CN100508628C - Digitized test method for CCD camera signal-to-noise ratio - Google Patents

Abstract

The present invention provides a digitized testing method for the signal-to-noise ratio of a CCD video camera. The fundamental principle of the present invention is that the video camera images the high- uniform illuminating white and clean target, and the fluctuation of the imaging signal reflects the noise of the video camera. The video signal is quantified into a digital image through an image acquisition card, the noise signal is separated by a specially designed digital high-pass filter, and the root-mean-square value of the noise signal is calculated as the noise intensity. The image acquisition card brings a quantizing noise into the original video signal, the quantizing noise is modified to estimate the value of the noise in the original video signal, and thereby the signal-to-noise ratio can be worked out. The present invention achieves the fast speed and objective testing method of the television camera signal-to-noise ratio based on digital image acquisition, the testing result is as good as the result of the video analyzer, thus the method can be used as a replacement method for video camera signal-to-noise testing.

Description

A kind of ccd video camera signal to noise ratio digitized test method

Technical field

The invention belongs in the photoelectricity test field the evaluation and test of discrete light electric imaging system performance, realized a kind of the ccd video camera signal to noise ratio being carried out the digitlization objective measurement method, be applied to the evaluation and test of visible light ccd video camera.

Background technology

Regulation according to IEC1146-1:1994 " video camera method of measurement: non-broadcasting single-sensor video camera ", the method of measurement of signal to noise ratio is: insert the video noise measuring instrument at the video camera output, disconnect in the camera chin and controlling automatically and correcting circuit, connect video noise measuring instrument 6MHz low pass filter and 100KHz high pass filter, the gray scale resolution chart of video camera picked-up EIAJ-C3, the video amplitude that makes the video camera output is 100% white level, reads signal to noise ratio on the video noise measuring instrument.

The expression formula of signal to noise ratio (S/N) is: $S/N=20\lg \frac{V_{\mathrm{ref}}}{V_{\mathrm{Nrms}}}$

V wherein _RefBe 100% white level amplitude, promptly the amplitude between blanking level and the white level is 700mV to the Phase Alternation Line system video camera, and the TSC-system video camera then is 714mV.V _NrmsBe the noise effective value of reading from noise meter (representing) with root-mean-square value.The signal to noise ratio test provides respectively under 2000Lx illumination and sensitivity minimization illumination conditions usually.

The key of signal to noise ratio test is to determine noise intensity, and noise testing is generally read noise signal with visual method on oscilloscope, and the number reading method of employing has two kinds of quasi-peak value method and tangent methods.Because the random fluctuation of noise voltage, the root-mean-square value of visual definite noise voltage exists stronger subjectivity, and the measurement result between the different people differs greatly.The objective examination that novel video noise measuring instrument (as the TEK VM700A of Imtech) has been realized signal to noise ratio by the method for spectrum analysis, but certainty of measurement and sensitivity are all than higher equipment manufacturing cost costliness (equipment of measuring video signal-to-noise ratio is immediately approximately wanted 30,000 U.S. dollars).Other has the way of relatively economical to be to use a filter, filters picture signal, measures the noise level that stays.But required filter is worth nearly 10,000 U.S. dollars.Therefore above-mentioned method of measurement exists dependence operator's subjective judgement ability or the high defective of testing cost.

Summary of the invention:

The objective of the invention is to, overcome the defective of existing visual method signal to noise ratio measuring technology, a kind of new ccd video camera signal to noise ratio digitlization objective examination method based on Digital Image Processing is proposed, this method is carried out analyzing and processing by the image information to the even illumination target target plate gathered, and then calculates the signal to noise ratio of video camera.

