KR20110103196A - System and method for controlling quality of image - Google Patents

Abstract

An image quality control system and method are provided. The image quality control system divides the difference image signal of the image signal into a plurality of blocks to perform discrete cosine transform, and generates a histogram of the discrete cosine transform coefficients for each component of the discrete cosine transform coefficients. Based on the histogram, when the coefficient is quantized, a cognitive image quality calculation unit according to the distortion value of the coefficient generated according to the quantization interval and a bit rate value of the image according to the quantization interval are calculated, and the calculated cognitive And a bit rate calculator for calculating a bit rate value for encoding the video signal into an image having a predetermined quality, based on the image quality value and the calculated generated bit rate value.

Description

Image quality control system and method {SYSTEM AND METHOD FOR CONTROLLING QUALITY OF IMAGE}

The present invention relates to an image quality adjusting system and method, and more particularly, to an image quality adjusting system and method for predicting a distortion value and a bit rate of a captured image so as to maintain a preset image quality. will be.

Due to the remarkable development of communication technology, the transmission and storage of video information through a network is becoming an essential living item for living in modern society, and video surveillance is being made through ubiquitous sensor network (USN). .

In such a video service, the capacity of the network can be used is limited. Therefore, in order to maximize the transmission and storage efficiency of the video information, an image for obtaining an optimal video quality within the capacity of a given network can be obtained. Compression technology is needed. As a representative technology of image compression, JPEG is standardized for still image compression, H.261, H.263, MPEG-1, MPEG-2, MPEG-4, etc. Uses a method of eliminating redundant information in time or space.

On the other hand, in the video surveillance system, conventionally, the captured surveillance video is encoded at a constant bit rate regardless of the situation. Therefore, when there is no problem with security, the bit rate is higher than necessary, which causes waste of network bandwidth. When a situation arises and a high level of image quality is required, the bit rate is insufficient to provide the desired level of image quality. There was a problem.

Accordingly, as image object recognition technology and motion recognition technology have been recently developed, a technology for automatically searching for an object to be monitored through video has become possible, and the situation can be determined from the captured surveillance image. For example, it analyzes the characteristics of unusual unusual behaviors or unusual objects such as attempting to withdraw the cash automatically from the masked person, dropping the bag at the airport, fire or intruder, and automatically detect them. Alarms can be displayed, which reduces the need and effort to monitor thousands of videos 24 hours a day.

However, the conventional technique of sending at a constant bit rate is difficult to send an image of a desired quality because the image quality is relatively poor when the movement of a subject or the screen is complicated, so that it is transmitted at a fixed bit rate regardless of the situation. As a result, waste of transmission bandwidth usually occurs and when a situation occurs, it is difficult to obtain desired image quality. In addition, even when the image quality is differentiated and provided according to the movement of the subject, when there is much change in the operation of the subject, there is a problem that it is difficult to continuously predict and provide an image of a desired quality accurately.

Therefore, the present invention proposes a technique that can calculate the relationship between the image quality and the bit rate of the image.When the relationship between the image quality and the bit rate is calculated by the present invention, the required image quality is set in advance for each situation, and the necessary bit rate is determined according to the parameter of the encoder. Can be given as

Some embodiments of the present invention provide an encoder that predicts an image distortion value according to a quantization interval from a difference image signal of an image, and calculates an image quality-bit rate relationship of the image based on the predicted image distortion value to obtain a desired bit rate encoder. It provides a video quality control system and method that outputs the data to the encoder for use as an input parameter.

In addition, an embodiment of the present invention, predicts the image quality according to the quantization interval of the difference image signal of the image reflecting the human visual characteristics, and predicts the bit rate value to be applied to the encoding of the image based on the quantization interval and the image quality And image quality control systems and methods.

In addition, an embodiment of the present invention provides an image quality control system and method capable of providing an accurate and continuous image of a predetermined quality according to a monitoring situation.

As a technical means for achieving the above technical problem, a first aspect of the present invention is a differential image signal processing unit for performing discrete cosine transform of the difference video signal of the video signal and generating a histogram of the discrete cosine transform coefficient for each frequency component, each component A distortion value-quantum relation calculation unit that calculates a distortion value of the coefficient according to the quantization interval, which is generated when the coefficient is quantized based on the respective histogram, and an amount that calculates the generated bit rate value of the image according to the quantization interval. A bit rate for encoding the video signal into an image having a predetermined quality, based on a self-bit rate relationship calculating unit and a distortion value of the coefficient according to the calculated quantization interval, and a generated bit rate value of the image according to the calculated quantization interval. Providing an image quality adjusting system including an image quality-bit rate relationship calculating unit for calculating a value Can.

