CN104408695B - Histogram FUZZY WEIGHTED is adjusted and infrared image enhancing method in a balanced way - Google Patents

Abstract

The invention discloses a kind of adjustment of histogram FUZZY WEIGHTED and infrared image enhancing method in a balanced way；Belong to infrared image enhancing method technical field；Technical points comprise the steps：(1) original infrared image histogram p is calculated_r(r_k)；(2) the objective fuzzy degree of membership μ (r of gray level are calculated_k)；(3) according to objective fuzzy degree of membership μ (r_k) to histogram p_r(r_k) adjustment is weighted, obtain new histogram p '_r(r_k)；(4) to new histogram p '_r(r_k) equalization processing is carried out, obtain enhancing image I'(x)；The present invention is intended to provide one kind can improve infrared image enhancement effect, and the small histogram FUZZY WEIGHTED adjustment of amount of calculation and infrared image enhancing method in a balanced way；For infrared image enhancement.

Description

Histogram FUZZY WEIGHTED is adjusted and infrared image enhancing method in a balanced way

Technical field

The present invention relates to a kind of image enchancing method, more specifically, more particularly to a kind of histogram FUZZY WEIGHTED is adjusted With infrared image enhancing method in a balanced way.

Background technology

Infrared image generally there are low signal to noise ratio, object edge and details it is fuzzy the shortcomings of, for the ease of eye-observation with And be conducive to follow-up infrared image target identification, tracking, detection etc. to process, it is necessary to enhancing treatment is carried out to infrared image.

Infrared image enhancing method has a lot, can be divided into transpositions domain and the major class of space domain method two.Transpositions domain is by elder generation Image carries out certain conversion (such as Fourier transformation, wavelet transformation), then carries out enhancing treatment to conversion coefficient, finally carries out Anti- transformation changes acquisition enhancing image, and such method enhancing effect is good, but due to needing to carry out positive inverse transform, operand is big.It is empty Between domain method be that enhancing treatment directly is carried out to pixel, operand is small, but enhancing effect is generally below transpositions domain.Histogram equalization Change is a kind of conventional spatial domain Enhancement Method, and the method tries to achieve image Nogata by each gray-level pixels quantity of statistical picture Figure information, carries out grey scale mapping based on cumulative distribution function converter technique, so as to reach the overall contrast of enhancing image, makes The purpose of image clearly.Histogram equalization have computing it is simple, it is good to visible light image enhancement effects the characteristics of.Due to red Outer image background and noise occupy substantial amounts of gray level, and the gray level of target is less, and infrared image is through histogram equalization Afterwards, the contrast of background and noise is enhanced, and the contrast of target is lowered, and bright phenomenon occurs in gray area high.Cause This, general histogram equalization is not suitable for the enhancing of infrared image.

The content of the invention

It is an object of the invention to be directed to above-mentioned the deficiencies in the prior art, there is provided a kind of objective fuzzy of use gray level is subordinate to Category degree is adjusted to histogram, and the gray level of background and noise is effectively suppressed, and equalization processing is then carried out again, With and improve infrared image enhancement effect, and the small histogram FUZZY WEIGHTED adjustment of amount of calculation increases with infrared image in a balanced way Strong method.

The technical proposal of the invention is realized in this way：A kind of histogram FUZZY WEIGHTED adjustment increases with infrared image in a balanced way Strong method, it is characterised in that the method comprises the steps：

(1) original infrared image histogram p is calculated_r(r_k)；

(2) the objective fuzzy degree of membership μ (r of gray level are calculated_k)；Fuzzy theory is specially introduced, grey scale pixel value point Into smaller and larger two fuzzy sets, smaller gray value constitutes blurred background collection, and larger gray value constitutes objective fuzzy collection；Adopt Belong to the measurement of target area possibility as pixel with the fuzzy membership of target gray；Using following S types fuzzy membership functions：

In formula, r_maxIt is the maximum gray scale of image, r_minIt is minimal gray level；r_qTo get over a little, i.e. r=r_qWhen, μ (r)= 0.5；Using the average gray value of image as getting over a little, calculating formula is：

(3) according to objective fuzzy degree of membership μ (r_k) to histogram p_r(r_k) adjustment is weighted, obtain new histogram p '_r (r_k), formula is：

p'_r(r_k)=μ^α(r_k)×p_r(r_k) k=0,1 ..., L-1,

In formula, μ (r_k)、p_r(r_k) it is respectively k-th gray level r_kFuzzy membership and histogram, p '_r(r_k) it is weighting Histogram after adjustment, α is constant；

(4) to new histogram p '_r(r_k) equalization processing is carried out, obtain enhancing image I'(x).

