KR960010194B1 - Hd-mac. system mode generator - Google Patents

Abstract

The mode determinator, to determine a mode according to the motion of an image signal using the video signals filtered in low-speed mode, middle-speed mode and high-speed mode, the original image signals, and a motion vector selected through the motion estimation in the middle-speed mode, includes: a frame exchanging means for changing the frame sequence of the image signals filtered in the low-speed mode; an interpolating means for a motion error using the motion vector and the image signals filtered in the middle-speed mode; and a mode selecting means for comparing the image signals filtered in the high-speed mode, interpolated in the middle-speed mode and having the frame sequence changed in the low-speed mode with the original image signals in order to determine the mode having the least error as the mode.

Description

Mode determiner for HD-MAC systems

1 shows frequency characteristics of a two-dimensional low pass filter.

2 is a diagram illustrating error comparison images of the modes used in the conventional method.

3 is a block diagram of an HD-MAC system with a mode determiner in accordance with the present invention.

* Explanation of symbols for main parts of the drawings

10, 12, 14: low pass filter 16, 18, 20: sample reduction (SS)

22, 24: line shuffling (LS) section 26: switch

28: motion estimation unit 30: mode determiner

32: Matching unit 34: DATV

36: motion reinforcement interpolation (MCI) unit 38: frame exchange unit

40: error calculation and comparison unit

The present invention relates to a High Definition Multiplex Analog Component (HD-MAC) system, and more particularly, to a movement mode determiner for determining a movement mode.

HD-MAC, a high definition television (HDTV) developed in Europe, divides the shape signal into 16 × 16 divided images, and then divides each divided image into three paths according to object movement. Divide into modes and send and receive. At this time, since the number of samples transmitted per 4 fields (2 frames) of 16x16 blocks must be kept constant, it is effective to increase the number of spatial sampling while reducing the number of temporal sampling in the static region, and spatial sampling in the region where the motion is correct. Frequent temporal sampling is effective even if the frequency is reduced.

In this case, the above-described three path modes are a low speed mode, a quasi-high speed mode, and a high speed mode. The low speed mode is a mode for processing a still image without motion, and the quasi-high speed mode has an image that can be followed by the human eye. The high speed mode is a mode for processing an image with a considerably fast motion. More specifically, for example, if there is a motion of about 0 to 0.5 pixels during a 40 ms period, the image is classified into a still mode or an 80 ms mode, and when there is a motion of about 0.5 to 12.5 pixels during a 40 ms period, The image is classified into a slow motion mode or a 40 Hz mode, and if there is more than 12.5 pixels of motion during the 40 Hz period, the image is classified into a fast motion mode or a 20 Hz mode. In this case, the 80 Hz mode may be referred to as a low speed mode. The 40 kHz mode is a semi-high speed mode, and 20 kHz is a high speed mode.

On the other hand, in such a low-speed, semi-fast, and fast mode determination method, conventionally, the mode is finally determined to have the least error by calculating an error between the raw data of the filtered data and the interpolated data for each mode. That is, in the related art, a mode is determined by comparing an error-cost function between an original image and a filtered image of each path (low speed mode path, quasi-fast mode path, and fast mode path) (hereinafter referred to as F80, F40, and F20). will be. In this case, the cost function can be thought of as various types, but mainly uses the mean square error shown in [Equation 1].

[Equation 1]

Where B is a block, and O (i, j) and F (i, j) represent the original image and the filtered image.

Assuming that the inputs F80, F40, and F20 in each field interval are filtered images of the original image of the interval, the MSE value between the original image and the filtered image is determined only by the filtering error. In this case, as shown in FIG. 1, the original image is filtered by the first passband 1 (F80), and the original image is filtered by the second passband (F40) and the original image. When comparing the image F20 filtered by the third passband 3 with each other, the image F80 filtered by the first passband having the widest passband will be closest to the original image. Therefore, if the MSE value between the original image and the filtered image is determined only by the filtering error, the filtering error is determined to be the low speed mode, for example, 80㎳ mode because the F80 that passes through the filter having the largest passband is the minimum. The likelihood is very high (see Figure 1). F40 also has a higher probability of determination than F20. In order to avoid such a determination error, a process of increasing the MSE value is required in the velocity portion of each route image that is not included in the velocity range of the route mode. A process of a conventional mode determination method of determining a mode by calculating an MSE value will be described with reference to FIG. 2.

First, a frame repetition technique is used in a low speed mode, for example, an image of 80 Hz. After constructing frames 1 and 2 into frames, two-dimensional filtering is used and it is used as input of decision module for 4 field intervals. Therefore, in the 3rd and 4th field sections, an error occurs due to frame repetition in the non-still region. This error increases the MSE of the F80's motor part, preventing it from being incorrectly determined to be in 80㎳ mode.

