CN109246406B - A Realization Method of Automatic White Balance of Color Camera - Google Patents

Abstract

The invention relates to a method for realizing automatic white balance of a color camera. This method improves the convergence speed of the automatic white balance algorithm of the color camera. The main steps of the method are: 1. According to the RGB and YCbCr space conversion relationship, obtain the current gain and the target gain relational expression; 2. The current frame image detects the white point detection; 3. Estimates the color temperature of the current frame image N, and judges whether Color cast occurs; 4. Calculate the target gain according to the relationship between the current gain and the target gain in step 1; 5. Adjust the gain of the R and B components of each pixel of the N+2 frame image.

Description

A Realization Method of Automatic White Balance of Color Camera

technical field

The invention relates to an image processing technology, in particular to a method for realizing automatic white balance of a color camera.

Background technique

The basic concept of white balance is to restore white objects to white regardless of the light source. The spectral characteristics are different under different light sources or weather conditions, which is called color temperature. In a physical sense, the reason why people see objects in a specific color is because when the light hits the object, the light reflected by the object presents this specific color in the human visual system. In other words, the color of an object that can be seen depends not only on its reflective properties, but also on the light that shines on the object. For example, a piece of white paper will be orange-yellow under tungsten light, and greenish under fluorescent light. And in daylight it's a bit bluish. The purpose of automatic white balance is to automatically adjust the ratio of the three primary color components of the image under different color temperature lighting, so that it looks consistent with the color taken under the standard light source.

Many color cameras with automatic white balance function evaluate the color temperature of images in the YCbCr space, and then adjust the gain values of Cb and Cr to make Cb and Cr meet the conditions of white pixels, thereby completing the automatic white balance function. Therefore, Cb, The adjustment method of Cr gain is an important point of automatic white balance.

The existing method is to adjust the gain step by step by setting the step size, but when this method is used for adjustment, the convergence speed is slow and cannot meet the requirements of high-frequency cameras.

Contents of the invention

In order to solve the defects of the prior art in the background technology, the present invention provides a method for realizing the automatic white balance of the color camera, which improves the fast convergence speed of the automatic white balance of the color camera.

Concrete technical scheme of the present invention is:

A kind of realization method of color camera automatic white balance, runs on the processor of color camera, and described processor type is FPGA or CPU or DSP; Realize by following steps:

[1] According to the spatial conversion relationship between RGB and YCbCr, obtain the relational expression of the current gain R component gain γ and the target gain R component gain γ′, and the relational expression of the current gain B component gain α and the target gain B component gain α′;

The specific formula is:

[2] Set the number threshold w of white dots, detect each pixel of the current frame image N, judge whether it is a white dot, and count the number w' of white dots in the current frame image N, and use the detected Compare the number w' of white dots in the current frame image N with the threshold w of white dots, if w'≥w, proceed to step [3], if w'<w, proceed to step [6];

[3] Estimate the color temperature of the current frame image N to determine whether there is a color cast;

Find the average of all white points in the current frame image N and

if If M is set to 9, it is judged that there is a color cast in the image, and the B or R component needs to be adjusted for gain, and then go to step [4];

if Then judge that there is no color cast in the image, and go to step [6];

【4】when When , the target gain R component gain γ' is calculated according to the formula (1); when , calculate the target gain B component gain α' according to the formula (2);

[5] Use the target gain R component gain γ' or the target gain B component gain α' obtained in step [4] to adjust the white balance of the current frame image.

[6] After waiting for the transmission of 1 frame image, repeat steps [2]-[5] from the N+2 frame image.

Further, the judgment condition that any pixel in the current frame image N in step [3] is a white point is: Y≥A, |Cb|≤B, |Cr|≤C; wherein, A=2048, B=10, C=20. Since the quantization bit of the camera implementing this method is 12bit, and the gray value range is 0-4095, the value of A is set to half of the gray value range in this method, so A is set to 2048.

Further, the threshold of the number of white points must be large enough to ensure the effect and correctness of the white balance, here w is set to 1/500 of the total number of pixels;

The beneficial effects of the present invention are:

The present invention adopts the method of directly calculating the gain to quickly converge the automatic white balance, and the convergence can be realized after two or several frame periods. For cameras with a frame rate of tens of frames per second or hundreds of frames per second, the Quickly achieve white balance, greatly improving work efficiency.

In general white balance methods, when adjusting the gain, a larger step size is set in order to ensure the convergence speed, which may lead to failure to converge. The method of directly calculating the gain adopted by the present invention can realize that the step size of two adjacent gain adjustments is very small, thereby ensuring the accuracy of the white balance method.

Description of drawings

Fig. 1 is a schematic diagram of the workflow of the present invention;

Fig. 2 is the linear relationship diagram of Cb and gain α;

Fig. 3 is a linear relation diagram of Cr and gain γ.

Detailed ways

The present invention provides a method of directly calculating the gain, which quickly and efficiently realizes the automatic white balance of the color camera, and its specific implementation steps are as follows:

[1] Obtain the relational expression of the current gain R component gain γ and the target gain R component gain γ′, and the relational expression of the current gain B component gain α and the target gain B component gain α′; the specific implementation steps of this step are as follows:

[1.1] Determine the formula for converting RGB space to YCbCr space;

Y＝a*γ*R+b*G+c*α*B (1)

Cb=α*B-Y=-a*γ*R-b*G+(1-c)*α*B (2)

Cr=γ*R-Y=(1-a)*γ*R-b*G-c*α*B (3)

Among them, Y is the brightness component of the pixel in the YCbCr space; Cb is the blue offset component of the pixel in the YCbCr space; Cr is the red offset component of the pixel in the YCbCr space; R is the red component of the pixel in the RGB space , G is the green component of the pixel in the RGB space, B is the blue component of the pixel in the RGB space, a is the coefficient of the R component in the space conversion formula, b is the coefficient of the G component in the space conversion formula, and c is the space conversion The coefficient of the B component in the formula, γ is the gain of the R component in the automatic white balance, and α is the gain of the B component in the automatic white balance.

