KR101401683B1 - Apparatus for detecting focus robust to low light noise and spot highlight - Google Patents

Abstract

The present invention relates to a digital image processing apparatus and method, and more particularly, to a focus detection apparatus capable of removing a low-illuminance noise and a spot highlight image to obtain a clear image. A focus detection apparatus robust against low-noise noise is a digital image processing apparatus comprising: an average calculation unit for calculating an average of a predetermined block including a pixel currently being processed from the input image, using an image signal including noise as an input image; A difference image calculation unit for calculating the difference image of the input image signal and the average calculated by the average calculation unit; A variance calculation unit for calculating a variance value of a video signal and noise from the input video signal, the calculated average, and the calculated difference image; And a focus indexing calculator for calculating a focus index by applying a weight calculated by the calculated variance value to the difference image.

Local average, local variance, focusing index, spot highlight

Description

BACKGROUND OF THE INVENTION Field of the Invention [0001] The present invention relates to a focus detection apparatus for detecting a low-

The present invention relates to a digital image processing apparatus and method, and more particularly, to a focus detection apparatus capable of removing a low-illuminance noise and a spot highlight image to obtain a clear image.

When the focus is matched, the image is captured clearly compared with the case where the focus is not coincident. This means that the high frequency component is maximized. Therefore, generally, the focus detection is performed by calculating the high frequency component of the image. However, when noise is added to the input image, image degradation due to noise occurs, which occurs at the same time in low frequency and high frequency. That is, in the case of the conventional automatic focusing method based on high frequency component detection, there is a problem that focus is misdetected due to the focus indexing together with the noise along with the high frequency component. Also, the position and bandwidth of the band required for focusing detection are changed according to the characteristics of the added noise. As the amount of noise increases, the required band is decreased and the bandwidth is decreased. In order to design an automatic focusing algorithm that reflects these characteristics, it is necessary to analyze characteristics of noise and design a filter according to the characteristics. Most conventional techniques do not reflect the influence of such noise, and even in the case of a technique considering noise image, it is not adaptable to the characteristics of noise.

Filter based auto focusing uses a filter that can detect the high frequency component of the image acquired through the sensor and uses the method of determining focus indexing based on the filter. Generally, there are a Tenengrad method that uses a high-frequency bandpass filter, a gradient computation method of each pixel, a sum modified difference (SMD) method that uses a first derivative, and a sum modified laplacian (SML) method that uses a second derivative .

The problem with filter-based auto-focusing algorithms is that additional algorithms are needed to efficiently remove the effects of noise. In general, in order to remove the influence of noise, noise elimination is preliminarily performed, or each band-pass filter is designed so that a large number of filters are held as filter banks. Then, the SNR of the image is measured and a filter . To overcome these disadvantages, a Robust Focus Measuring Operator has been proposed which removes the influence of noise by using one filter. This focuses on the fact that noise has no correlation between adjacent pixels and that the image has a correlation and calculates the focus indexing by obtaining a difference by a distance k as shown in Equation 1, Method.

Although the influence of noise is reduced as the value of k increases in Equation (1), the focusing detection capability is also reduced when there is no noise, and k is selected by estimating the SNR.

In the case of WT (wavelet transform) or DCT (discrete cosine transform), there is a Transform-based automatic focusing method for calculating the focus indexing by selecting a required frequency band based on the reconstructed image for each frequency band. However, such a transform-based automatic focusing method also has a disadvantage in that it is necessary to first determine the parameters according to the characteristics and the amount of noise by estimating the SNR.

As described above, in the conventional automatic focusing algorithm, a cascade method of detecting a focus after removing noise of an input image requires a preprocessing process such as setting an SNR of an input image, and a focus detection failure due to noise is a problem .

SUMMARY OF THE INVENTION It is an object of the present invention to provide a focus detection apparatus which is robust against low-noise noises and spot highlights, which prevents a focus detection failure due to noise and improves the problem of focusing at a point where a spot highlight occurs in an image.

According to a first aspect of the present invention, there is provided a focus detection apparatus robust to low-noise noises in a digital image processing apparatus, comprising: an image processing unit for processing an image signal including noise into an input image, An average calculating unit for calculating an average of a predetermined block including a pixel in the current block; A difference image calculation unit for calculating the difference image of the input image signal and the average calculated by the average calculation unit; A variance calculation unit for calculating a variance value of a video signal and noise from the input video signal, the calculated average, and the calculated difference image; And a focus indexing unit for calculating a focus index by applying a weight calculated by the calculated variance value to the difference image.

In the present invention, when the number of video signals is larger than the noise of the input video signal, the weight may be close to 1.