In addition, the second aspect of the present invention, the step of dividing the difference video signal of the video signal into a plurality of blocks, discrete cosine transforming the divided difference video signal, and generating a histogram of the discrete cosine transform coefficient for each frequency component Estimating a distortion value of the coefficient generated according to a quantization interval when the coefficient is quantized based on the histogram, and predicting a bit rate value of an image according to a quantization interval based on the histogram. And calculating a bit rate value for encoding into an image having a predetermined quality, based on the distortion value and the generated bit rate value.

According to the above-described problem solving means of the present invention, by predicting the image quality and bit rate value according to the quantization interval for each frequency component of the obtained image, it is possible to provide the image obtained by encoding the desired quality.

In addition, according to the above-described problem solving means of the present invention, by predicting the image quality by reflecting the human visual characteristics, it is possible to more effectively encode the obtained image to the image of a specific quality.

In addition, according to one of the other problem solving means of the present invention, it is possible to provide a more accurate and continuous image of the preset quality according to the monitoring situation.

1 is a schematic diagram illustrating an example of a video surveillance system for determining a surveillance situation and encoding an image with an appropriate image quality according to a determination result according to an embodiment of the present invention.
2 is an overall configuration diagram of a video surveillance system according to an embodiment of the present invention.
3 is a detailed block diagram of the image quality control system 3000 according to an embodiment of the present invention.
4 is a detailed block diagram of the image quality-bit rate calculator 3200 according to an embodiment of the present invention.
5 is an example of an image quality setting table for each monitoring situation according to an embodiment of the present invention.
FIG. 6 is an example of a histogram of discrete cosine transform coefficients of respective components of a difference image signal divided into 4 × 4 pixel units according to an embodiment of the present invention.
7 is a graph illustrating distortion generation of quantized DCT coefficients according to an embodiment of the present invention.
8 is an example of a graph illustrating a relationship between an image distortion value (RMSE) and cognitive image quality according to an embodiment of the present invention.
9 is a flowchart illustrating a method of controlling image quality for each monitoring situation according to an embodiment of the present invention.
10 is a detailed flowchart of an image quality adjusting method according to an embodiment of the present invention.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In the drawings, parts irrelevant to the description are omitted in order to clearly describe the present invention, and like reference numerals designate like parts throughout the specification.

Throughout the specification, when a part is "connected" to another part, this includes not only "directly connected" but also "electrically connected" with another element in between. . In addition, when a part is said to "include" a certain component, which means that it may further include other components, except to exclude other components unless otherwise stated.

Hereinafter, with reference to the accompanying drawings will be described in detail the present invention.

FIG. 1 is a schematic diagram illustrating an example of a video surveillance system that determines a surveillance situation and encodes an appropriate image quality according to a determination result according to an embodiment of the present invention. As shown in FIG. 1, in a video surveillance system according to an exemplary embodiment of the present invention, a recognition apparatus recognizes a content or a situation of an image and based on this, analyzes an image quality-bit rate relationship of the image in real time based on the situation. In this case, the encoder may inform the encoder of a bit rate suitable for the optimal image quality. The encoder performs rate control according to the bit rate input as a parameter, so that the encoder can encode the bit rate according to the input bit rate. Accordingly, according to an embodiment of the present invention, it is possible to provide an optimal picture quality adjusted to a surveillance situation and to reduce waste of transmission network bandwidth.

2 is an overall configuration diagram of a video surveillance system according to an embodiment of the present invention.

As shown in FIG. 2, the video surveillance system according to an embodiment of the present invention includes an image capturing unit 1000, an object / motion recognition system 2000, an image quality adjusting system 3000, an encoder 4000, a network. And 5000 and a monitoring server 6000.

The image capturing unit 1000 photographs a subject. The image capturing unit 1000 may be installed in a ceiling of a specific place to photograph a subject and a subject in the region, and provide the captured image to the object / motion recognition system 2000 and the image quality control system 3000 which will be described later. can do.

The object / motion recognition system 2000 analyzes the input image and provides the image quality control system 3000 with information on the situation being monitored. The object / motion recognition system 2000 may generate information about the current monitoring situation based on the motion information of the subject in the received image. In addition, the situation determination unit 2200 may collect and use a movement change rate of the subject or a temperature of a monitoring target by using a separate sensor (motion detection sensor, temperature sensor, etc.), but is not limited thereto.

The image quality adjusting system 3000 sets an image quality according to the surveillance situation, and outputs an optimal bit rate to encode the received image signal with the image quality corresponding to the surveillance situation.

Also, the image quality control system 3000 may predict an image distortion value according to a quantization interval based on the difference image signal of the received image signal, and predict the image quality based on the predicted image distortion value and human visual characteristics. have. The difference video signal refers to a signal indicating a difference between a previous picture signal and a current picture signal. The image quality control system 3000 may predict a bit rate value according to the quantization interval from the received image signal, and calculate a bit rate value for a specific image quality in consideration of the quantization interval.