Above-mentioned histogram FUZZY WEIGHTED is adjusted with infrared image enhancing method in a balanced way, and step (1) calculates original red Outer image histogram p_r(r_k) be specially：If the total pixel number of original infrared image I (x) is N, intensity profile scope is [0, L- 1], r is made_kRepresent k-th gray level, n_kRepresent r_kThe number of the pixel of appearance, then r_kHistogram p_r(r_k) be：

p_r(r_k)=n_k/ N k=0,1 ..., L-1.

Above-mentioned histogram FUZZY WEIGHTED is adjusted with infrared image enhancing method in a balanced way, and step (4) is described to new straight Scheme p ' in side_r(r_k) equalization processing is carried out, obtain enhancing image I'(x) be specially：

Calculate gray accumulation distribution function S'_k, computing formula is：

According to gray accumulation distribution function, calculating weighted histogram equalization transforming function transformation function is：

r'_k=round ((M-1) S '_k/S′_L-1)

K=0,1 ..., L-1,

In formula, r'_kIt is enhanced k-th gray level, L is the number of greyscale levels of original image, and M is the ash of image after equalization Degree series.

In above-mentioned histogram FUZZY WEIGHTED adjustment and infrared image enhancing method in a balanced way, in step (4), number of greyscale levels For 8bit images, value is 256 to M.

Above-mentioned histogram FUZZY WEIGHTED is adjusted with infrared image enhancing method in a balanced way, and in step (3), constant is increasing The strong factor, α is bigger, and stronger to histogrammic corrective action, the humidification to image is also stronger, and α spans are:1<α<2； When α=0, p '_r(r_k)=p_r(r_k), the present invention deteriorates to common histogram equalization.

The present invention is first adjusted using the objective fuzzy degree of membership of gray level using after the above method to histogram, will The gray level of background and noise is effectively suppressed, and equalization processing is then carried out again, so as to improve infrared image enhancement Effect, and amount of calculation is small.Using fuzzy membership functions so that integrated curved is S types, functional, and gets over point r_qCan Independent selection, getting over of overcoming that other type fuzzy membership functions (such as standard fuzzy S function, classics PAL functions) are present a little is received The constraint of gray scale upper and lower limit lacks limit.

Brief description of the drawings

The present invention is described in further detail for embodiment in below in conjunction with the accompanying drawings, but does not constitute to of the invention Any limitation.

Fig. 1 is schematic flow sheet of the invention；

Fig. 2 is one of experimental result of the embodiment of the present invention；

Fig. 3 is the two of the experimental result of the embodiment of the present invention.

Specific embodiment

Refering to shown in Fig. 1, a kind of histogram FUZZY WEIGHTED adjustment of the invention and in a balanced way infrared image enhancing method should Method comprises the steps：

(1) original infrared image histogram p is calculated_r(r_k)：If the total pixel number of original infrared image I (x) is N, gray scale point Cloth scope is [0, L-1], makes r_kRepresent k-th gray level, n_kRepresent r_kThe number of the pixel of appearance, then r_kHistogram p_r(r_k) For：

p_r(r_k)=n_k/ N k=0,1 ..., L-1；

(2) the objective fuzzy degree of membership μ (r of gray level are calculated_k)；Fuzzy theory is specially introduced, grey scale pixel value point Into smaller and larger two fuzzy sets, smaller gray value constitutes blurred background collection, and larger gray value constitutes objective fuzzy collection；Adopt Belong to the measurement of target area possibility as pixel with the fuzzy membership of target gray；Using following S types fuzzy membership functions：

In formula, r_maxIt is the maximum gray scale of image, r_minIt is minimal gray level；r_qTo get over a little, i.e. r=r_qWhen, μ (r)= 0.5；Using the average gray value of image as getting over a little, calculating formula is：

(3) according to objective fuzzy degree of membership μ (r_k) to histogram p_r(r_k) adjustment is weighted, obtain new histogram p '_r (r_k), formula is：

p'_r(r_k)=μ^α(r_k)×p_r(r_k) k=0,1 ..., L-1,

In formula, μ (r_k)、p_r(r_k) it is respectively k-th gray level r_kFuzzy membership and histogram, p '_r(r_k) it is weighting Histogram after adjustment, α is constant, and constant is enhancer, and α is bigger, stronger to histogrammic corrective action, the enhancing to image Effect is also stronger, and α spans are:1<α<2；When α=0, p '_r(r_k)=p_r(r_k), it is equal that the present invention deteriorates to common histogram Weighing apparatusization.