In the quasi-high speed mode, for example, 40 ms mode, an odd frame is formed by filtering in consecutive odd-numbered fields, and an even frame is formed by a motion estimation / motion compensation method. Each field component is extracted from the even frame and the odd frame formed through filtering to calculate an error with the raw field and use the error value for mode determination.

A more detailed description of the even frame configuration is as follows. First, 1, 3, 5, 7... Filtering fields 1, 3, 5, 7... It consists of one frame. Then, the motion vector is detected by performing normal motion estimation on frames 1 and 3, and then motion compensation is performed based on the detected motion vector to determine the motion compensated frame as frame 2. In addition, frame 4 performs motion estimation on frames 3 and 5, respectively, using the previous 9 motion vectors detected by the above-described motion estimation process. After selecting a large motion vector, motion compensation is performed based on the selected motion vector to determine the motion compensated frame as the fourth frame.

In this case, the motion estimation estimates how much the pixels (or blocks) of the current frame move in the direction compared to the previous frame. The motion estimation is performed twice in four field sections (two frames), and each 16 × 16 block is divided into six pixels up, down, left, and right. As you move, you find the location of the smallest error.

The motion vector is obtained using an unconditional search algorithm that obtains a correlation between all search points in the search area and then estimates the search point having the largest correlation as the motion vector.

On the other hand, in the high speed mode, unlike the high speed mode or the low speed mode, samples of the even fields (2, 4, 6 ...) are not transmitted. Instead, only sample data of the odd fields is transmitted. Therefore, the motion vectors obtained at the time of motion estimation are separately transmitted. This is for the decoder to generate data of even fields that are not transmitted using sample data of odd fields and motion vectors.

In the high speed mode, for example, 20 ms mode, each field is filtered and compared with the raw image, and the error value is used for mode determination.

However, conventionally, in determining the moving mode, the filtered image F80 of the low speed mode path includes only the filtering error in the first and second field sections, and the filtered image F40 of the quasi high speed mode path is 1, Since only the filtering error is included in the field section 3, the error probability for the mode decision is still high. Although the error of the mode decision is reduced by using the motion compensation / frame repetition technique, as shown in FIG. 1, since the passband of the filter has a great influence on the mode decision, it is most often determined as the low speed mode. In addition, even in the low speed mode, the quasi-high speed mode, in which the passband of the filter is larger, is more likely to be determined than in the high speed mode, and there is still a problem in that the error probability for mode determination is still high.

Accordingly, the present invention is to solve the above problems, in order to avoid the determination error when determining the mode in HD-MAC, to determine the correct mode by increasing the MSE value in the speed portion of the path mode that is not included in the path range of the path mode It is to provide a mode determiner that can.

In the present invention, in order to achieve the above object, in the HD-MAC system which separates and processes a video signal into a low speed mode, a semi high speed mode, and a high speed mode, Determining a movement mode according to a movement degree of an image signal using a motion vector detected through motion estimation; Frame exchange means for exchanging a frame order of the filtered video signal in a low speed mode; Means for performing motion compensation interpolation using the motion vector and the video signal filtered in the quasi-fast mode; The video mode filtered in the high speed mode, the motion compensation interpolated video signal in the quasi-high speed mode, and the video signal in which the frame order in the low speed mode is exchanged are compared with the original video signal before filtering, respectively, to select the mode having the smallest error value. Characterized in that it comprises a means for determining as.

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

3 shows an encoder of an HD-MAC system with a mode determiner according to the present invention.

The protector of the HD-MAC system uses three luminance coding paths. At this time, the three mode paths are a low speed mode path for realizing high resolution in a still area, a quasi-high speed mode motion compensation path for up to 12 pixels per frame, and a case of fast moving or sudden change of screen. Fast mode path used.

The low pass filters 10, 12, and 14 connected corresponding to the respective paths distinguish respective modes (low speed mode, quasi high speed mode, and high speed mode), and in the signal processing process by reducing the sample, which will be described later. In order to prevent overlapping, the frequency band is limited. The subsampling sections 16, 18, and 20, corresponding to the low pass filters 10, 12, and 14, are sampled at regular intervals in the current samples. Again, reduce the number of samples by reducing the number of data transferred.

The line shuffling sections 22 and 24 are connected on the low speed and high speed mode paths, respectively, and combine the samples of the two lines to form a line. In the quasi high speed mode path, interfield shuffling is used. .

The switching unit 26 selects either the output of the line shuffling unit 22, the line shuffling unit 24, or the output of the sample reduction unit 20 by the mode determination signal provided from the matching unit 32 to be described later. To the decoder side.