In the present invention, when the noise of the input image signal is higher than that of the image signal, the weight may be close to zero.

According to a second aspect of the present invention, there is provided a digital image processing apparatus, which is robust against a low-illuminated noise and a spot highlight, according to a second embodiment of the present invention. The digital image processing apparatus includes a noise- An average calculating unit for calculating an average of a predetermined block including pixels currently being processed from the input image; A difference image calculation unit for calculating the difference image of the input image signal and the average calculated by the average calculation unit; A variance calculation unit for calculating a variance value of a video signal and noise from the input video signal, the calculated average, and the calculated difference image; A brightness determining unit determining a brightness of the input image that is greater than or less than a specific brightness value; And a focus indexing unit for calculating a focus index by applying a weight calculated by the calculated variance value to the input image according to the brightness determination result.

In the present invention, the focus indexing calculation unit may include a first focus index calculation unit for calculating a first focus index by applying a weight calculated by the calculated variance value to the input image having a brightness value as a result of the brightness determination; A second focus indexing calculation unit for calculating a second focus indexing by applying a weight calculated by the calculated variance value to the input image having a brightness value greater than or equal to a specific brightness value as a result of the brightness determination; And outputting first and second focus indexes when an image having a brightness value equal to or greater than a predetermined brightness value is greater than or equal to a predetermined portion of the input image, And outputting the first focus indexing if the first focus indexing is not performed.

In the present invention, the first and second focus indexing calculators may apply the weighting factor of 1 when there are more image signals than the noise of the input image signal.

In the present invention, the first and second focus indexing calculators may apply the weight to zero if the noise is larger than the image signal of the input image signal.

As described above, according to the present invention, it is possible to prevent a focus detection failure due to noise and to fix a problem of focusing at a point where a spot highlight occurs in an image.

In addition, since it has a simple structure, it can be effectively mounted on a product using a low processor, and it is possible to effectively eliminate noise by using an additional weight on other filters, and to use only local average and local dispersion Therefore, it is easy to improve the performance when changing the characteristics of the product.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. The description of the drawings and the drawings is for illustrative purposes only and is not intended to limit the scope of the invention.

FIG. 1 is a block diagram showing a configuration of a focus detection apparatus robust against low-illuminance noise according to the present invention. The focus detection apparatus includes a local average calculation unit 110, a difference image calculation unit 120, a local variance calculation unit 130, (140).

와 노이즈(n)가 첨가된 영상을

라고 할 때, 노이즈가 첨가된 입력 영상의 국부 평균 영상을

라고 가정한다.The local average calculation unit 110 calculates an average of localized averages of pixels included in a pixel currently processed from the input image, i.e., a local average, with the input image (y = x + n). The image before noise (n) is added

And noise (n)

, The local average image of the input image to which the noise is added

.

와 국부 평균 산출부(110)에서 산출된 평균

의 차영상

을 산출한다. The difference image calculating unit 120 calculates a difference image

And the average value calculated by the local average calculation unit 110

Car image

.

, 산출된 평균

및 산출된 차영상

으로부터 영상신호

및 노이즈 n에 대해 현재 처리 중인 픽셀이 포함된 소정 블록의 각 분산값

및

을 산출한다.The local variance calculating section 130 calculates a local variance

, The calculated average

And the calculated car image

Lt; / RTI &

And a variance value of a predetermined block including a pixel currently being processed for noise n

And

.

에 대해 산출된 분산값

및

으로 계산된 가중치를 적용하여 포커스 인덱싱을 산출한다.The focus indexing calculation unit 140 calculates a focus index

≪ / RTI >

And

To calculate the focus indexing.

와 노이즈(n)가 첨가된 입력 영상

에 대해, 포커스 인덱싱을 산출하는 수학식은 수학식 2와 같이 계산될 수 있다.The image before noise (n) is added

And the input image with noise (n)

, The equation for calculating the focus indexing can be calculated as shown in Equation (2).

는 노이즈(n)가 첨가된 입력 영상이 국부 평균 영상으로, 차영상

는 입력 영상의 고주파 성분을 의미하며, 이는 고주파 성분을 추출하는 다른 필터로 대체될 수 있다. 영상의 각 픽셀에서 고주파 성분의 반영을 결정하게 되는 가중치는 수학식 3과 같다.here,

The input image to which the noise (n) is added is a local average image,

Means a high frequency component of the input image, which can be replaced with another filter for extracting a high frequency component. The weight that determines the reflection of the high-frequency component in each pixel of the image is expressed by Equation (3).

