CN106023190A - Automatic focusing image focusing degree evaluation method - Google Patents

Abstract

The invention discloses an automatic focusing image focusing degree evaluation method. The method adopts a phase consistency image and gradient image combined method, solves the problem of peak collapse caused when a phase consistency method is adopted as a focusing evaluation function, has good expression performance no matter in a normal scene, or a scene, where signal-to-noise ratio is low or brightness is changed, is high in practical value, and has the advantages of good unbiasedness, unimodality, sensitivity and stability.

Description

A kind of evaluation methodology of auto-focusing image quasi-Jiao degree

Technical field

The present invention relates to image quality evaluating method, particularly relate to the evaluation side of a kind of auto-focusing image quasi-Jiao degree Method.

Background technology

Along with digital picture and the fast development of multimedia technology, various types of optical imaging systems also get more and more, Autofocus Technology plays vital effect in optical imaging systems.Autofocus Technology mainly has two kinds, is respectively Active focusing and passive type focusing.Actively focusing needs to add extra range-measurement system in systems, adds the one-tenth of system Basis and the complexity of system, simultaneously for having glass or having the scene of shelter, range finding there will be problem, therefore, active Focusing technology is replaced by self-action focusing technology at leisure.

Passive focusing is to judge out of focus degree by the fog-level of image obtained in analysis focus process, thus refers to Lead focus process.This Atomatic focusing method is to utilize certain Digital Image Processing algorithm, obtains image definition eigenvalue, And control motor according to this eigenvalue, to drive eyeglass or imageing sensor to change position, do corresponding adjustment, until this Till one eigenvalue meets a certain condition made an appointment.Therefore, the auto focus evaluation function of superperformance is selected to judge The fog-level of image thus to guide autofocus system to focus exactly be the key of problem.Weight due to focusing evaluation function The property wanted, the evaluation of image definition has become a popular research field.Preferably focusing evaluation function should have unbiased Property, unimodality, susceptiveness, stability, some the most also should to scene light and shade change and noise have preferably Robustness.Existing auto focus evaluation function is very sensitive to the brightness and contrast of scene, because its principle is utilization figure The intensity of picture judges out of focus degree.And along with imaging system popularizing in people live, people like at various environment Lower use focusing system, uses the situation of focusing system also to get more and more under or scene that signal to noise ratio is relatively low inconsistent in brightness. In consideration of it, still have the evaluation function of good behaviour performance to become the urgent needs of people this kind of scene.

Summary of the invention

Present invention aims to use phase equalization method to have what peak value subsided as focusing evaluation function Problem, proposes the evaluation methodology of a kind of auto-focusing image quasi-Jiao degree.

It is an object of the invention to be achieved through the following technical solutions: the evaluation side of a kind of auto-focusing image quasi-Jiao degree Method, the method comprises the following steps:

(1) image I is obtained；

(2) from image I, focusing window region I is chosen_c；

(3) utilize log-Gabor small echo to calculate focusing window region I_cPhase equalization collection of illustrative plates PC, each pixel Phase equalization is defined as:

$PC(x,y)=\frac{\underset{i=1}{\overset{m}{\Σ}}\underset{j=1}{\overset{n}{\Σ}}{W}_i(x,y)f_p\{A_{ji}(x,y)\[{\mathrm{\Δ\φ}}_{ji}(x,y)\]-T_i\}}{\underset{i=1}{\overset{m}{\Σ}}\underset{i=1}{\overset{n}{\Σ}}{A}_{ji}(x,y)}---\left(1\right)$

Wherein, m and n represents the direction yardstick number of Log-Gabor wavelet filter, A respectively_ji(x, y) and φ_ji(x,y) The amplitude being illustrated respectively in the jth logarithmic scale of i-th direction and local phase deviation.T_iIt is estimating on i-th direction Meter noise, W_i(x y) is the weighting function on i-th direction, f_p(u)=(u > 0) u.

