KR20110131612A - Auto white balance setting method by white detection considering sensor characteristic - Google Patents

Abstract

PURPOSE: An automatic white balance setting method is provided to apply an image to each pixel by calculating gain by the RGB average of the detected image pixels. CONSTITUTION: A pixel by primary white detection is detected to a final white in case the value of r,b variation is less than a reference value(S4, S5). In case the value of the dispersion is bigger than a standard value, second white is detected from the primarily detected white(S6). The RGB average is calculated from finally detected white pixels. The automatic white balance setting is performed in the use of the calculated gain(S8).

Description

Automatic white balance setting method by white detection considering sensor characteristic}

The present invention relates to, for example, an automatic white balance setting method of a mobile phone camera, and more particularly, by detecting white in consideration of sensor characteristics, the color according to a light source for an image including white in the image. The present invention relates to an automatic white balance setting method of a mobile phone camera for minimizing distortion.

In general, white objects appear in different colors depending on the light source (sunlight, fluorescent light, incandescent lamp, etc.). White balance means that white objects appearing in different colors according to these light sources are adjusted to appear as the original white color. It means work.

In addition, in the case of a person, even if the lighting or the place is different, the color of the object does not feel different, but since the camera is not compatible with the color unlike the human eye, the color of the same subject may be different depending on the light and the place. Will be represented.

Therefore, in the camera, a function for controlling a phenomenon in which the color of the object changes according to the light source is necessary as described above, and automatic white balance is a function in which the camera automatically determines and controls the above phenomenon. In recent years, the camera is a feature that is essential to almost all.

Conventional techniques for such automatic white balance can be classified according to whether the light source is estimated as follows.

That is, first, a method of estimating a light source and performing automatic white balance includes finding a white region of an input image and using each channel by using an average of the corresponding RGB values of the image information or an average of the color difference signals RY (BY). There is a way to adjust the gain.

As a method of not performing light source estimation, there is a method of adjusting the gain of each channel by using an average value for each channel of the entire image.

The above method is based on the assumption that if the input image has a sufficient color change, the average color of each channel is gray.

That is, if the RGB gain of the image is adjusted so that the average value of each channel is the same, the principle of reproducing the white balance of the image is used.

However, among the above-described methods, the automatic white balance method based on the light source estimation uses a simple white detection method without taking into account the characteristics of the sensor at the time of imaging, thereby causing a problem of poor white balance performance.

Therefore, in order to solve the above problems and improve the performance of the white balance, it is desirable to consider the sensor characteristics in the automatic white balance method by the light source estimation, but the automatic white balance technology that satisfies such requirements is still It is not provided.

The present invention has been made to solve the problems of the conventional automatic white balance setting method as described above. Accordingly, an object of the present invention is to provide an automatic white balance of a mobile phone camera, for an image including white in the image. An object of the present invention is to provide an automatic white balance setting method for minimizing color distortion caused by a light source.

More specifically, the present invention, in the automatic white balance of the mobile phone camera, in consideration of the sensor characteristics, by calculating the gain from the R, G, B average of the last detected white pixels and applying to each pixel of the image, The purpose is to provide an automatic white balance setting method of the mobile phone camera so that there is no difference in the color of the object even if this is changed.

In order to achieve the above object, according to the present invention, in order to minimize the distortion of the color according to the light source for the image including the white in the image to automatically set the white balance by the white detection in consideration of the sensor characteristics A white balance setting method, comprising: a conversion step of converting R, G, and B values of each pixel of an input image into r, g, and b, which are standardized R, G, and B, and the input from the r, g, and b values A primary detection step of detecting white in the image as a primary, a calculation step of calculating an average and a variance of the standardized R, G, and B of the white pixels detected in the primary detection step, and a calculation step a discriminating step of determining whether the variance of r and b is smaller or larger than a predetermined reference value; and if the variance of r and b is smaller than the reference value in the determination result of the determining step, A white detection step of detecting the pixels by the primary white detection as the final white by judging that they have a similar color temperature, or, if the value of the variance is greater than the reference value in the determination result of the determination step, detection And a second detection step of detecting the second white again from the first detected white by determining that the difference in the color temperature of the pixels is large, and the white pixels finally detected from the white detection step or the second detection step. And a white balance execution step of calculating an R, G, and B averages and gains from the control unit; and a white balance execution step of performing automatic white balance setting using the gain calculated in the gain calculating step. A balance setting method is provided.