The present invention includes: evenly lighting lamphouse, ccd video camera to be measured, image pick-up card, data processing are with computer and be used for the oscilloscope of auxiliary reading; Ccd video camera wherein to be measured is positioned at even lighting lamphouse exit, is used to take the transmission-type blank in even lighting lamphouse exit, and then the fluctuation in the imaging signal has reflected the noise of ccd video camera to be measured; Image pick-up card is connected on the ccd video camera to be measured, and vision signal is quantified as digital picture sends into data processing and handle with computer; The The data digital high-pass filter that data processing is sent image pick-up card here with computer is handled, and isolates noise signal, and the root-mean-square value of calculating noise is as noise intensity; The described oscilloscope that is used for auxiliary reading is connected ccd video camera to be measured, and conventional noise testing means are provided, for the digitlization snr measurement provides contrast and reference;

Concrete method of testing is as follows:

1) be 2000Lx in illumination, colour temperature is under the standard illumination conditions of 3100K, regulates the aperture of ccd video camera camera lens to be measured, takes the outlet blank of even lighting lamphouse; It is saturated to guarantee that output image does not have: under N position quantification situation, and average gray＜2 after promptly guaranteeing to quantize ^N-1, and the pixel maximum gradation value is also less than 2 ^N-1; Grasp two frame of digital image P1 and P2 respectively;

2) digital picture P1 and P2 are subtracted each other as gray value by pixel, obtain digital picture P3, no longer contain of the influence of the even property of lamp box uneven illumination among the P3 noise testing;

3) use high pass filter H that image P3 is done digital filtering, obtain noise image P4;

The high pass filter template H that the signal to noise ratio test is selected for use is:

$H=\left[\begin{array}{ccc}1/6& -1/3& 1/6\\ -1/3& 2/3& -1/3\\ 1/6& -1/3& 1/6\end{array}\right]$

This filter not only can effectively leach noise, and there is not enlarge-effect in the noise to original image simultaneously;

4) variances sigma of calculating noise image P4 ², when adopting the N position to quantize, the computing formula of ccd video camera signal to noise ratio to be measured (S/N) is:

$S/N=20\lg \frac{2^N-1}{N_{\mathrm{rms}}}=20\lg \frac{2^N-1}{\sqrt{{\mathrm{\σ}}^2/2-1/12}},$

N wherein _RmsRoot-mean-square value for noise;

When image pick-up card quantization digit N=8, described snr computation is as follows:

$S/N=20\lg \frac{255}{N_{\mathrm{rms}}}=20\lg \frac{255}{\sqrt{{\mathrm{\σ}}^2/2-1/12}}$

In order to give prominence to the noise signal in the video, lighting lamphouse need provide the area source of high uniformity, its irregularity in brightness≤5%; In order to test the signal to noise ratio of ccd video camera to be measured under different illumination conditions, evenly the outlet illumination of lighting lamphouse can be regulated.

Need in the test to select the image pick-up card quantisation depth, promptly guarantee to quantize signal to noise ratio much larger than ccd video camera signal to noise ratio to be measured according to ccd video camera signal to noise ratio scope to be measured; If quantized level is the N position, quantized interval is a Δ, and then the maximum signal amplitude value of Liang Huaing is 2 ^NThe quantification snr computation that Δ, quantizing process are brought into is as follows:

$S/N=20\lg \frac{V_{\mathrm{ref}}}{N_q}=20\lg \frac{2^N\mathrm{\Δ}}{\sqrt{{\mathrm{\Δ}}^2/12}}=(6.02N+10.8)\mathrm{dB},$

N wherein _qRoot-mean-square value for the image pick-up card quantizing noise.

More than in all mathematical formulaes the implication explanation in this manual of symbol be consistent.

The invention has the beneficial effects as follows objective, quick, the accurate measurement that has realized the video camera signal to noise ratio with very low cost, have practical value.

Description of drawings

Fig. 1-television camera signal-to-noise test system figure; Wherein: 1-evenly lighting lamphouse 2-ccd video camera 3-image pick-up card 4-computer 5-oscilloscope to be measured;

Fig. 2-a: the digital picture P1 of the even illumination target of camera acquisition;

B: the gray scale isometric chart of corresponding P1;

Fig. 3-a: the digital picture P2 of the even illumination target of camera acquisition;

B: the gray scale isometric chart of corresponding P2;

Fig. 4-a:P1 and P2 subtract each other the digital picture P3 that obtains by pixel;

B: the gray scale isometric chart of corresponding P3;

Fig. 5-a: use template H that P3 is the noise image P4 that digital filtering obtains;

B: the gray scale isometric chart of corresponding P4;

The frequency domain filter that Fig. 6-template H (airspace filter device) is corresponding.