According to formulas (1), (2) and (3), it can be seen that when adjusting the gain of γ, α is used as a constant; when adjusting the gain of α, γ is used as a constant. In this way, γ has a linear relationship with Cr, and α has a linear relationship with Cb. The specific linear relationship is shown in Figure 2 and Figure 3 .

【1.2】According to the formulas (1), (2) and (3), we can get:

Y'=a*γ'*R+b*G+c*α'*B (4)

Cb'=α'*B-Y'=-a*γ'*R-b*G+(1-c)*α'*B (5)

Cr'=γ'*R-Y'=(1-a)*γ'*R-b*G-c*α'*B (6)

Among them, Cr' is the red offset component when the automatic white balance method converges, that is, the target value, which is a constant; Cb' is the blue offset component when the automatic white balance method converges, that is, the target value, which is a constant; γ' is The gain of the R component when the automatic white balance method converges, and α' is the gain of the B component when the automatic white balance method converges.

From formula (5) and formula (6), we can get

It can be seen from this that, through the current gains γ and α, the target gains γ′ and α′ can be calculated.

[2] Set the number threshold w of white dots, detect each pixel of the current frame image N, judge whether it is a white dot, and count the number w' of white dots in the current frame image N, and use the detected Compare the number w' of white dots in the current frame image N with the threshold w of white dots, if w'≥w, go to step [3], if w'<w, go to step [6]; this In the embodiment, w is 1/500 of the total number of detector pixels. ;

The judgment condition that any pixel in the current frame image N is a white point is: Y≥A, |Cb|≤B, |Cr|≤C; wherein, A=2048, B=10, C=20. Since the quantization bit of the camera implementing this method is 12bit, and the gray value range is 0-4095, the value of A is set to half of the gray value range in this method, so A is set to 2048.

[3] Estimate the color temperature of the current frame image N to determine whether there is a color cast;

Find the average of all white points in the current frame image N and

if If M is set to 9, it is judged that there is a color cast in the image, and the B or R component needs to be adjusted for gain, and then go to step 4;

if Then it is judged that there is no color cast in the image, and go to step 6;

【4】when When , the target gain R component gain γ' is calculated according to the formula (1); when , calculate the target gain B component gain α' according to the formula (2);

[5] Use the target gain R component gain γ' or the target gain B component gain α' obtained in step [4] to adjust the white balance of the current frame image.

[6] After waiting for the transmission of 1 frame image, repeat steps [2]-[5] from the N+2 frame image. The reason for this is that after calculating the new gains γ' and α' based on the Nth frame image, if the frame rate is fast enough, the N+1th frame image has already been output, and the new gain has no effect. The new gain will be applied to the N+2th frame image. So in order to ensure the correctness of the automatic white balance, this method is adopted.

Since many parameters such as w, A, B, and C in this invention are self-set, if the set parameters are inappropriate, the automatic white balance will also fail to converge. At this time, the automatic white balance can be set The number of adjustments T. If this threshold is exceeded, the automatic white balance will temporarily stop, and after a period of time, the gain will return to the default value, and the calculation will start again.

Claims (2)

1. A method for realizing automatic white balance of a color camera is characterized by comprising the following steps:

【1】 Acquiring a relational expression of a current gain R component gain gamma and a target gain R component gain gamma 'and a relational expression of a current gain B component gain alpha and a target gain B component gain alpha' according to the space conversion relation between RGB and YCbCr;

the concrete formula is as follows:

cb is a blue offset component of a pixel point in a YCbCr space; cr is a red offset component of the pixel point in a YCbCr space; g is the green component of the pixel point in the RGB space, B is the blue component of the pixel point in the RGB space, a is the coefficient of the R component in the space conversion formula, B is the coefficient of the G component in the space conversion formula; cr' is a red offset component when the automatic white balance method converges, namely a target value, and is a constant; cb' is a constant, which is the blue shift component when the automatic white balance method converges, i.e., the target value;

【2】 Setting a white point number threshold value w, detecting each pixel of the current frame image N, judging whether the pixel is a white point, counting the number w 'of the detected white points of the current frame image N, comparing the detected white point number w' of the current frame image N with the white point number threshold value w, if w 'is more than or equal to w, performing the step (3), and if w' is less than w, performing the step (6);

【3】 Estimating the color temperature of the current frame image N, and judging whether color cast occurs or not;

the judgment condition that any pixel in the current frame image N is a white point is as follows: y is more than or equal to A, | Cb | is less than or equal to B, | Cr | is less than or equal to C; wherein, A is 2048, B is 10, and C is 20;

calculating the average value of all white points of the current frame image NAnd

if it is notIf M is set to be 9, judging that the image has color cast, and adjusting the gain of the component B or R, and entering the step (4);

if it is notJudging that the image has no color cast, and entering the step (6);

【4】 When in useCalculating a target gain R component gain gamma' according to a formula (1); when in useCalculating a target gain B component gain alpha' according to a formula (2);

【5】 Performing white balance adjustment on the current frame image through the target gain R component gain gamma 'or the target gain B component gain alpha' obtained in the step (4);

【6】 After waiting for the 1 frame image transmission, the steps [ 2 ] to [ 5 ] are repeated from the N +2 frame image.

2. The method for realizing automatic white balance of a color camera according to claim 1, wherein: w is 1/500 of the total number of detector pixels.