The essence of infrared image is the Temperature Distribution for characterizing scenery, and the temperature of background area is relatively low, and image intensity value is relatively It is small；The temperature of target area is higher, and image intensity value is relatively large.It is to the enhancing of target and to the back of the body to infrared image enhancement purpose The suppression of scape.It is this it is contemplated that the complexity of practical IR image, it is very difficult to accurate to divide target area and background area, introduces Grey scale pixel value, is divided into smaller and larger two fuzzy sets by fuzzy theory, i.e., smaller gray value constitutes blurred background collection, compared with High-gray level value constitutes objective fuzzy collection.Fuzzy membership using target gray belongs to the degree of target area possibility as pixel Amount.

Fuzzy membership functions advantage is that integrated curved is S types, functional, and gets over point r_qCan independently select, overcome What other type fuzzy membership functions (such as standard fuzzy S function, classics PAL functions) were present get over a little receives gray scale upper and lower limit about Beam lacks limit.

(4) to new histogram p '_r(r_k) equalization processing is carried out, obtain enhancing image I'(x), specially：

Calculate gray accumulation distribution function S'_k, computing formula is：

According to gray accumulation distribution function, calculating weighted histogram equalization transforming function transformation function is：

r'_k=round ((M-1) S '_k/S′_L-1)

K=0,1 ..., L-1,

In formula, r'_kIt is enhanced k-th gray level, L is the number of greyscale levels of original image, and M is the ash of image after equalization Degree series.For 8bit images, value is 256 to number of greyscale levels M.

Experimental example enhancing effect compares

The present invention carries out enhancing effect with tradition histogram equalization and plateau equalization popular both at home and abroad at present Fruit contrast experiment.Using 2 evaluation indexes of contrast and index of fuzziness, the objective quantification of enhancing effect is carried out to it.

(1) contrast (Contrst).Computing formula is as follows：

Contrst values are bigger, and the contrast of image is higher, and visual quality is better.

(2) index of fuzziness (Fuzzy exponent).It is defined as：

Q (x)=sin (0.5 π (1-I (x)/I_max))

In formula, I_maxIt is the maximum gradation value of image.According to the definition of index of fuzziness, index of fuzziness F_BIt is smaller, image It is more clear.

Experimental Hardware environment is：The x2 Dual core Processor 5200+2.7GHz of AMD Athlon (tm) 64, 1.75GB internal memories；Software environment is:Windows XP Sp2+Matlab R2009b.Using two width power equipment Infrared Thermograms As test image.The enhancing effect image of three kinds of methods is shown in Fig. 2~Fig. 3, and experimental result data is as shown in table 1~2.

1 Fig. 2 (a) of table, three kinds of Enhancement Method evaluation index contrasts

Enhancement Method	Contrast (Contrst)
			Histogram equalization	5337.8	0.2044
Plateau equalization	3908.7	0.1869
			Context of methods	5362.5	0.1696

2 Fig. 3 (a) of table, three kinds of Enhancement Method evaluation index contrasts

Enhancement Method	Contrast (Contrst)
			Histogram equalization	5300.4	0.2027
Plateau equalization	5015.9	0.1912
			The present invention	6115.6	0.1489

From Fig. 2 and Fig. 3, original image gray scale narrow dynamic range, picture contrast and definition are low；Histogram equalization Image, gray scale dynamic model is big, but overall partially bright, the loss in detail of image, and some mixed and disorderly ambient noises are exaggerated；Plateau histogram The enhancing effect of equalization is better than histogram equalization, but clutter and noise in background are still larger；The inventive method increases The gray scale dynamic range of strong image is big, and picture contrast is high, and target sharpness is high, and ambient noise is far below plateau equalization Change and conventional histogram equalization, the subjective vision effect for strengthening image is better than other two kinds of methods.From Tables 1 and 2, this The contrast value of inventive method enhancing image is maximum, and index of fuzziness is minimum, and this is just consistent with the result of subjective assessment. Experimental result illustrates that the enhancing effect of the inventive method is better than other two kinds of congenic methods

Embodiment provided above is better embodiment of the invention, only for the convenient explanation present invention, not to this hair It is bright to make any formal limitation, any those of ordinary skill in the art, if putting forward skill the present invention is not departed from In the range of art feature, using the Equivalent embodiments for locally being changed done by disclosed technology contents or modify, and Without departing from technical characteristic content of the invention, still fall within the range of the technology of the present invention feature.