Embodiment

The invention will be further described below in conjunction with drawings and Examples.For convenience of description, this to sentence 8 bit image capture cards be that example is introduced:

As shown in Figure 1, according to the ccd video camera signal to noise ratio digital test system of the present invention design mainly by even lighting lamphouse 1, ccd video camera to be measured 2, image pick-up card 3, data processing with computer 4 and be used for assisting the oscilloscope 5 etc. of reading partly to constitute.

Basic implementation process is: by the transmission-type blank in the even lighting lamphouse of ccd video camera 2 shootings to be measured 1 exit, then the fluctuation in the imaging signal has reflected the noise of ccd video camera 2 to be measured.By image pick-up card 3 vision signal of ccd video camera 2 to be measured being quantified as digital picture sends into data processing and handles with computer 4.Adopt specially designed digital high-pass filter to isolate noise signal, the root-mean-square value of calculating noise is as noise intensity.Brought quantizing noise into because of the raw video signal of 3 pairs of ccd video cameras 2 to be measured of image pick-up card, quantizing noise has been revised the noise size in the raw video signal that promptly can estimate ccd video camera 2 to be measured.The oscilloscope 5 that is used for auxiliary reading uses conventional means to measure the noise of ccd video camera 2 to be measured, for the digitlization snr measurement provides contrast and reference.

1) even lighting lamphouse

It is 2000Lx that even lighting lamphouse can provide illumination, and colour temperature is the standard illumination conditions of 3100K, and in order to give prominence to the noise signal in the video, evenly lighting lamphouse needs to provide the area source of high uniformity, its irregularity in brightness≤5%.This can realize by evenly place many group fluorescent tubes in multilayer opal glass plate back, also can directly select bigbore integrating sphere illumination for use.In order to test the signal to noise ratio of ccd video camera to be measured under different illumination conditions, evenly the outlet illumination of lighting lamphouse can be regulated.

2) selection of image pick-up card quantisation depth

Image pick-up card is actually A/D converter, and its quantization digit has determined the size of quantizing noise.If quantized interval is a Δ, quantization error distributes [Δ/2, Δ/2] interval interior obedience evenly, then because noise variance (D (the N that quantification is brought into _q)) can be calculated as follows:

$D\left(N_q\right)=\underset{-\mathrm{\Δ}/2}{\overset{\mathrm{\Δ}/2}{\∫}}\frac{1}{\mathrm{\Δ}}{x}^2\mathrm{dx}=\frac{{\mathrm{\Δ}}^2}{12}---\left(1\right)$

If quantized level is the N position, then the maximum signal amplitude value of Liang Huaing is 2 ^NΔ, so the quantification snr computation that quantizing process is brought into is as follows:

$S/N=20\lg \frac{V_{\mathrm{ref}}}{N_q}=20\lg \frac{2^N\mathrm{\Δ}}{\sqrt{{\mathrm{\Δ}}^2/12}}=(6.02N+10.8)\mathrm{dB}---\left(2\right)$

As seen, increasing quantization digit can the lower quantization noise.Quantization digit of every increase approximately can improve signal to noise ratio 6dB.(corresponding 256 gray scales) quantize the about 59dB of signal to noise ratio, thereby can satisfy most of television camera signal to noise ratio testing requirement (signal to noise ratios about 45～55dB) during 8 quantifications.If the signal to noise ratio of ccd video camera to be measured own very high (〉=59dB), just must selecting for use more, the image pick-up card of high quantization level could quantize noise signal.

3) the basic implementation process of television camera signal to noise ratio digital test is as follows:

A) be 2000Lx in illumination, colour temperature is under the standard illumination conditions of 3100K, regulate the aperture of ccd video camera camera lens to be measured, take the outlet blank of even lighting lamphouse, it is saturated (under 8 quantification situations to guarantee that output image does not have, average gray＜255 after promptly guaranteeing to quantize, and the pixel maximum gradation value is also≤255.Grasp two frame of digital image P1 and P2 respectively, as shown in Figures 2 and 3.