The detailed configuration of the image quality adjusting system 3000 will be described later.

The encoder 4000 encodes an image signal received from the image capturing unit 1000 according to a bit rate output from the image quality adjusting system 3000.

The network 5000 may be a wired or mobile radio communication network or satellite, such as a local area network (LAN), a wide area network (WAN), or a value added network (VAN). It includes all kinds of wireless networks such as a communication network and the like, and a combination thereof, and means a data communication network in a comprehensive sense so that each network constituent illustrated in FIG. 2 can communicate with each other smoothly.

In FIG. 2, the image capturing unit 1000, the object / motion recognition system 2000, and the encoder 4000 are described as being separate from the image quality adjusting system 3000, but the present disclosure is not limited thereto. It may be included in the image quality control system 3000.

The monitoring server 6000 receives and stores the encoded video signal from the encoder 4000 and performs a necessary function such as alarming as necessary. In addition, the monitoring server 6000 may monitor the monitoring situation based on the received signal.

Hereinafter, a detailed configuration of an image quality adjusting system 3000 according to an embodiment of the present invention will be described with reference to FIG. 3.

3 is a detailed block diagram of the image quality control system 3000 according to an embodiment of the present invention.

As illustrated in FIG. 3, the image quality adjusting system 3000 includes an image quality setting unit 3100, an image quality-bit rate relationship calculating unit 3200, and a controller 3300. . The image quality control system 3000 receives a current situation determined from the object / motion recognition system 2000, receives a video signal photographed from the photographing unit 1000, and inputs a bit rate to the encoder 4000 as a parameter. You can output

The image quality setting unit 3100 sets a surveillance situation and an image quality corresponding to the surveillance situation. The video quality setting unit 3100 may set a plurality of monitoring situations in advance, and may set the video quality to be provided for each set monitoring situation.

For example, the image quality setting unit 3100 may distinguish a surveillance situation into a general situation, an abnormal symptom situation, and an emergency situation, and set the image quality to be encoded by varying the image quality or resolution according to each situation. Can be. Therefore, it is possible to encode a video with a low picture quality in a normal situation and to encode a video with a high picture quality in an emergency situation, thereby saving network resources and efficiently obtaining a surveillance video.

The image quality setting unit 3100 may interpret a signal related to a current monitoring situation received from the object / motion recognition system 2000 and map it to any one of the set monitoring situations.

The image quality-bit rate relationship calculator 3200 calculates and provides a relationship between the quality of the image and the bit rate to the controller 3300. The image quality-bit rate relationship calculator 3200 predicts an image distortion value according to the quantization interval based on the difference image signal of the image signal, and relates the image quality and the quantization value based on the image distortion value and human visual characteristics. Can be predicted. The image distortion value may be a root mean square error (RMSE) of a difference between a DCT coefficient of the difference image signal and a DCT coefficient of the quantized signal of the difference image signal. In addition, the difference image signal is a signal representing the difference between the previous image signal and the current image signal, and may be generated by subtracting the previous image signal having motion compensation to the current image signal.

The image quality-bit rate relationship calculating unit 3200 performs a discrete cosine transform (DCT) on the difference image signal, generates a histogram for each component of the discrete cosine transformed coefficients (calculates the cumulative occurrence frequency), and histograms. From the image distortion value generated according to the quantization interval can be calculated. In addition, the image quality-bit rate relationship calculator 3200 may calculate cognitive image quality reflecting human visual characteristics, based on the image distortion value calculated according to the quantization interval. Furthermore, the image quality-bit rate relationship calculator 3200 may calculate a bit rate according to the quantization interval, and thus may calculate a relationship between cognitive image quality and the bit rate.

The controller 3300 controls the operation of the entire image quality adjusting system 3000, receives an image quality value necessary for the current situation from the image quality setting unit 3100, and receives an image from the image quality-bit rate relationship calculating unit 3200. Receiving information on the relationship between the quality and the bit rate, the optimum bit rate necessary for encoding the current video into the video quality corresponding to the current situation is output.

The bit rate value output from the controller 3300 is input as an input parameter of the encoder 4000, and the encoder 4000 encodes a video signal according to the received bit rate and sends the bit stream to the monitoring system 6000 through the network 5000. ).

Hereinafter, a detailed configuration of the image quality-bit rate relationship calculator 3200 according to an embodiment of the present invention will be described with reference to FIG. 4.

4 is a detailed block diagram of the image quality-bit rate relationship calculator 3200 according to an embodiment of the present invention.