Δ φ (x, y)=cos [φ_r(x,y)-φ(x,y)]-|sin[φ_r(x,y)-φ(x,y)]| (2)

(x is y) at (x, y) the local phase weighted mean of position, φ to φ_r(x, y) represent local phase (x, y) The r level Fourier expansion series component of position；

(4) utilize gradient calculation method obtain phase equalization collection of illustrative plates PC gradient magnitude spectrum G:

(5) suing for peace all gradient magnitudes in G, as final focusing function evaluation of estimate, gradient magnitude summation is the highest, Current image location is closer to focus position.

The beneficial effects of the present invention is: 1. solve scene by phase equalization collection of illustrative plates PC and there is brightness flop and make an uproar Problem in terms of sound, further combined with gradient calculation method overcome phase equalization method as during evaluation function in accurate burnt position Putting neighbouring curve and there is the problem subsided, the evaluation function curve finally given has good unbiasedness, unimodality, susceptiveness And stability, practical value is higher, can be widely used in and include brightness flop, and the situations such as signal to noise ratio is low are interior various Actual photographed scene.

Accompanying drawing explanation

Fig. 1 is the FB(flow block) of inventive method.

Fig. 2 is the schematic diagram of optional 5 width in one group of emulation out of focus sequence chart.

Fig. 3 is as the criterion burnt image.

Fig. 4 is the focusing window image chosen.

Fig. 5 is the phase equalization collection of illustrative plates of focusing window area image.

Fig. 6 is the gradient magnitude spectrum G of phase equalization collection of illustrative plates PC.

Fig. 7 is the evaluation function curve utilizing phase equalization method to obtain.

Fig. 8 is the evaluation function curve utilizing phase equalization and gradient associated methods to obtain.

Fig. 9 is the evaluation function curve utilizing existing method to obtain.

Detailed description of the invention

The evaluation methodology of the present invention a kind of auto-focusing image quasi-Jiao degree, the method uses phase equalization collection of illustrative plates and ladder The degree method that combines of collection of illustrative plates, solves asking of having when using phase equalization method as focusing evaluation function that peak value subsides Topic, the scene that or brightness relatively low for normal scene or signal to noise ratio changes all has good performance performance, practical valency Being worth higher, range of application is relatively wide, has good unbiasedness, unimodality, susceptiveness, stability.

It is described in detail with example below in conjunction with the accompanying drawings:

Fig. 1 is the simple process block diagram of the inventive method.Below in conjunction with embodiment, the invention will be further described.

The present embodiment is as a example by one group of emulation out of focus sequence chart, and Fig. 2 is for arbitrarily to choose 5 width figures in one group of out of focus sequence chart As schematic diagram, wherein (a)-(e) is respectively the 1st width, the 9th width, the 17th width, the 25th width, the 33rd width out of focus sequence analogous diagram, wherein 17th width is as the criterion burnt figure.Wherein, taper into from the 1st width～the 17th width defocusing amount, from the 17th width～the 33rd width defocusing amount gradually Become big.

The present embodiment carries out defocusing amount evaluation initially with existing method to 33 width figures, owing to brightness is inconsistent, existing Method can not truly reflect the defocusing amount of each figure, as shown in Figure 9.

The present embodiment have employed the method for simple employing phase equalization method and the present invention further to above-mentioned 33 width Image is evaluated as follows (as a example by the 17th width image):

(1) image I is obtained as shown in Figure 3；

(2) from image I, choose the focusing window region I of 200 × 200 pixels as shown in Figure 4_c；

(3) utilize log-Gabor small echo to calculate focusing window area image I_cPhase equalization collection of illustrative plates PC, 2D scheme In Xiang, the phase equalization of each pixel is defined as:

$PC(x,y)=\frac{\underset{i=1}{\overset{m}{\Σ}}\underset{j=1}{\overset{n}{\Σ}}{W}_i(x,y)f_p\{A_{ji}(x,y)\[{\mathrm{\Δ\φ}}_{ji}(x,y)\]-T_i\}}{\underset{i=1}{\overset{m}{\Σ}}\underset{i=1}{\overset{n}{\Σ}}{A}_{ji}(x,y)}---\left(1\right)$

Wherein, m and n represents the direction yardstick number of Log-Gabor wavelet filter, A respectively_ji(x, y) and φ_ji(x,y) The amplitude being illustrated respectively in the jth logarithmic scale of i-th direction and local phase deviation.T_iIt is estimating on i-th direction Meter noise, W_i(x y) is the weighting function on i-th direction, f_p(u)=(u > 0) u.