Claims (5)

1. a kind of histogram FUZZY WEIGHTED adjustment and in a balanced way infrared image enhancing method, it is characterised in that under the method includes State step：

(1) original infrared image histogram p is calculated_r(r_k)；

(2) the objective fuzzy degree of membership μ (r of gray level are calculated_k)；Fuzzy theory is specially introduced, grey scale pixel value is divided into smaller With larger two fuzzy sets, smaller gray value composition blurred background collection, larger gray value composition objective fuzzy collection；Using target The fuzzy membership of gray scale belongs to the measurement of target area possibility as pixel；Using following S types fuzzy membership functions：

$\mathrm{\&mu;}\left(r\right)=\left\{\begin{array}{ccc}\frac{r-r_{\min }}{2(r_q-r_{\min })}sin(\frac{\mathrm{\&pi;}}{2}\&times;\frac{r-r_{\min }}{r_q-r_{\min }})& if& r_{\min }\&le;r<r_q\\ 1+\frac{r-r_{max}}{2(r_{max}-r_q)}cos(\frac{\mathrm{\&pi;}}{2}\&times;\frac{r-r_q}{r_{\max }-r_q})& if& r_q\&le;r\&le;r_{max}\end{array}\right.,$

In formula, r_maxIt is the maximum gray scale of image, r_minIt is minimal gray level；r_qTo get over a little, i.e. r=r_qWhen, μ (r)=0.5； Using the average gray value of image as getting over a little, calculating formula is：

$r_q=\frac{1}{N}\underset{x=1}{\overset{N}{\&Sigma;}}I\left(x\right);$

In formula, I (x) is original infrared image；N is the total pixel number of original infrared image I (x)；

(3) according to objective fuzzy degree of membership μ (r_k) to histogram p_r(r_k) adjustment is weighted, obtain new histogram p_r′(r_k), it is public Formula is：

p'_r(r_k)=μ^α(r_k)×p_r(r_k) k=0,1 ..., L-1,

In formula, μ (r_k)、p_r(r_k) it is respectively k-th gray level r_kFuzzy membership and histogram, p_r′(r_k) it is weighting adjustment Histogram afterwards, α is constant；L is the number of greyscale levels of original image；

(4) to new histogram p_r′(r_k) equalization processing is carried out, obtain enhancing image I'(x).

2. histogram FUZZY WEIGHTED adjustment according to claim 1 and in a balanced way infrared image enhancing method, its feature exist In step (1) calculates original infrared image histogram p_r(r_k) be specially：If the total pixel number of original infrared image I (x) is N, Intensity profile scope is [0, L-1], makes r_kRepresent k-th gray level, n_kRepresent r_kThe number of the pixel of appearance, then r_kNogata Figure p_r(r_k) be：

p_r(r_k)=n_k/ N k=0,1 ..., L-1.

3. histogram FUZZY WEIGHTED adjustment according to claim 1 and in a balanced way infrared image enhancing method, its feature exist In step (4) is described to new histogram p_r′(r_k) equalization processing is carried out, obtain enhancing image I'(x) be specially：

Calculate gray accumulation distribution function S'_k, computing formula is：

${S^{\&prime;}}_k=\underset{j=0}{\overset{k}{\&Sigma;}}{p^{\&prime;}}_r\left(r_j\right)=\underset{j=0}{\overset{k}{\&Sigma;}}{\mathrm{\&mu;}}^{\mathrm{\&alpha;}}\left(r_j\right)p_r\left(r_j\right);$

According to gray accumulation distribution function, calculating weighted histogram equalization transforming function transformation function is：

r'_k=round ((M-1) S_k′/S′_L-1)

K=0,1 ..., L-1,

In formula, r'_kIt is enhanced k-th gray level, L is the number of greyscale levels of original image, and M is the gray level of image after equalization Number.

4. histogram FUZZY WEIGHTED adjustment according to claim 3 and in a balanced way infrared image enhancing method, its feature exist In in step (4), for 8bit images, value is 256 to number of greyscale levels M.

5. histogram FUZZY WEIGHTED adjustment according to claim 1 and in a balanced way infrared image enhancing method, its feature exist In in step (3), constant is enhancer, and α is bigger, stronger to histogrammic corrective action, and the humidification to image is also got over By force, α spans are:1 ＜ α ＜ 2；When α=0, p_r′(r_k)=p_r(r_k), the present invention deteriorates to common histogram equalization.