strengthen, and even gray level merging is led Cause loss in detail.

Automatic growth control, first rejects the extreme value in scene, and then overall dynamic range is mapped to the gray space of 0-255, This method causes contrast low, and loss in detail is serious；

Image layered process, mainly image layered levels of detail and Primary layer, the method operand is big, edge ash easily occurs Degree reversion, strengthens inconspicuous to minor detail.

Summary of the invention

It is an object of the invention to overcome the deficiencies in the prior art, it is provided that one is applied to thermal infrared imager image detail and strengthens and make an uproar Sound suppressing method, extracting respectively and processing edge image, detail pictures and Primary layer image, then will process after figure As merging so that picture contrast and details after process are obviously enhanced.

It is an object of the invention to be achieved through the following technical solutions: one is applied to thermal infrared imager image detail and strengthens and make an uproar Sound suppressing method, comprises the following steps: include that edge image extraction step S1, detail pictures extraction step S2, primary image carry Take step S3, details strengthens step S4 and image co-registration step S5；

Described edge image extraction step S1 includes following sub-step:

S11. use the first edge smoothing filter to original image I process, obtain the first smoothed image T1；

S12. extraction edge image DI1:

DI1=I-T1；

It is primarily present in the first smoothed image T1 further, for most noise, utilizes original graph image I to deduct first Smoothed image T1 extracts edge image, it is possible to noise is played good inhibiting effect.

Described detail pictures extraction step S2 includes following sub-step:

S21. use the second edge smoothing filter to original image I process, obtain the second smoothed image T2；

S22. the second levels of detail DI2 corresponding for smoothed image T2 is calculated:

DI2=I-T2；

Further, owing to most noise is primarily present in the second smoothed image T2, deduct second with original image I and smooth Image T2 obtains levels of detail, it is possible to noise is played good inhibiting effect.

S23. extraction detail pictures DI:

DI=DI2-DI1；

Described primary image extraction step S3 includes: uses holding edge filter device to original image I process, obtains base This tomographic image T3；

Described details strengthens step S4 and includes:

S41. edge image DI1 is carried out details enhancement process, obtain the edge image DI1 ' strengthened；

S42. detail pictures DI is carried out details enhancement process, obtain the detail pictures DI ' strengthened；

S43. primary image P3 is carried out details enhancement process, obtain the primary image T3 ' strengthened；

Described image co-registration step S5 includes: the edge image DI1 ' of enhancing, detail pictures DI ' are entered with primary image T3 ' Row merges, and obtains final enhancing image DOUT_B:

DOUT_B=T3 '+P1*DI0 '+P2*DI '；

In formula, P1, P2 are proportionality coefficients, and the interval of P1 is (0,3), and the interval of P2 is (0,2).

Step S41 carries out details enhancing by the conversion of GAMMA curve to edge image DI1, obtains the edge image strengthened DI1′。

Step S42 carries out details enhancing by the conversion of GAMMA curve to detail pictures DI, obtains the detail pictures DI ' strengthened.

Described step S43 includes: by Primary layer image T3 to carrying out contrast enhancing, and pixel in Primary layer image T3 Intensity value ranges is compressed to the gray space of 0-255.

Described one is applied to thermal infrared imager image detail and strengthens and noise suppressing method, also includes a final enhancing figure As number field restriction step S6, including following sub-step:

S61. search the gray value pixel less than 0 in final enhancing image DOUT_B, and its gray value is defined to 0；

S62. search the gray value pixel more than 255 in final enhancing image DOUT_B, and its gray value is defined to 255.

The invention has the beneficial effects as follows: by extracting respectively and processing edge image, detail pictures and Primary layer image, Image after processing again merges so that picture contrast and details after process are obviously enhanced；Simultaneously at the second smooth figure In corresponding levels of detail, eliminate obvious edge, do not have edge gray inversion phenomenon completely.

Accompanying drawing explanation

Fig. 1 is the flow chart of the present invention.

Detailed description of the invention

Technical scheme is described in further detail below in conjunction with the accompanying drawings, but protection scope of the present invention is not limited to following institute State.