As shown in FIG. 4, the image quality-bit rate relationship calculator 3200 according to an embodiment of the present invention may include a difference image signal processor 3210, a distortion value-quantum relationship calculator 3220, and image quality-distortion. A value relationship calculator 3230, a quantum-bit rate relationship calculator 3240, and an image quality-bit rate relationship calculator 3250.

The difference image signal processor 3210 generates a difference image signal through motion compensation from the image signal received from the image capturing unit 1000, and generates a histogram based on the generated difference image signal.

The difference image signal processor 3210 may divide the difference image into a plurality of blocks, respectively, and perform discrete cosine transform on each of the divided images, and generate a histogram of discrete cosine transform coefficients for each frequency component.

For example, the difference image signal processor 3210 may divide the difference image into blocks having an N * N pixel number, and perform discrete cosine transform (DCT) on each block. In this case, one block may be distinguished by components {(u, v), 0 ≦ u ≦ N−1 and 0 ≦ v ≦ N−1}, where u is a frequency in the horizontal direction and v is a frequency in the vertical direction. . In addition, the difference image signal processor 3210 may generate a histogram of the discrete cosine transform coefficients after the discrete cosine transform for each block. That is, for example, when the difference image signal of 352 x 288 pixels is divided into 4 x 4 blocks, the result is 88 x 72 blocks. In this case, a histogram can be generated by collecting only (0,0) components of the discrete cosine transform coefficients in all 88 x 72 blocks. Similarly, 16 histograms can be created, from (0,1) to (3,3).

The histogram for each component of each discrete cosine transform coefficient of the difference image signal may have a Laplace distribution, and the histogram distribution of the coefficient of the (u, v) th DCT component is modeled using the Laplace model as in Equation 1 below. Can be represented.

는 (u,v) 번째 DCT 성분의 계수의 히스토그램의 라플라스 모델 인자이며, 아래의 수학식 2와 같이 계산될 수 있다.
Here, u and v are indexes of horizontal and vertical components in the DCT block. Also,

Is a Laplace model factor of the histogram of the coefficient of the (u, v) th DCT component, and may be calculated as in Equation 2 below.

는 (u,v) 번째 DCT 성분의 계수 히스토그램의 표준편차이다.In Equation 2,

Is the standard deviation of the coefficient histogram of the (u, v) th DCT component.

Based on the histogram of the DCT component coefficients, the distortion value-quantum relationship calculating unit 3220 calculates an image distortion value (eg, a root mean square error value) generated according to the quantization interval Q, as shown in Equation 3 below. Can be calculated or predicted. The distortion value-quantum relationship calculator 3220 may measure the distortion of the encoded image by predicting the distortion of the coefficient in the process of quantizing the difference image signal.

The distortion value-quantum relationship calculating unit 3220 may calculate a distortion value (RMSE) according to a change in the quantization interval, so as to prevent an increase in the amount of computation due to repetitive quantization, the histogram of the coefficient distribution of each DCT component. The distribution function can be modeled and used.

In one embodiment of the present invention, the image distortion value may be defined as a root mean square (RMSE: Root Mean Sqaure) of the difference between the DCT coefficient of the original signal of the difference image signal and the DCT coefficient of the quantized signal. It doesn't work. The distortion prediction value of the DCT coefficient of the (u, v) th component with respect to the quantization interval Q may be calculated as in Equation 3 below.

In Equation 3, Du, v (Q) is a distortion value (RMSE) according to the quantum Q of the (u, v) th DCT component, and Pu, v is a histogram distribution modeling equation of the (u, v) th DCT component Equation 1), x is the horizontal axis of the histogram shown in Figure 7, iQ is the quantization level interval of each of the histogram shown in Figure 7, N is the number of blocks in one image.

The image quality-distortion relationship calculation unit 3230 calculates a relationship between an image quality perceived by a human being (cognitive image quality) and an image distortion value.

In general, the relationship between the image quality perceived by human vision and the image distortion value RMSE may be represented by a curve including two inflection points rather than a linear relationship as shown in FIG. 8. FIG. 8 is a diagram illustrating an example of a relationship between quality of vision perceived by human vision and RMSE. In FIG. 8, the horizontal axis represents image distortion value RMSE, and the vertical axis represents degree of image quality perceived by human vision. The vertical axis value may range from a maximum of 1 to a minimum of zero. For example, the quality perceived by human vision may be divided into intervals I, II, and III, as shown in FIG. 8, and even if the distortion value RMSE increases until the interval I, the quality of human vision does not change much. Do not. However, if the distortion value RMSE increases above the inflection point B (section II), the quality perceived by human vision begins to decrease with the distortion value RMSE. Then, if the distortion value RMSE exceeds the inflection point C (section III), the quality perceived by human vision no longer decreases significantly even if the distortion value RMSE increases. These intervals I, II and III can be defined as follows.