Here, the conversion step is characterized in that performed using the following Equation 1.

[ Equation  One]

r = R / (R + G + B)

b = G / (R + G + B)

g = B / (R + G + B)

In the primary detection step, white is detected using standardized R, G, B, and luminance of pixels of the input image, and at this time, white is detected using a detection equation according to the sensor. do.

Here, the sensor is a CMOS sensor, the detection equation, tol is a value representing the tolerance of the y-intercept, Y represents the luminance of one pixel, Y 'is a luminance threshold derived from the luminance histogram In this case, it is characterized by detecting the white color through a detection formula such as the following [Equation 2].

[ Equation  2]

0.235 <r <0.485

-0.7009r + 0.4731-tol <b <-0.7009r + 0.4731 + tol

Y '<Y

The luminance threshold may be a luminance level when the luminance threshold value has a value greater than the frequency of the average luminance at a level greater than the average luminance of the input image.

In addition, when the luminance histogram of the input image does not form a double peak, the average luminance is used as the luminance threshold.

In addition, the secondary detection step has a color temperature larger than the average of normalized R, G, and B values of white detected in the first detection step by using a formula as shown in Equation 3 below. It is characterized in that it is configured to exclude only the red-based white by detecting only the white of the color temperature higher than the average (rm, bm) of, r and b.

[ Equation  3]

r <rm

b> bm

In the gain calculating step, R, G, and B gains are calculated for each channel from the R, G, and B averages of the finally detected pixels of white using a formula as shown in Equation 4 below. And configured to.

[ Equation  4]

는 각각 상기 최종적으로 검출된 화이트 화소들의 R, G, B 평균을 나타낸다.)
(here,

Respectively represent R, G, and B averages of the finally detected white pixels.)

In addition, the white balance execution step, using the gain for each channel of the R, G, B calculated by the gain calculation step, automatically white balance based on the equation as shown in Equation 5 below It is characterized in that it is configured to perform a process for setting.

[ Equation  5]

In addition, the reference value is configured to be set to an arbitrary value as needed, and may have the same value for the variance of the r and b, or may be configured to have different values.

As described above, the present invention, in order to minimize the distortion of the color by the light source for the image including the white in the image by calculating the gain by the R, G, B average of the pixels finally detected as white angle of the image It is configured to apply to the pixel.

That is, according to the present invention, the primary white detection is performed in consideration of the sensor characteristics, the average and the variance of the normalized R, G, B of the primary detected pixels are calculated, and then the standardized R, G, B If the variance satisfies the preset condition, only the pixels of the color temperature higher than the standardized R, G, and B are secondarily detected among the first detected pixels, and the variance of the standardized R, G, B does not satisfy the preset condition. If not, only the first detection is performed to calculate the average of R, G, and B from the last detected white pixels, calculate the gain for each of the R, G, and B channels from the calculated average, and calculate the image of the calculated gain. Automatic white balance is performed by multiplying each pixel R, G, and B values of.

Therefore, according to the present invention, the conventional automatic white balance method based on light source estimation solves the problem that the performance of the white balance was not good because the simple white detection method was used without considering the characteristics of the sensor when taking a picture, and the sensor characteristics were considered. By calculating the gain from the R, G, B average of the last detected white pixels and applying it to each pixel of the image, it provides an automatic white balance setting method of the mobile phone camera so that there is no difference in the color of the object even if the illumination changes By doing so, it is possible to minimize the distortion of the color according to the light source with respect to the image including white in the image, and to improve the quality of the image.

