KR20030009624A - White balance correction method by using Kalman filter - Google Patents

Abstract

PURPOSE: A method for correcting white balance of a video system using a Kalman filter is provided to improve accuracy and calculation processing capability of automatic white balancing. CONSTITUTION: The first image data is received from a selected image block(41), to measure an average value of a color difference signal(42). The average value is input to Kalman filter to obtain an offset estimation value of the color difference signal(43). The second image data different from the first image data is received from the selected image block(44), to calculate an offset measurement value of the color difference signal. The difference between the offset estimation value and the offset measurement value is compared with an allowable value. When the difference is larger than the allowable value, the step of calculating the offset measurement value and the comparison step are repeated at least once. When the difference is smaller than the allowable value, a gain is calculated from the offset measurement value.

Description

White balance correction method by using Kalman filter

The present invention relates to a white balance correction method of an image device, and more particularly, to a white balance correction method using a Kalman filter.

The conventional white balance correction method does not use the correlation between subsequent frames, or otherwise does not use systematic concepts such as correcting the white balance every frame. Therefore, since the average of the color difference signals Cb and Cr is applied to the entire image, a large amount of computation is required. In order to solve this problem, there is a method of correcting the white balance by extracting a specific part of the image, but this method has a disadvantage in that an incorrect result is obtained.

SUMMARY OF THE INVENTION An object of the present invention is to provide a white balance correction method of an image apparatus using a Kalman filter which can more accurately correct white balance and not perform white balance correction every frame.

1 is a schematic view showing a method of correcting a white balance using a Kalman filter according to the present invention;

2 is a graph showing a difference between a color difference signal offset prediction value and a measured value in a two-dimensional plane formed by a first color difference signal Cb axis and a second color difference signal Cr axis;

3 is an explanatory diagram showing an image sampling area;

Figure 4 is a flow chart showing a white balance correction method according to the present invention.

The present invention for achieving the above object is a first step of obtaining the average value of the color difference signal by receiving the first image data from the plurality of image block selected image block; Obtaining an offset prediction value of the color difference signal by inputting the average value to a Kalman filter; Obtaining an offset measurement value of a color difference signal by receiving second image data different from the first image data from a selected block; A fourth step of comparing a difference between the offset prediction value obtained from the Kalman filter and the offset measurement value with an allowable value; A fifth step of repeating the third and fourth steps at least once when the difference between the offset prediction value and the offset measurement value is larger than an allowable value; And calculating a gain from the offset measurement value when the difference between the offset prediction value and the offset measurement value is smaller than an allowable value.

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for correcting white balance using a Kalman Filter, wherein an offset of a color difference signal (Cb, Cr) of consecutive frames in an image device such as a digital camera or a camcorder is used. It is characterized by using the prediction.

The Kalman filter is designed to estimate the state of the actual vehicle even if there is an error in modeling or sensor noise in consideration of noise caused by modeling or sensor noise caused by a detection device.

According to the present invention, since the offset of the color difference signals Cb and Cr is continuously tracked using the Kalman filter, the offset of the color difference signals Cb and Cr of the next frame can be predicted, thereby reducing the calculation amount. On the other hand, the accuracy of the Kalman filter's next prediction depends on the accuracy of the current measurement. In the present invention, the accuracy of the measured value is further improved by determining the inaccuracy based on the difference between the predicted value and the measured value.

Hereinafter, a white balance correction method according to an embodiment of the present invention will be described with reference to FIGS. 1 to 4.

As shown in FIG. 1, the present invention inputs prediction values and image data measurements obtained from a Kalman filter into a sampler, and calculates offsets of each of the first color difference signal Cb and the second color difference signal Cr from the sampler. ), It calculates the blue (B) gain and then generates the sensor gain control input signal.

As shown in FIG. 2, the first color difference signal Cb and the second color difference signal Cr are identified by checking whether the difference L between the predicted value obtained from the Kalman filter and the measured value of the image data is larger than the allowable values Cr_max and Cb_max. Find each offset.