B) digital picture P1 and P2 are subtracted each other as gray value by pixel, obtain new digital picture P3, as shown in Figure 4.No longer contain of the influence of the even property of lamp box uneven illumination among the new figure to noise testing.

C) use high pass filter H that digital picture P3 is done digital filtering, obtain noise image P4, as shown in Figure 5.

The high pass filter template of selecting for use is:

$H=\left[\begin{array}{ccc}1/6& -1/3& 1/6\\ -1/3& 2/3& -1/3\\ 1/6& -1/3& 1/6\end{array}\right]$

The frequency domain filter of filters H correspondence as shown in Figure 6, as seen this filter is a kind of typical high pass filter.Use this template that image is equivalent to as convolution algorithm image is made airspace filter, the calculating process of airspace filter is:

$f\′(x,y)=\frac{1}{6}f(x-1,y-1)-\frac{1}{3}f(x-1,y)+\frac{1}{6}f(x-1,y+1)$

$-\frac{1}{3}f(x,y-1)+\frac{2}{3}f(x,y)-\frac{1}{3}f(x,y+1)$

$+\frac{1}{6}f(x+1,y-1)-\frac{1}{3}f(x+1,y)+\frac{1}{6}f(x+1,y+1)$

Wherein (x is that coordinate is that (x y) locates gray values of pixel points, and (x y) is the gray value of corresponding points after the filtering to f ' in the original image y) to f.

If the noise of each pixel is obeyed same distribution in the image, then filtered pixel noise variance is:

$D(f\′(x,y))=({\left(\frac{1}{6}\right)}^2\×4+{(-\frac{1}{3})}^2\×4+{\left(\frac{2}{3}\right)}^2)D\left(f(x,y)\right)=D\left(f(x,y)\right)$

Wherein D (f ' (x, y)) is filtered noise spot variance, and D ((x y) is the noise spot variance before the filtering to f.

As seen, select for use this filter not only noise effectively can be leached, there is not enlarge-effect in the noise to original image simultaneously.

D) variances sigma of calculating P4 ², wherein comprised ccd video camera intrinsic noise (N to be measured _Rms) and image pick-up card quantizing noise (N _q) influence, and D (N is arranged _Rms+ N _q)=σ ²/ 2.Separate because of ccd video camera intrinsic noise to be measured and image pick-up card quantizing noise, and by (1) formula as can be known, quantizing noise is about 1/12 (in variance) of a quantized interval.So ccd video camera intrinsic noise to be measured can be calculated as follows:

D(N _rms)＝σ ²/2-1/12

So when adopting 8 to quantize, ccd video camera snr computation formula to be measured is:

$S/N=20\lg \frac{255}{N_{\mathrm{rms}}}=20\lg \frac{255}{\sqrt{{\mathrm{\σ}}^2/2-1/12}}---\left(3\right)$

4) test result:

According to the ccd video camera signal-to-noise test system that above-mentioned method of testing is built, evenly the lighting lamphouse outlet size is 250 * 190mm, and after tested, its irregularity in brightness is less than 3%.

The signal to noise ratio test is carried out under 2000Lx illumination, uses 8 bit image capture cards, and tested object is the quick logical MTV-2821 B that a station symbol claims signal to noise ratio 50dB.Four width of cloth images of continuous acquisition are as shown in table 1 in the experiment:

In table 1, the experiment to four width of cloth images of even illumination target continuous acquisition

Picture number	The pixel average gray	The image variance
			No.1	212.08	9.48
No.2	212.19	9.46
			No.3	212.29	9.56
No.4	212.10	9.48

The result that obtains signal to noise ratio by (3) formula is as shown in table 2:

Table 2, the signal to noise ratio of using the signal to noise ratio digital measuring method to obtain

Process object	Filtered noise variance	Remove the noise variance of quantizing noise	Signal to noise ratio (dB)
				No.1—No.2	0.69	0.61	50.3
No.1—No.3	0.68	0.60	50.3
				No.1—No.4	0.67	0.60	50.3
No.2—No.3	0.68	0.60	50.3
				No.2—No.4	0.69	0.61	50.3
No.3—No.4	0.67	0.60	50.3

By test data as seen, the present invention has realized quick, the objective examination method of a kind of television camera signal to noise ratio based on digital image acquisition, its measurement reproducibility is good, test result is consistent with usefulness video analyzer measurement result, thereby can be used as a kind of alternative method of ccd video camera signal to noise ratio test.