Interval I (Point a to Point b): An area where the RMSE distortion of the signal occurs, but where there is little noticeable deterioration in cognitive image quality (corresponding to cognitive quality 1 to 0.9).

Interval II (point b to point c): An area where the degradation of cognitive image quality increases as the signal's RMSE distortion increases (corresponding to cognitive quality 0.9 to 0.1).

Interval III (Point c to Point d): An area where the RMSE distortion of the signal increases, but where the degradation of cognitive image quality no longer increases (corresponding to cognitive quality of 0.1 to 0).

Points a and d can then be determined theoretically. That is, in theory, the maximum distortion that may occur due to the quantization of the image codec may occur at the maximum quantizer value (for example, 51 for the H.264 codec and 31 for the MPEG-2 codec). In this case, in the case of the H.264 codec, the quantization level interval is 224. In the uniform quantizer assumption, the maximum distortion due to quantization is 112, and in the case of MPEG-2, the quantization level interval is 32 and the maximum distortion is 31.

When the distortion value RMSE is minimum (0) and maximum (H.264 codec: 112, MPEG-2 codec: 31), theoretically may be mapped to cognitive image quality 1.0 and 0.0, respectively. Therefore, the theoretical points a and d can be set identically for all images. However, points b and c depend on the characteristics of the input image. Point b represents a maximum distortion value (RMSE) representing an image quality similar to that of an original image, and point c represents a minimum distortion value (RMSE) representing an image quality similar to an image encoded with the maximum quantization value. The cognitive image quality of points b and c may be defined as 0.9 and 0.1, respectively.

)를 갖는 로지스틱(logistic) 함수를 다음과 같이 정의할 수 있다.
In order to derive the cognitive image quality of human vision according to the distortion value RMSE, four factors ((a, b, c, d))

We can define a logistic function with) as

In Equation 4, D is a distortion value (RMSE) of an image, and PQ is a cognitive image quality of human vision.

는 아래 수학식 5와 같이 지점(a, b, c, d)와 모델의 제곱 오차를 최소화하는 모델 인자로서 구할 수 있다.In equation (4)

Can be obtained as a model factor that minimizes the square error of points (a, b, c, d) and the model as shown in Equation 5 below.

In Equation 5, a and d are theoretically determined by the minimum and maximum distortion values, and b and c may be determined by characteristics and experiments of human vision. Existing studies on the sensitivity of video signals to frequency are known that the human visual system is more sensitive to low frequency signals.

Accordingly, at the lowest distortion three components {(u, v) = (0,0), (0,1) and (1,0)} for the total distortion value, that is, the sum of the distortion values generated in each DCT component We can define the point where 40% of the sum of the distortion values of is b and the point where 75% is c.

In Equation 4, the closer the cognitive image quality value PQ (D) is to 1, the closer to the original image with little cognitive distortion, and the closer the cognitive image quality value is to 0, the more visually human the image is. The distortion of the image becomes large.

The quantum-bit rate relationship calculator 3240 calculates a relationship between the quantum and the bit rate in order to calculate a bit rate value necessary for encoding the image having the set quality.

The quantizer-bit rate relationship calculator 3240 may estimate the generated bit rate according to the quantization interval in each DCT component based on the histogram information of each DCT component. The generation bit rate of each DCT component may be estimated based on entropy according to the quantization factor, and the generation rate in the (u, v) th DCT component in the DCT block may be estimated as in Equation 6.

은 양자화 간격 Q로 영상을 부호화 했을 때, 양자화 레벨 iQ로 맵핑되는 DCT 계수의 분포 확률을 나타내며,

은 아래 수학식 7과 같이 정리될 수 있다.
here,

Represents the probability of distribution of the DCT coefficients mapped to the quantization level iQ when the image is encoded at the quantization interval Q.

Can be summarized as in Equation 7 below.

In addition, the generated bit rate of the image according to the quantization interval may be calculated as the sum of the generated bit rates of all DCT components, as shown in Equation 8, by using Equation 6.

The quantum-bit rate relationship calculator 3240 may calculate a relationship between the quantization interval value Q and the bit rate based on Equation 8, and calculates the calculated quantum-bit rate relationship information from the image quality-bit rate relationship calculator ( 3250).

The image quality-bit rate relationship calculating unit 3250 is configured to the values calculated by the distortion value-quantum relationship calculating unit 3220, the cognitive image quality-distortion relationship calculating unit 3230, and the quantum-bit rate relationship calculating unit 3240. Based on this, the relationship between the image quality and the bit rate can be calculated. The image quality-bit rate relationship calculating unit 3250 calculates an image distortion value (the relationship between the image distortion value and the quantizer) according to the quantization interval, a relationship between the image quality and the distortion value, and a bit rate value of the image according to the quantization interval (between the quantum bit rate). Relationship), it is possible to accurately calculate the relationship between the image quality and the bit rate.