는 현재 처리 중인 픽셀에서의 영상 성분과 노이즈 성분의 국부 분산을 이용하여, 처리 중인 픽셀에서 노이즈 성분 보다 영상 성분이 주효한 경우에는 수학식 4와 같이 영상의 고주파 성분을 효율적으로 추출하게 된다.This weight

Using the local variance of the image component and the noise component in the currently processed pixel, when the image component is more effective than the noise component in the pixel under processing, the high frequency component of the image is extracted as shown in Equation (4).

On the contrary, when the noise component is more effective than the image component, the focus indexing in the pixel is 0 as shown in Equation (5), thereby effectively blocking the influence of the noise.

및

을 사용해야 하며, 이를 위해 노이즈가 첨가되지 않은 영상 성분과 노이즈는 다음 수학식 6과 같은 근사화를 이용하게 된다.At this time, the local dispersion of the image component and the noise component

And

For this purpose, the image components and noise having no added noise are approximated by the following Equation (6).

If the approximation as shown in Equation (6) is used, there is a disadvantage that the image component and the high-frequency component are partially damaged and the edge component of the image is partially included in the noise component. However, The image component and the noise component can be estimated by a simple method using only the local average of the image. As shown in FIG. 2A, it can be confirmed that the image in which the noise is estimated by the method with 10 dB AWGN (adaptive white Gaussian nose) is included and some edge information is included. However, as shown in the histogram of FIG. , And the distribution of Gaussian noise is relatively well represented. The prior art and the present invention are visualized for 10 dB AWGN added images, as shown in FIGS. 3A and 3B, respectively. In FIG. 3A, the noise is detected together with the high frequency component of the focus detection result image. In FIG. 3B, it is confirmed that the influence of the noise is minimized and the high frequency component is detected.

FIG. 4 is a block diagram showing the configuration of a focus detection apparatus robust to low-light noise and spot highlight according to the present invention. The local-average calculation unit 410, the difference-image calculation unit 420, the local variance calculation unit 430, A determination unit 440, and a focus indexing calculation unit 450. In the present invention, the focus indexing calculation unit 450 includes a first focus indexing calculation unit 451, a second focus indexing calculation unit 452, and a focus indexing output unit 453.

The focus detection apparatus shown in FIG. 4 solves the problem that when a strong light source is present in the focusing window, the high frequency component of the light source overwhelms the high frequency component of the other portion, .

Conventionally, a method of not performing focus indexing calculation by clipping more than a specific value for all pixels in an image is used. This is a method which can effectively remove a strong light source (hereinafter referred to as spot highlight), but there is a problem that focus detection is failed by performing an exception process for a general highlight area instead of a spot highlight. Therefore, the focus detection apparatus shown in FIG. 4 prevents focus detection failure due to such spot highlighting, and even if the majority of the image corresponds to the clipping region, the focus detection apparatus shown in FIG. 4 can highlight / And calculates the focus indexing by dividing the dark region and the general region.

The local average calculator 410, the difference image calculator 420, and the local variance calculator 430 of FIG. 4 are the same as those of FIG. 1, and thus description thereof is omitted.

의 밝기 중 가장 밝은 영역 즉 특정 밝기값 이상(T_h)인 부분과 가장 어두운 영역 즉, 특정 밝기값 이하(T_l)인 부분을 판단한다. 여기서 특정 밝기값 이상(T_h)의 기준값은 그 밝기값이 예를 들어 250 이상인 경우이고, 특정 밝기값 이하(T_l)의 기준은 그 밝기값이 예를 들어 5 이하인 경우이다.The brightness determination unit 440 determines brightness

(T _h ) and the darkest region, that is, a portion that is equal to or less than a specific brightness value (T _l ) is determined. Here, the reference value of a specific brightness value (T _h ) is a brightness value of, for example, 250 or more, and a reference of a specific brightness value or less (T ₁ ) is a brightness value of, for example, 5 or less.

에 대해 상기 산출된 분산값

및

으로 계산된 가중치

를 적용하여 포커스 인덱싱(F 또는 F+F_H)을 산출한다.The focus indexing calculator 450 calculates a focus index

The calculated variance value < RTI ID = 0.0 >

And

Weight calculated by

To calculate the focus indexing (F or F + F _H ).

의 밝기 판단부(440)의 밝기 판단 결과, 일반 밝기 값을 갖는 영상(T_h 이하 및 T_l이상)에 대해 수학식 7과 같은 제1 포커스 인덱싱(F)을 산출한다.The first focus indexing calculation unit 451 calculates a focus index

The first focus indexing F as shown in Equation (7) is calculated for the image having the normal brightness value (T _h or less and T _l or more) as the brightness determination result of the brightness determination unit 440.