1 is a flowchart showing the overall flow of the automatic white balance setting method according to the present invention.
FIG. 2 is a diagram illustrating a relationship between standardized R, G, and B of gray patches for each light source in a Macbeth SG chart.
3 is a diagram illustrating an example of a primary white detection region.
4 is a diagram for describing a luminance threshold used when detecting white.
5 is a diagram illustrating an example of a white detection result image by primary white detection.
6 is a diagram illustrating an example of a primary white detection region.
7 is a diagram illustrating an example of a white detection result image by secondary white detection.
FIG. 8 is a diagram illustrating a detection result by primary white detection and secondary white detection of a Macbeth color checker in a light source.
9 is a diagram illustrating an example of an image of a result of performing automatic white balance.

Hereinafter, with reference to the accompanying drawings, it will be described in detail with respect to the automatic white balance setting method of the mobile phone camera according to the present invention as described above.

Here, the content described below is only one embodiment for carrying out the present invention, the present invention is not limited only to the content of the embodiment described below.

In the following description of the present invention, detailed descriptions of related well-known technologies or constructions are already well known to those skilled in the art, or are considered to unnecessarily obscure the subject matter of the present invention. The detailed description is omitted for the sake of brevity.

In addition, the terms used in the following description are terms defined in consideration of functions in the present invention, and it should be noted that this may vary depending on the user's intention or custom.

First, with reference to Figure 1 will be described a specific embodiment of the automatic white balance setting method of the mobile phone camera according to the present invention.

Referring to FIG. 1, FIG. 1 is a flowchart illustrating the overall flow of the automatic white balance setting method according to the present invention.

As shown in FIG. 1, the automatic white balance setting method according to the present invention relates to an automatic white balance technology for an image including white. As such, the automatic white balance technology for an image including white includes: A conversion step S1 of converting the R, G, and B values of each pixel of the input image into standardized R, G, and B r, g, and b; And a calculation step S3 for calculating the average and the variance of the standardized R, G, and B pixels of the white pixels detected in the step, and determining whether the variance calculated in the calculation step S3 satisfies a predetermined condition. A white detection step S5 of finally detecting pixels by primary white detection if the value of the variance is smaller than a predetermined value in the determination step S4 and the determination result of the determination step; Of the determination step If the result of the determination indicates that the value of the variance is larger than a predetermined value, the second detection step S6 of detecting the second white color from the first detected white color, R, from the last detected white pixels from the detection steps, It is characterized in that it comprises a gain calculation step (S7) for calculating the G, B mean and gain, and a white balance execution step (S8) for performing an automatic white balance using the calculated gain.

Here, the converting step S1 of converting the R, G, and B values of the input image into the normalized R, G, and B values is performed using Equation 1 below.

[ Equation  One]

r = R / (R + G + B)

b = G / (R + G + B)

g = B / (R + G + B)

Next, FIG. 2 shows a gray patch of a Macbeth SG chart with a CMOS sensor at each light source, and then converts the gray patches of the Macbeth SG chart into standardized R, G, and B, and shows them on the standardized R, G, and B planes.

2A, 2B and 2C, when the trend line is drawn in the linear distribution of the gray patches, it can be seen that the error is the smallest in the relationship between the normalized R and the normalized B in FIG. 2C.

2D shows these results for each light source. As shown in FIG. 2D, as the color temperature of the light source increases (arrow direction), the value of normalized R decreases, and the value of normalized B increases.

3 is a diagram showing an example of a detection area for primary white detection from the characteristics as shown in FIG. 2D, and the detection at this time is performed using the following detection formula (Equation 2) (step S2).

[ Equation  2]

0.235 <r <0.485

-0.7009r + 0.4731-tol <b <-0.7009r + 0.4731 + tol

Y '<Y

[Equation 2] is a formula derived for the CMOS sensor used to derive the relationship as shown in FIG. 2, and the equation representing the parallelogram boundary in the r and b planes is used. It should be noted that each CMOS sensor may vary, and in this embodiment, in Equation 2, tol represents a tolerance of the y-intercept, and 0.03 is used.

In addition, Y represents the luminance of one pixel, and Y 'is the luminance threshold, which is derived from the luminance histogram as shown in FIG.