Strengthen and noise suppressing method as it is shown in figure 1, one is applied to thermal infrared imager image detail, comprise the following steps: bag Include edge image extraction step S1, detail pictures extraction step S2, primary image extraction step S3, details strengthen step S4 and Image co-registration step S5；

Described edge image extraction step S1 includes following sub-step:

S11. use the first edge smoothing filter to original image I process, obtain the first smoothed image T1；

S12. extraction edge image DI1:

DI1=I-T1；

Described detail pictures extraction step S2 includes following sub-step:

S21. use the second edge smoothing filter to original image I process, obtain the second smoothed image T2；

S22. the second levels of detail DI2 corresponding for smoothed image T2 is calculated:

DI2=I-T2；

S23. extraction detail pictures DI:

DI=DI2-DI1；

This step, in the second levels of detail DI2 corresponding for smoothed image T2, has deducted edge image DI1 and (has eliminated obvious edge Information), completely will not edge gray inversion phenomenon.

Described primary image extraction step S3 includes: uses holding edge filter device to original image I process, obtains base This tomographic image T3；

Described details strengthens step S4 and includes:

S41. edge image DI1 is carried out details enhancement process, obtain the edge image DI1 ' strengthened；

S42. detail pictures DI is carried out details enhancement process, obtain the detail pictures DI ' strengthened；

S43. primary image P3 is carried out details enhancement process, obtain the primary image T3 ' strengthened；

Described image co-registration step S5 includes: the edge image DI1 ' of enhancing, detail pictures DI ' are entered with primary image T3 ' Row merges, and obtains final enhancing image DOUT_B:

DOUT_B=T3 '+P1*DI0 '+P2*DI '；

In formula, P1, P2 are proportionality coefficients, and the interval of P1 is (0,3), and the interval of P2 is (0,2).

Step S41 carries out details enhancing by the conversion of GAMMA curve to edge image DI1, obtains the edge image strengthened DI1′。

Step S42 carries out details enhancing by the conversion of GAMMA curve to detail pictures DI, obtains the detail pictures DI ' strengthened.

Described step S43 includes: by Primary layer image T3 to carrying out contrast enhancing, and pixel in Primary layer image T3 Intensity value ranges is compressed to the gray space of 0-255.

Described one is applied to thermal infrared imager image detail and strengthens and noise suppressing method, also includes a final enhancing figure As number field restriction step S6, including following sub-step:

S61. search the gray value pixel less than 0 in final enhancing image DOUT_B, and its gray value is defined to 0；

S62. search the gray value pixel more than 255 in final enhancing image DOUT_B, and its gray value is defined to 255.

Embodiment 1, is detected entire image successively by the sliding window of (2k+1) * (2k+1):

$delta(x,y)=\underset{t=-k:k}{\overset{s=-k:k}{MAX}}\left(I\right(x+s,y+t\left)\right)-\underset{t=-k:k}{\overset{s=-k:k}{MIN}}\left(I\right(x+s,y+t\left)\right);---\left(1\right)$

In formula, x represents image line coordinate, and y represents image column coordinate, k window size；

-k:k represents that the value of s and t is the integer between-k to k, i.e. s=-k ,-k+1 ..., k-1, k；

T=-k ,-k+1 ..., k-1, k.

Represent, be s=-k ,-k+1 in s span ..., k-1, k, t span is

T=-k ,-k+1 ..., when k-1, k, the maximum of I (x+s, y+t), i.e. pixel maximum in current window.

In like manner,Represent the minima of I (x+s, y+t), i.e. pixel minimum in current window.

Delta in formula (1) (x, y) represents that (2k+1) * (2k+1) window that slides is sought local delta value by image, Big value and descend most the difference of value, it is possible to details and edge in phenogram picture, generally:

The region of delta < 100 is not comprise details smooth region；The region of 100 < delta < 2000 is mainly details；Delta> The region of 2000 is obvious edge.