In addition, the image quality-bit rate relationship calculator 3250 provides the controller 3300 with relationship information between the calculated image quality and the bit rate. The controller 3300 may receive image quality information set according to a situation from the image quality setting unit 3100 and output a bit rate value necessary for the set image quality, and the encoder 4000 outputs the bit rate value output from the controller 3300. Can be used as an encoding parameter to encode the received video signal.

According to an embodiment of the present invention, even if the moving amount of the subject in the image changes, in order to maintain a certain quality image, the image quality according to the quantization interval is predicted from the captured image signal, and the bit rate value according to the quantization interval is predicted. Through this, it is possible to accurately calculate a bit rate value for a specific image quality, and at the same time maintain a desired image quality.

Hereinafter, an example of an image quality setting table for each monitoring situation according to an embodiment of the present invention will be described with reference to FIG. 5.

5 is an example of an image quality setting table for each monitoring situation according to an embodiment of the present invention.

As illustrated in FIG. 5, the image quality setting unit 3100 may set a plurality of monitoring situations and may set image quality to be provided for each set monitoring situation. In detail, the image quality setting unit 3100 may classify a monitoring situation into a general situation, an abnormal symptom situation, and an emergency situation, and may set an image quality or resolution according to each situation. In addition, the cognitive image quality value can be normalized to a value between 0 and 1.

Therefore, according to an embodiment of the present invention, a high-definition image is generated in an emergency situation in which an image is encoded at low image quality in a normal situation and a high image quality is encoded in an emergency situation, thereby saving network resources. It is possible to efficiently obtain the surveillance image by providing a.

Hereinafter, an example of a histogram of discrete cosine transform coefficients for each component of a difference image signal according to an embodiment of the present invention will be described with reference to FIG. 6.

FIG. 6 is an example of a histogram of discrete cosine transform coefficients of respective components of a difference image signal divided into 4 × 4 pixel units according to an embodiment of the present invention.

In FIG. 6, the first graph in the upper left shows a histogram of coefficients of the (0,0) th DCT component, that is, the DC component. The remaining graphs show histograms of the coefficients of the AC component. The graphs to the right and the bottom of the DC component show a histogram of the coefficients of the (0,1) and (1,0) th DCT components. The bottom right graph shows a histogram of the coefficients of the (3,3) th DCT components. In addition, as shown in FIG. 6, the histogram of the coefficient of each DCT component of the difference image signal may have a Laplace distribution.

Hereinafter, the distortion of the quantized DCT coefficients according to an embodiment of the present invention will be described with reference to FIG. 7.

7 is a graph illustrating distortion generation of quantized DCT coefficients according to an embodiment of the present invention.

As shown in FIG. 7, the DCT coefficient of the difference image signal modeled as the Laplace distribution has a distorted value through the quantization process. 7 is the value of the DCT coefficient, and the vertical axis is the number of degrees of the histogram. Originally, DCT coefficients of an original signal are mapped to quantization levels distributed at regular intervals through a quantization process, and image distortion occurs due to a difference between DCT coefficients of the original signal and DCT coefficients due to quantization.

Hereinafter, a relationship between an image distortion value RMSE and an image quality perceived by human vision according to an embodiment of the present invention will be described with reference to FIG. 8.

8 is an example of a graph illustrating a relationship between an image distortion value (RMSE) and cognitive image quality according to an embodiment of the present invention.

As shown in FIG. 8, the image quality perceived by human vision can be normalized from a maximum of 1 to a minimum of 0. In theory, the maximum image quality is when the distortion value RMSE is 0 (point a) and the minimum image quality. Is when the distortion value RMSE is maximum (point d). However, experiments show that the cognitive image quality between a and d is not inversely inversely related to the distortion value RMSE. As shown in FIG. 8, two inflection points b and c may exist. Even if the distortion value RMSE increases from a to the inflection point b, the cognitive image quality does not decrease much. As the distortion value RMSE increases from the inflection point b to the inflection point c, the cognitive image quality decreases. From the inflection point c to d, the cognitive image quality no longer decreases even when the distortion value RMSE increases. The relationship between the distortion value RMSE and cognitive quality can be obtained by obtaining the inflection points b and c.

Hereinafter, a method of adjusting image quality for each monitoring situation according to an embodiment of the present invention will be described with reference to FIG. 9.

9 is a flowchart illustrating a method of controlling image quality for each monitoring situation according to an embodiment of the present invention.