의 밝기 판단부(440)의 밝기 판단 결과 상기 밝기 판단 결과 특정 밝기값 이상(T_h) 및 특정 밝기값 이하(T_l)인 영상에 대해 수학식 8과 같은 포커스 인덱싱(F_H)를 산출한다.The second focus indexing calculation section 452 calculates the second focus index &

The focus determination unit 440 calculates a focus index F _H as shown in Equation 8 for an image having a brightness value T _h and a brightness value T _l less than a specific brightness value .

The focus indexing output unit 453 calculates a ratio of a portion of the image that is higher than a specific brightness value T _{h to a} darkest region, that is, a portion that is lower than a specific brightness value T _l , . A specific gray level or more (T _h) and a part specific brightness value or less (T _l) in case the portion occupied by, for example more than 1/4 across the image, a first focus index and the second indexing summing the focus (F + F _H ), and when the portion having a specific brightness value equal to or higher than the specific brightness value T _h and the portion having a specific brightness value equal to or less than the specific brightness value T _l occupy less than 1/4, for example, in the entire image, the first focus indexing F Output.

After completing the focus indexing operation for one frame, if the area corresponding to the highlight / dark area in the image does not exceed a certain portion of the entire image, it is determined as a spot highlight and an exception process is performed. , It is recognized as focus indexing normally and added to the general area focus indexing to detect the focus.

The present invention has been described with reference to the preferred embodiments. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is defined by the appended claims rather than by the foregoing description, and all differences within the scope of equivalents thereof should be construed as being included in the present invention.

1 is a block diagram showing the configuration of a focus detection apparatus robust against a low-illuminance noise according to the present invention.

FIG. 2 is a diagram showing a noise estimation result and a histogram distribution diagram for an image to which 10 Db AWGN is added in FIG.

FIG. 3 is a view comparing a conventional focus detection result and a focus detection result according to the present invention.

4 is a block diagram showing the configuration of a focus detection apparatus robust against low-light noise and spot highlight according to the present invention.

Claims (7)

A digital image processing apparatus comprising: An average calculating unit for calculating an average of a predetermined block including pixels currently being processed from the input image by using a video signal including noise as an input image; A difference image calculation unit for calculating the difference image of the input image signal and the average calculated by the average calculation unit; A variance calculation unit for calculating a variance value of a video signal and noise from the input video signal, the calculated average, and the calculated difference image; And And a focus indexing calculation unit for calculating a focus index by applying a weight calculated by the calculated variance value to the difference image. Claim 2 has been abandoned due to the setting registration fee. 2. The focus detection apparatus of claim 1, wherein when the number of the video signals is larger than the noise of the input video signal, the weight value is close to unity. Claim 3 has been abandoned due to the setting registration fee. 2. The focus detection apparatus according to claim 1, wherein when the noise is larger than a video signal of the input image signal, the weight is close to zero. A digital image processing apparatus comprising: An average calculating unit for calculating an average of a predetermined block including pixels currently being processed from the input image by using a video signal including noise as an input image; A difference image calculation unit for calculating the difference image of the input image signal and the average calculated by the average calculation unit; A variance calculation unit for calculating a variance value of a video signal and noise from the input video signal, the calculated average, and the calculated difference image; A brightness determination unit for determining a brightness value (T _h ) and a specific brightness value (T _l ) of brightness of the input image; And And a focus indexing unit for calculating a focus index by applying a weight calculated by the calculated variance value to the input image according to the brightness determination result. Claim 5 has been abandoned due to the setting registration fee. 5. The apparatus of claim 4, wherein the focus indexing unit And applying a weight calculated by the calculated variance value to the input image having a general brightness value (less than a specific brightness value (T _h ) and a specific brightness value (T _l )) as a result of the brightness determination, A first focus indexing calculation unit for calculating a focus index; And calculating a second focus index by applying a weight calculated by the calculated variance value to the input image having a brightness value T _h or more and a brightness value T _l or less as a result of the brightness determination, A focus indexing calculation unit; And Said particular brightness value (T _h) or higher and the specific brightness value (T _l) If the image having less than a predetermined portion of the input image the first and second output to the focus index and the specific brightness value (T _h) above And an output unit for outputting the first focus indexing if an image having a specific brightness value ( _Tl ) or less is not equal to or more than a predetermined portion of the input image. Claim 6 has been abandoned due to the setting registration fee. 6. The apparatus of claim 5, wherein the first and second focus indexing units Wherein when the number of video signals is larger than the noise of the input video signal, the weighting factor is set to 1. Claim 7 has been abandoned due to the setting registration fee. 6. The apparatus of claim 5, wherein the first and second focus indexing units Wherein when the noise is larger than the video signal of the input image signal, the weight is applied as 0.

Images