Referring to FIG. 4, the average of the histogram represented by the thin solid line and the histogram represented by the thick solid line in FIG. 4 is the same.

As shown in FIG. 4, when the image contains a large amount of white (white), that is, when the image includes a large number of pixels having high luminance, the luminance histogram of the image forms a bimodal shape, and the average luminance of the image is It is located in the middle of the double rod.

As described above, when white is detected using the average luminance located in the middle of the double peak as a threshold value, non-white color is detected as white, so the level having the highest frequency at the peak of the higher luminance is used as the luminance threshold value. It is desirable to.

If the luminance histogram of the image does not form a double peak, the average luminance is used as the luminance threshold.

Subsequently, Fig. 5 shows an example of the white detection result image by primary white detection obtained as described above.

Referring to FIG. 5, the reason why a red-based color is detected as white despite being a color other than white in FIG. 5 is because it is determined that the color temperature belongs to white with low color.

Therefore, in order not to detect the red series color with a low color temperature, secondary white detection is performed as follows.

That is, the second white detection calculates an average (rm, bm) and variance of r and b of the pixels detected by the primary white (step S3), and determines whether the variance of these r and b satisfies a preset condition. (Step S4), i.e., if the primary detected white pixels have a nearly similar color temperature (if the r and b dispersions are small), the pixels by primary white detection are finally detected as white (step S5).

If the determination result of step S4 does not satisfy the preset condition, that is, if the color temperature difference of the primary detected white pixels is large (r and b are large dispersion), the primary detected pixels Among them, only whites having a color temperature higher than the r and b averages are detected to exclude red series white (step S6).

Here, in the above-described step (S4), the reference value preset for comparison with the variance of r and b is arbitrarily set to an appropriate value as necessary at the level of those skilled in the art, and different reference values for r and b may be set. It can also be configured to have.

In addition, detection of white whose color temperature is higher than r and b average is performed based on the following [Equation 3].

[ Equation  3]

r <rm

b> bm

Here, the threshold values used for the r and b variances are values determined by experiments, and the detection region is as shown in FIG. 6.

As described above, the result of detecting white by performing the second detection is shown in FIG. 7.

Referring to FIG. 7, when comparing FIG. 5 with FIG. 7, it may be confirmed that white having a low color temperature is excluded.

Next, referring to FIG. 8, FIG. 8 is a diagram illustrating detection results of primary white detection and secondary white detection of a Macbeth color checker in a light source.

In FIG. 8, the white detection area is displayed in color and the undetected area is displayed in black.

In order to perform automatic white balance, R, G, and B gains must be calculated from the R, G, and B averages of the finally detected white pixels, and this process is expressed by Equation 4 below. Is performed using (step S7).

는 각각 검출된 화이트 화소들의 R, G, B 평균을 나타낸다.
Also, in [Equation 4]

Denotes R, G, and B averages of the detected white pixels, respectively.

[ Equation  4]

Subsequently, automatic white balance is performed using the R, G, and B gains calculated by Equation 4 above. At this time, automatic white balance is performed by Equation 5 below (step S8). ).

[ Equation  5]

As a result of performing the auto white balance by applying the R, G, and B gains obtained as described above, the image is shown in FIG. 9.

Referring to FIG. 9, it can be seen that the color distortion of the image of FIG. 9B is less than that of FIG. 9A and the image quality is good.

Therefore, as described above, the average and the variance of the normalized R, G, B of the primary detected pixels are calculated, and if the variance of the standardized R, G, B satisfies a preset condition, among the primary detected pixels Only pixels with a color temperature higher than the normalized R, G, and B are detected secondarily, and if the dispersion of the standardized R, G, and B does not satisfy a preset condition, only the first detection is performed, and the final detected white pixels are According to the present invention, the average of R, G, and B is calculated, and the gain for each of the R, G, and B channels is calculated from the calculated average, and the calculated gain is multiplied by each pixel R, G, and B values of the image. An automatic white balance setting method may be performed in consideration of sensor characteristics.