Step S900 is a step of setting the video quality according to the monitoring situation. In operation S900, a plurality of monitoring situations may be set, and an image quality to be provided for each set monitoring situation may be set. In addition, for example, the monitoring situation may be classified into a general situation, an abnormal symptom situation, and an emergency situation, and may be set to encode an image by changing image quality or resolution according to each situation. Therefore, it is possible to obtain a surveillance video efficiently while saving network resources by encoding an image with low image quality in normal situations and encoding an image with high image quality in emergency situations.

Step S902 is a step of determining the monitoring situation. In step S902, the information on the current video signal monitoring situation determined from the object / motion recognition system 2000 is received, and the received signal is interpreted to map the monitoring situation of the current video signal to any one of the monitoring situations set in S900. can do.

In step S904, the image signal to be photographed is encoded into an image having a predetermined quality, and the image quality is adjusted.

In operation S904, the difference image signal after motion compensation may be obtained from the captured image signal, and the image distortion value and the cognitive image quality according to the quantization interval may be predicted, and based on this, a bit rate value for providing an image having a predetermined quality is obtained. Can be calculated in consideration of the quantization interval.

Hereinafter, a detailed method of controlling image quality according to an embodiment of the present invention will be described with reference to FIG. 10.

10 is a detailed flowchart of an image quality adjusting method according to an embodiment of the present invention.

In operation S1000, the difference image signal is analyzed. In operation S1000, the difference image signal after motion compensation may be generated from the image signal received and received, and a histogram based on the generated difference image signal may be generated.

In operation S1000, the difference image may be divided, and the divided images may be discrete cosine transformed, and a histogram of the discrete cosine transform coefficients may be generated for each frequency component of each block. For example, a difference image may be divided into blocks having N * N pixels, and when each block is discrete cosine transformed, one block may be {(u, v), 0 ≤ u ≤ N-1 and 0 ≦ v ≦ N−1} components. In addition, a histogram of discrete cosine transform coefficients may be generated for each component of each block.

Step S1002 is a step of calculating or predicting an image distortion value (RMSE distortion value) generated according to the quantization interval based on the histogram of the DCT component coefficients. In operation S1002, the RMSE distortion of the encoded image may be measured by predicting the distortion of the coefficient in the process of quantizing the difference image signal. For example, when the difference image signal is divided into 4 x 4 components, the distortion of coefficients in a total of 16 DCT components must be predicted. In this case, the distortion of the coefficients is calculated using 16 distortion predictors, or one The distortion predictor may be repeatedly used to calculate the distortion of the coefficient.

Further, in step S1002, in order to prevent an increase in the amount of computation due to repetitive quantization in calculating a distortion value according to the change in the quantization interval, the distribution function of the histogram of the coefficient distribution of each DCT component may be modeled and used. Accordingly, by using the histogram's distribution function, each DCT value distribution function may be multiplied by the distortion to calculate the RMSE distortion caused by the quantization.

In operation S1004, the cognitive image distortion value is calculated by reflecting the human visual characteristic to the predicted image distortion value. In operation S1004, a relationship between the cognitive image distortion value and the RMSE image distortion value may be calculated. In step S1004, the image quality perceived by human vision may be obtained using the RMSE distortion value obtained in S1002.

Step S1006 is a step of predicting cognitive image quality according to the quantization interval. In step S1006, the cognitive image quality according to the quantization interval may be calculated based on the relationship between the RMSE image distortion value according to the quantization interval calculated in step S1002, the cognitive image distortion value calculated in S1004, and the cognitive image quality. have.

Step S1008 is a step of predicting a bit rate according to the quantization interval and calculating a bit rate value required for encoding into a video of a specific quality.

In operation S1008, the generation bit rate according to the quantization interval in each DCT component may be estimated based on the histogram information of each DCT component. An occurrence bit rate prediction of each DCT component may be estimated based on entropy according to a quantization factor. In operation S1008, the generation bit rate of the image according to the quantization interval may be calculated as the sum of the generation bit rates of all DCT components.

Therefore, in step S1008, a bit rate value for providing an image having a predetermined quality is provided through the quantization interval value Q based on the image quality value according to the quantization interval and the generated bit rate value of the image according to the quantization interval. Can be calculated.

In operation S1010, an image signal is encoded based on the calculated bit rate value. When the bit rate value obtained in S1008 is outputted to the encoder, the encoder can obtain the image quality required for the current situation by encoding using the bit rate value obtained in S1008 as an encoding parameter.

Therefore, according to an exemplary embodiment of the present invention, even if the situation of the subject in the image is changed, the image quality according to the quantization interval and the bit rate value according to the quantization interval are predicted from the captured image signal, so It is possible to calculate the bit rate value accurately and to maintain the set image quality continuously.