The motion estimation unit 28 estimates motion for each block on the quasi-fast mode path, and uses a block matching algorithm (hereinafter referred to as BMA) for this purpose. The motion estimation is performed twice in four field sections to move the 16 × 16 block six pixels up, down, left, and right to find the position of the smallest error. At this time, unlike the conventional technique, motion estimation forms a frame for every field to obtain a motion vector using an unconditional search BMA. At this time, the motion vector detected by the motion estimation is transmitted through the DATV channel.

The mode determiner 30 determines the mode for each block by increasing the MSE value in the speed portion not included in the speed range of the path mode in each path.

In this case, the speed range is not quantitatively indicating the speed range suitable for each mode, but it means that the block has the speed component corresponding to the corresponding path mode when the error value is preempted in the smallest mode. . Therefore, increasing the MSE value means that the MSE value is increased by using a method such as frame exchange to increase an error in calculating the MSE of the low speed mode in the motion region, for example.

In the present invention, the still mode path, the filtered image, and the filtered image of the quasi-movement path undergo a preprocessing process in order to prevent any determination error due to the comparison of only the filtering error. Through this process, the part not included in the speed range of the path mode increases the MSE value to induce accurate mode determination.

In the mode determiner 30, the frame exchanger 38 changes the order of fields 1 and 2 and fields 3 and 4 as preprocessing of the low speed mode path. For example, the order of 1,2,3,4 fields is changed to 3,4,1,2 fields. In addition, the motion compensation interpolator 36 performs the motion compensation interpolation in each field section based on the motion vector provided by the motion estimation unit 28 as a preprocessing process of the quasi-fast mode path to generate an image.

In the mode determiner 30, the error calculation and comparison unit 40 inputs the image of the low speed mode path and the quasi high speed path, which have undergone preprocessing, and the filtered image of the high speed mode path, and receives the error in units of the original image and blocks. Calculate and output the mode decision signal.

The matching unit 32 provides the switching unit 26 with a mode determining signal provided by the error calculating and comparing unit 40, that is, a decision signal for selecting one of a low speed mode, a quasi high speed mode, and a high speed mode path. The selection of the switching section 26 is controlled. In this case, when the mode decision signal is a quasi-fast mode decision signal, the mode determiner 30 provides the motion vector detected by the motion estimator 28 to the matcher 32, and the matcher 32 provides the motion vector with the DATV channel. It is provided to the DATV 34 to be transmitted through.

In this case, the mode determination of the mode determiner 30 will be described in more detail as follows.

The low speed mode, for example, the path image of 80 kHz undergoes a frame exchange process in the frame exchanger 38. For example, the fields 1, 2, 3, and 4 are changed to 4, 3, 1, and 2 fields, and are input to the error calculation and comparison unit 40. If you change the position of the frame, the still area does not change much, but the moving area is significantly different from the original image. Therefore, when the error calculation and comparison unit 40 calculates an error by comparing the raw field and the field in which the position is exchanged, the MSE value is increased only in the moving part, thereby preventing the block of this part from being determined as the low speed mode.

The motion compensation interpolation technique is used for the path image in the quasi-high speed mode, for example, 40 ms mode. That is, the motion compensation unit 36 generates a motion compensated image in every field section, extracts each field component from the motion compensated image, calculates an error with the raw field, and uses the error value for mode determination. In this case, since the real motion vector cannot be obtained in the fast motion region (for example, the fast mode portion), the MSE value increases. Therefore, it is possible to prevent the fast movement area from being determined to the quasi-high speed mode by this error.

That is, when the motion estimation is performed by the block matching algorithm in the fast motion region, the speed of the motion exceeds the search region, and the motion vector used for the motion compensation is not the actual motion vector. The error is large when detected.

Also, in high speed mode, for example, 20㎳ mode, each field is filtered and compared with the raw image, and the error value is used to determine the mode.

The mode determiner of the HD-MAC system according to the present invention can determine the correct mode by increasing the MSE value in the speed portion not included in the speed range of the path mode in each path image.

Claims (2)

In the HD-MAC system that separates and processes the video signal into the low speed mode, the semi high speed mode, and the high speed mode, the motion vector detected through the motion estimation in the quasi high speed mode and the filtered video signal for each mode are processed. Frame exchange means for determining a movement mode according to the movement degree of the video signal by using the image signal, and for exchanging the frame order of the video signal filtered in the low speed mode; Means for performing motion compensation interpolation using the motion vector and the video signal filtered in the quasi-fast mode; The video mode filtered in the high speed mode, the motion compensation interpolated video signal in the quasi-high speed mode, and the video signal in which the frame order in the low speed mode is exchanged are compared with the original video signal before filtering, respectively, and the mode having the smallest error value is obtained. The mode determiner of the HD-MAC system comprising means for determining. The mode determiner of claim 1, wherein the movement mode is determined in units of blocks.