Δ φ (x, y)=cos [φ_r(x,y)-φ(x,y)]-|sin[φ_r(x,y)-φ(x,y)]| (2)

(x is y) at (x, y) the local phase weighted mean of position, φ to φ_r(x, y) represent local phase (x, y) The r level Fourier expansion series component of position, phase equalization collection of illustrative plates PC is as shown in Figure 5；

(4) utilize gradient calculation method obtain phase equalization collection of illustrative plates PC gradient magnitude spectrum G, as shown in Figure 6,

$G=\sqrt{G_x^2+G_y^2}---\left(3\right)$

Wherein, G_xWith G_yRepresent phase equalization collection of illustrative plates PC partial derivative in the horizontal and vertical directions respectively；

(5) all gradient magnitudes in G are sued for peace, as final focusing function evaluation of estimate.

Repeat the above steps carries out definition evaluation of estimate calculating to the 33 width out of focus sequence analogous diagram obtained, and obtains this group The evaluation function curve of out of focus sequence focusing window, as shown in Figure 8.Preferably solve as shown in Figure 7 use merely phase place The problem that peak value subsides is there is in coherence method as focusing evaluation function.

The present invention proposes the auto focus evaluation function computational methods that a kind of phase equalization is combined, the method with gradient Solve employing phase equalization method and there is, as focusing evaluation function, the problem that peak value subsides, for normal scene the most also Being that relatively low or brightness change the scene of signal to noise ratio all has good performance performance, practical value is higher, and range of application is relatively wide, tool There are good unbiasedness, unimodality, susceptiveness, stability.

Claims (1)

1. the evaluation methodology of auto-focusing image quasi-Jiao degree, it is characterised in that the method comprises the following steps:

(1) image I is obtained；

(2) from image I, focusing window region I is chosen_c；

(3) utilize log-Gabor small echo to calculate focusing window region I_cPhase equalization collection of illustrative plates PC, the phase place of each pixel Conformance definition is:

$PC(x,y)=\frac{\underset{i=1}{\overset{m}{\Σ}}\underset{j=1}{\overset{n}{\Σ}}{W}_i(x,y)f_p\{A_{ji}(x,y)\[{\mathrm{\Δ\φ}}_{ji}(x,y)\]-T_i\}}{\underset{i=1}{\overset{m}{\Σ}}\underset{i=1}{\overset{n}{\Σ}}{A}_{ji}(x,y)}---\left(1\right)$

Wherein, m and n represents the direction yardstick number of Log-Gabor wavelet filter, A respectively_ji(x, y) and φ_ji(x, y) respectively Represent the amplitude in the jth logarithmic scale of i-th direction and local phase deviation.T_iIt is that the estimation on i-th direction is made an uproar Sound, W_i(x y) is the weighting function on i-th direction, f_p(u)=(u > 0) u.

Δ φ (x, y)=cos [φ_r(x,y)-φ(x,y)]-|sin[φ_r(x,y)-φ(x,y)]| (2)

(x is y) at (x, y) the local phase weighted mean of position, φ to φ_r(x y) represents local phase in (x, y) position R level Fourier expansion series component；

(4) utilize gradient calculation method obtain phase equalization collection of illustrative plates PC gradient magnitude spectrum G:

(5) suing for peace all gradient magnitudes in G, as final focusing function evaluation of estimate, gradient magnitude summation is the highest, currently Picture position is closer to focus position.