One embodiment of the present invention may also be embodied in the form of a recording medium including instructions executable by a computer, such as program modules, being executed by a computer. Computer readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. In addition, computer readable media may include both computer storage media and communication media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Communication media typically includes computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave, or other transmission mechanism, and includes any information delivery media.

The foregoing description of the present invention is intended for illustration, and it will be understood by those skilled in the art that the present invention may be easily modified in other specific forms without changing the technical spirit or essential features of the present invention. will be. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined form.

The scope of the present invention is shown by the following claims rather than the above description, and all changes or modifications derived from the meaning and scope of the claims and their equivalents should be construed as being included in the scope of the present invention. do.

3000: video quality control system
3100: image quality setting unit
3200: image quality-bit rate relationship calculator
3300: control unit
3210: difference image signal processing unit
3220: distortion value-quantum relationship calculation unit
3230: image quality-distortion relationship calculation unit
3240: quantizer-bit rate relationship calculation unit
3250: image quality-bit rate relationship calculation unit

Claims (15)

A differential image signal processing unit for discrete cosine transforming the difference image signal of the video signal and generating a histogram of the discrete cosine transform coefficients for each frequency component;
A distortion value-quantum relationship calculation unit for calculating a distortion value of the coefficient according to the quantization interval, which occurs when the coefficient is quantized based on the histogram for each component,
A quantizer-bit rate relationship calculating unit for calculating an occurrence bit rate value of an image according to a quantization interval;
An image quality-bit rate for calculating a bit rate value for encoding the video signal into an image of a predetermined quality, based on the distortion value of the coefficient according to the calculated quantization interval and the generated bit rate value of the image according to the calculated quantization interval Relationship calculation department
Including, the image quality control system.
The method of claim 1,
And an image quality-distortion relationship calculation unit calculating a relationship between the distortion value and cognitive image quality.
And the image quality-bit rate calculation unit calculates a bit rate value for encoding the video signal into an image having a predetermined quality based on the relationship between the calculated distortion value and cognitive image quality.
The method of claim 1,
An image quality setting unit which sets a plurality of monitoring situations and video quality corresponding to the monitoring situation, and determines the monitoring situation and the video quality corresponding to the video signal.
More,
And the image quality-bit rate calculator calculates a bit rate value for encoding the video signal at the determined image quality.
The method of claim 1,
The difference image signal processor divides the difference image signal into a plurality of blocks, and performs discrete cosine transform on the divided blocks to generate a histogram of discrete cosine transform coefficients for each frequency component of each block. system.
The method of claim 1,
The difference video signal is generated by subtracting a previous video signal to a current video signal in units of motion compensation blocks.
The method of claim 1,
And the distortion value-quantum relationship calculating unit models a distribution function of a histogram of each component and calculates the distortion value generated due to quantization using the modeled distribution function.
The method of claim 1,
And the distortion value is a root mean square (RMSE) value of a difference between a DCT coefficient of the difference image signal and a DCT coefficient of the quantized signal of the difference image signal.
The method of claim 2,
And an image quality-distortion relationship calculation unit models a relationship between the distortion value and cognitive image quality as a logistic function.
The method of claim 2,
The cognitive image quality value is normalized to a value of 1 to 0, image quality control system.
The method of claim 2,
The quantizer-bit rate calculation unit estimates a generation bit rate of each component according to a quantization interval based on entropy based on the histogram of each component, and estimates a generation bit rate of an image according to the quantization interval. And calculating the sum of the generated bit rates in each estimated component.
The method of claim 10,
The image quality-bit rate calculator is configured to provide a bit rate for providing the image having the preset quality based on the image quality value according to the quantization interval and the generated bit rate value of the image according to the quantization interval. Calculating a value.
Dividing the difference video signal of the video signal into a plurality of blocks;
Discrete cosine transforming the divided difference image signal, and generating a histogram of the discrete cosine transform coefficients for each frequency component;
Predicting a distortion value of the coefficients generated according to a quantization interval when the coefficients are quantized based on the histogram,
Predicting an occurrence bit rate value of an image according to a quantization interval based on the histogram;
Calculating a bit rate value for encoding into an image having a predetermined quality based on the distortion value and the generated bit rate value
Image quality control method comprising a.
The method of claim 12,
Calculating a relationship between the distortion value and cognitive image quality.
The method of claim 13,
Predicting cognitive image quality according to a quantization interval based on the relationship between the distortion value and cognitive image quality
More,
The calculating of the bit rate value may include calculating a bit rate value for encoding the image having the predetermined quality based on the cognitive image quality value according to the quantization interval.
The method of claim 12,
Setting a plurality of monitoring situations and image quality corresponding to each of the monitoring situations
More,
The calculating of the bit rate value comprises calculating a bit rate value for encoding the video signal with an image quality corresponding to the set monitoring situation.

Images