That is, as shown in FIG. 3, the present invention divides the group into a first image block 1 selected primarily from among a plurality of image blocks, a second image block selected secondly from the second image, and the like, and the first image data from the selected blocks. Is obtained, and the average value is input to the Kalman filter to obtain an offset prediction value. Second image data different from the first image data is input from the selected block, and the second image data is averaged to obtain a first color difference signal ( Measurement values of the offset of Cb) and the offset of the second color difference signal Cr are obtained. Next, when the difference between the offset prediction value obtained from the Kalman filter and the offset measurement value is smaller than the allowable value, the offset measurement value is determined from the offset values of the first color difference signal and the second color difference signal, respectively, in red (R), green (G), and blue (B). Calculate the gain. If the difference between the offset prediction value and the offset measurement value obtained from the Kalman filter is larger than the allowable value, the offset measurement value is obtained by selecting another image block among the image blocks not selected at the time of the first selection, as described above, and from the Kalman filter. The difference between the offset estimate obtained and the offset measurement is compared with the tolerance. The process of selecting an image block, calculating an offset measurement, and comparing the size with the tolerance are repeated until the difference between the offset prediction value and the offset measurement becomes smaller than the allowable value.

This will be described in more detail with reference to FIG. 4.

First, the first image data of the first image block 1 selected first among the plurality of image blocks is received 41 and each of the first color difference signal Cb and the second color difference signal Cr of the first image block is received. Average values Cb_1 [n-1] and Cr_1 [n-1] are calculated (42). Each of these average values is input to a Kalman filter to obtain offset prediction values Cb '[n-1] and Cr' [n-1] of the first color difference signal Cb and the second color difference signal Cr, respectively (44). ).

Subsequently, the second image data of the first image block 1 is input 44 and the average values Cb_1 [n] and Cr_1 of each of the first color difference signal Cb and the second color difference signal Cr of the first image block 1 are received. [n]) is calculated (45).

Next, the measured values Cb_1 [n] and Cr_1 [n] and the offset predicted values of the first color difference signal Cb and the second color difference signal Cr obtained from the second image data of the first image block 1, respectively. The difference (Dr_1, Db_1) of each of Cb '[n-1] and Cr' [n-1]) is obtained (46).

The difference between the measured values (Cb_1 [n], Cr_1 [n]) and the offset prediction values (Cb '[n-1], Cr' [n-1]) (Dr_1, Db_1) obtained according to the above procedure is allowed. Cr_max). If smaller than the allowable value, the red (R), green (G), and blue (B) gains are calculated (51). If large, the measured values Cb_2 [n] and Cr_2 [n] are calculated from an average value of each of the first color difference signal Cb and the second color difference signal Cr for a second block different from the first block (48). The measured values Cb_2 [n] and Cr_2 [n] and the offset predicted values Cb '[n-1, respectively, of the first color difference signal Cb and the second color difference signal Cr obtained from the second image block 2. ], And the differences (Dr_2 and Db_2) of Cr '[n-1]) are obtained (49). The difference between the measured values (Cb_2 [n], Cr_2 [n]) and the offset predicted values (Cb '[n-1], Cr' [n-1]) obtained by the above procedure (Dr_2, Db_2) is allowed to be allowed (Cr_max, Cr_max). If smaller than the allowable value, the offset measurement (Cb_2 [n], Cr_2 [n]) calculates the gains of red (R), green (G) and blue (B) (51), and if larger, the first block and the second. The above-described process is repeated for the third block different from the block.

The present invention described above is not limited to the above-described embodiments and the accompanying drawings, and various substitutions, modifications, and changes are possible in the art without departing from the technical spirit of the present invention. It will be apparent to those of ordinary knowledge.

The present invention as described above can improve the accuracy and the calculation processing ability in the implementation of the auto white balance (auto white balance) which is an essential function in digital cameras and digital camcorders. By tracking the offset of Cr) using a Kalman filter, the offset of the color difference signals Cb and Cr of the next frame can be predicted, thereby reducing the amount of computation.

Claims (1)

In the white balance correction method of the image device, A first step of receiving first image data from a plurality of image blocks selected image blocks to obtain an average value of a color difference signal; Obtaining an offset prediction value of the color difference signal by inputting the average value to a Kalman filter; Obtaining an offset measurement value of a color difference signal by receiving second image data different from the first image data from a selected block; A fourth step of comparing a difference between the offset prediction value obtained from the Kalman filter and the offset measurement value with an allowable value; A fifth step of repeating the third and fourth steps at least once when the difference between the offset prediction value and the offset measurement value is larger than an allowable value; And A sixth step of calculating a gain from the offset measurement when the difference between the offset prediction value and the offset measurement is smaller than an allowable value White balance correction method of the image device comprising a.