Claims (4)

1. ccd video camera signal to noise ratio digitized test method is characterized in that: comprising: evenly lighting lamphouse (1), ccd video camera to be measured (2), image pick-up card (3), data processing are with computer (4) and be used for assisting the oscilloscope (5) of reading; Ccd video camera wherein to be measured (2) is positioned at even lighting lamphouse (1) exit, is used to take the transmission-type blank in even lighting lamphouse exit, and then the fluctuation in the imaging signal has reflected the noise of ccd video camera to be measured (2); Image pick-up card (3) is connected on the ccd video camera to be measured (2), and vision signal is quantified as digital picture sends into data processing and handle with computer (4); The The data digital high-pass filter that data processing is sent image pick-up card (3) here with computer (4) is handled, and isolates noise signal, and the root-mean-square value of calculating noise is as noise intensity; The described oscilloscope (5) that is used for auxiliary reading is connected ccd video camera to be measured (2), and conventional noise testing means are provided, for the digitlization snr measurement provides contrast and reference;

Concrete method of testing is as follows:

1) evenly lighting lamphouse (1) requirement is 2000Lx in illumination, and colour temperature is under the standard illumination conditions of 3100K, regulates the aperture of ccd video camera to be measured (2) camera lens, takes the outlet blank of even lighting lamphouse (1); It is saturated to guarantee that output image does not have; Data processing grasps two frame of digital image P1 and P2 with computer (4) respectively by image pick-up card (3);

2) digital picture P1 and P2 are subtracted each other as gray value by pixel, obtain digital picture P3;

3) use digital high-pass filter H that image P3 is done digital filtering, obtain noise image P4;

The high pass filter template H that the signal to noise ratio test is selected for use is:

$H=\left[\begin{array}{ccc}1/6& -1/3& 1/6\\ -1/3& 2/3& -1/3\\ 1/6& -1/3& 1/6\end{array}\right]$

4) variances sigma of calculating noise image P4 ², when adopting the N position to quantize, the snr computation formula of ccd video camera to be measured (2) is:

$S/N=20\lg \frac{2^N-1}{\sqrt{{\mathrm{\σ}}^2/2-1/12}}$

2. a kind of ccd video camera signal to noise ratio digitized test method according to claim 1 is characterized in that: when the quantization digit N=8 of image pick-up card (3), the snr computation of described ccd video camera to be measured (2) is as follows:

$S/N=20\lg \frac{255}{\sqrt{{\mathrm{\σ}}^2/2-1/12}}$

3. a kind of ccd video camera signal to noise ratio digitized test method according to claim 1 is characterized in that: in order to give prominence to the noise signal in the video, evenly lighting lamphouse (1) need provide the area source of high uniformity, its irregularity in brightness≤5%; Test the signal to noise ratio of ccd video camera to be measured (2) under different illumination conditions by the outlet illumination of regulating even lighting lamphouse (1).

4. a kind of ccd video camera signal to noise ratio digitized test method according to claim 1, it is characterized in that: select the quantisation depth of image pick-up card (3) according to the signal to noise ratio scope of ccd video camera to be measured (2), guarantee to quantize the signal to noise ratio of signal to noise ratio much larger than ccd video camera to be measured (2); If quantized level is the N position, quantized interval is a Δ, and then the maximum signal amplitude value of Liang Huaing is 2 ^NThe quantification snr computation that Δ, quantizing process are brought into is as follows:

$S/N=20\lg \frac{2^N\mathrm{\Δ}}{\sqrt{{\mathrm{\Δ}}^2/12}}=(6.02N+10.8)\mathrm{dB}.$

Images