In addition, by performing the automatic white balance setting method considering the sensor characteristics as described above, it is possible to minimize the distortion of the color according to the light source for the image including the white in the image, thereby further improving the image quality. .

As mentioned above, although this invention was demonstrated through the Example of this invention, this invention is not limited only to the content described in the above Example, Therefore, this invention is common in the technical field to which this invention belongs. Naturally, various modifications, changes, combinations and substitutions can be made by a knowledgeable person depending on the design needs and various other factors.

Claims (10)

In the automatic white balance setting method for automatically setting the white balance by the white detection in consideration of the sensor characteristics in order to minimize the distortion of the color according to the light source for the image including the white in the image,
A conversion step of converting the R, G, and B values of each pixel of the input image into r, g, and b, which are standardized R, G, and B;
A first detection step of first detecting white in the input image from the r, g, and b values;
A calculation step of calculating an average and a variance of the normalized R, G, and B of the white pixels detected in the first detection step;
A discriminating step of determining whether the variance of r and b calculated in the calculating step is smaller or larger than a predetermined reference value;
If it is determined from the determination result of the determining step that the variance of the r and b is smaller than the reference value, it is determined that the detected pixels have a similar color temperature, and white for detecting the pixels by the first white detection as the final white. Detection step,
Alternatively, if it is determined that the value of the variance is larger than the reference value in the determination result of the determination step, it is determined that the difference in the color temperature of the detected pixels is large and the secondary white is detected again from the first detected white. Detection step,
A gain calculation step of calculating R, G, and B averages and gains from the white pixels finally detected from the white detection step or the second detection step;
And a white balance execution step of performing automatic white balance setting by using the gain calculated in the gain calculating step. The method of claim 1,
The conversion step, the automatic white balance setting method characterized in that performed using the following [Equation 1].

[ Equation 1]
r = R / (R + G + B)
b = G / (R + G + B)
g = B / (R + G + B)
The method of claim 1,
In the first detection step, white is detected using standardized R, G, B, and luminance of pixels of the input image.
At this time, the automatic white balance setting method, characterized in that for detecting the white by using the detection formula according to the sensor. The method of claim 3, wherein
The sensor is a CMOS sensor,
In the above detection formula, when tol is a value representing the tolerance of the y-intercept, Y is the luminance of one pixel, and Y 'is the luminance threshold derived from the luminance histogram. Automatic white balance setting method characterized in that the detection of the white color through a detection formula such as.

[ Equation 2]
0.235 <r <0.485
-0.7009r + 0.4731-tol <b <-0.7009r + 0.4731 + tol
Y '<Y
The method of claim 4, wherein
The luminance threshold value is an automatic white balance setting method using a luminance level when the value is greater than the frequency of the average luminance at a level greater than the average luminance of the input image among the luminance histogram of the input image. . The method of claim 4, wherein
And if the luminance histogram of the input image does not form a double peak, an average luminance is used as a luminance threshold. The method of claim 1,
The secondary detection step has a color temperature greater than the average of the normalized R, G, and B values of white detected in the first detection step, using a formula as shown in Equation 3 below. And detecting only whites having a color temperature higher than the mean (rm, bm) of b to exclude white of a red series.

[ Equation 3]
r <rm
b> bm
The method of claim 1,
The gain calculating step is configured to calculate R, G, and B gains for each channel from the R, G, and B averages of the finally detected pixels of white using a formula as shown in Equation 4 below. Auto white balance setting method, characterized in that.

[ Equation 4]

(here,

Respectively represent R, G, and B averages of the finally detected white pixels.) The method of claim 1,
In the white balance execution step, the white balance is automatically set based on the equation as shown in Equation 5 below by using the gain for each channel of R, G, and B calculated by the gain calculation step. Automatic white balance setting method, characterized in that configured to perform the processing.

[ Equation 5]

The method of claim 1,
The reference value is configured to be set to any value as needed,
The reference value may be configured to have the same value or different values for the variance of the r and b, respectively.

Images