Introduce parameter eps, to delta (x, y) is normalized, and obtains:

$a(x,y)=\frac{delta(x,y)}{delta(x,y)+eps},---\left(2\right)$

According to formula (2) it can be seen that

If delta (x, y) >=eps, then a (x, y) >=0.5；

If delta (x, y) ＜ eps, then a (x, y) ＜ 0.5；

Ask image to slide (2k+1) * (2k+1) window ask local average:

$I\_avg(x,y)=\frac{\underset{s=-k}{\overset{s=k}{\&Sigma;}}\underset{t=-k}{\overset{t=k}{\&Sigma;}}I(x+s,y+t)}{{(2k+1)}^2}---\left(3\right)$

For edge smoothing filter:

I_s (x, y)=I (x, y) * (1-a (x, y))+I_avg (x, y) * a (x, y) (4)

In formula, (x, y) represents edge smoothing filter output image to I_s, and (x y) represents original image to I.

(x, y) the biggest then average ratio is the biggest, and image original value ratio is the least, say, that the region of delta < eps for visible a Delta is the least closer to artwork；Delta=eps region artwork and average respectively account for half；The region delta of delta > eps is the biggest The most smooth；Therefore the first edge smoothing filter and the second edge smoothing filter just can realize by arranging different eps, It is between 2000～5000 that such as first edge smoothing filter can arrange eps, and the second edge smoothing filter can arrange eps and be Between 30～100.

For holding edge filter device:

I_bs (x, y)=I_avg (x, y) * (1-a (x, y))+I (and x, y) * a (x, y), (5)

I_bs (x, y) be holding edge filter device output image.

From formula (5): the smooth smaller region of delta, keep region bigger for delta, i.e. can reach edge Purpose, such as eps is kept to may be configured as 30～100.

Claims (5)

1. one kind is applied to the enhancing of thermal infrared imager image detail and noise suppressing method, it is characterised in that: comprise the following steps: include Edge image extraction step S1, detail pictures extraction step S2, primary image extraction step S3, details strengthen step S4 and figure As fusion steps S5；

Described edge image extraction step S1 includes following sub-step:

S11. use the first edge smoothing filter to original image I process, obtain the first smoothed image T1；

S12. extraction edge image DI1:

DI1=I-T1；

Described detail pictures extraction step S2 includes following sub-step:

S21. use the second edge smoothing filter to original image I process, obtain the second smoothed image T2；

S22. the second levels of detail DI2 corresponding for smoothed image T2 is calculated:

DI2=I-T2；

S23. extraction detail pictures DI:

DI=DI2-DI1；

Described primary image extraction step S3 includes: uses holding edge filter device to original image I process, obtains Primary layer Image T3；

Described details strengthens step S4 and includes:

S41. edge image DI1 is carried out details enhancement process, obtain the edge image DI1 ' strengthened；

S42. detail pictures DI is carried out details enhancement process, obtain the detail pictures DI ' strengthened；

S43. primary image P3 is carried out details enhancement process, obtain the primary image T3 ' strengthened；

Described image co-registration step S5 includes: the edge image DI1 ' of enhancing, detail pictures DI ' are melted with primary image T3 ' Close, obtain final enhancing image DOUT_B:

DOUT_B=T3 '+P1*DI0 '+P2*DI '；

In formula, P1, P2 are proportionality coefficients, and the interval of P1 is (0,3), and the interval of P2 is (0,2).

One the most according to claim 1 is applied to thermal infrared imager image detail and strengthens and noise suppressing method, it is characterised in that: Step S41 carries out details enhancing by the conversion of GAMMA curve to edge image DI1, obtains the edge image DI1 ' strengthened.

One the most according to claim 1 is applied to thermal infrared imager image detail and strengthens and noise suppressing method, it is characterised in that: Step S42 carries out details enhancing by the conversion of GAMMA curve to detail pictures DI, obtains the detail pictures DI ' strengthened.

One the most according to claim 1 is applied to thermal infrared imager image detail and strengthens and noise suppressing method, it is characterised in that: Described step S43 includes: by Primary layer image T3 to carrying out contrast enhancing, and pixel gray scale in Primary layer image T3 Value Ratage Coutpressioit is to the gray space of 0-255.

One the most according to claim 1 is applied to thermal infrared imager image detail and strengthens and noise suppressing method, it is characterised in that: Also include final enhancing image number field restriction step S6, including following sub-step:

S61. search the gray value pixel less than 0 in final enhancing image DOUT_B, and its gray value is defined to 0；

S62. search the gray value pixel more than 255 in final enhancing image DOUT_B, and its gray value is defined to 255.