JP2002064748A - Telop area detector in moving image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a telop area detector that can detect the appearance of a telop at a high-speed with high accuracy from compressed coded data themselves or information decoding only part of the coded data and extract a telop position in a frame. SOLUTION: A variable length decoding section 1 partially decodes only required information from coded data, and decoded information sets A, B, C are respectively fed to a time transition decision section 2, a telop position deciding section 3, and an appeared frame decision section 4. The time transition decision section 2 sets an area to be a telop area object on an I image. The telop position decision section 3 extracts a block, having a coded mode corresponding to the telop from the image. The appeared frame decision section 4 conducts frame decision processing denoting from which frame the telop appears.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for detecting a telop region in a moving image, and more particularly to a moving image capable of extracting a telop region with high speed and high accuracy from information obtained by decoding compressed encoded data itself or only a part thereof. And a telop area detection device.

[0002]

2. Description of the Related Art As a conventional method of detecting a telop area (hereinafter, referred to as a first detection method), particularly a method of detecting a telop from a news video, a method of using a temporal luminance distribution difference value to determine telop detection is used. is there. In addition, a method of obtaining a telop from geometrical properties such as locality of an appearance position and a regular arrangement, and a method of obtaining a telop from character feature amounts such as edge distribution deflection and color similarity have been reported. Further, as a method of analyzing these feature amounts, a telop region detection method incorporating a neural network or a genetic algorithm has been proposed. In these methods, a telop area is extracted by applying various algorithms to the luminance value of each pixel in a moving image.

In another telop detection system, there has been proposed a system in which coded data of a moving image which has been compression-encoded is used instead of using pixels themselves. This method achieves a telop area detection process by directly operating various parameters and encoded data required at the time of compression.

A conventional method (hereinafter, referred to as a second detection method) that focuses on a temporal change of a motion prediction error observes a temporal variation of motion prediction information for a code unit block in an image, and instantaneously generates a telop. Is detected. Only the necessary minimum information is selected and the telop area is extracted. For this reason,
The extracted resolution is set to the size of the coding unit block of each compression method having motion prediction information.

[0005] Further, a method focusing on the coding mode of a code unit block (hereinafter, a third detection method) is based on the correlation between the characteristics of a stationary telop and the coding mode, and increases or decreases the count counter in accordance with the coding mode. Let it. An area having a threshold value or more is recognized as a telop area candidate, and the shape is determined to extract a telop. References disclosing the third detection method include, for example, IEICE Transactions D-II, Vol.J81-D-IINo.8, PP1847-1855, August 1998, “MPEG Fast Telop Region Detection Method "
There is.

[0006]

In order to apply the first detection method to moving image data that has been compression-encoded, it is necessary to decode the encoded data once, and the compression-encoded data is converted to the information of the pixel area. The telop detection process cannot be applied unless it is returned to. Therefore, the first detection method is realized by, for example, the configuration shown in the block diagram of FIG.

[0007] The variable length decoding unit 51 shown in FIG. 10 receives compressed image coded data as input and performs variable length decoding, decoding such as inverse quantization and coefficient range limitation. The encoding mode information of the decoded frame or block from the variable length decoding unit 51,
Motion prediction information, motion prediction error information, and the like are output. The input of the image generation unit 52 is prediction error information from the variable length decoding unit 51, and generates image data of one frame through inverse transformation. Further, the motion compensating unit 52 extracts a block from the image memory 53 holding a reference frame for motion prediction using the motion prediction information, and outputs the block with the output image data of the image generating unit 52 to form one complete frame. to add. This image is input to the telop detection unit 54, and if the frame is a reference image for the next frame and subsequent frames, the image memory 5
3 is stored. After these series of decoding processes, the telop detection unit 54 performs the telop region detection process in the pixel region for the first time. The detection result by the telop detection unit 54 is sent to the image display unit 55.

In the first detection method, there is a problem that a large calculation cost is required for the decoding processing of the compressed data and the telop area detection processing using the decoded image.

On the other hand, since the second and third detection methods use the compressed and coded data itself, the decoding process can be omitted and the detection process can be executed at high speed. However, in actual moving images, the motion prediction error information changes greatly due to factors such as camera work such as pan and tilt and video effects edited after shooting such as wipe and dissolve, and it is determined that a telop appears. Is difficult. In particular, this effect is significant in a scene change, and the second detection method has a problem that the detection accuracy is poor, such as a frame after the scene change is erroneously recognized as a telop area.

The third detection method has a characteristic that the detection rate is increased when a large number of motion vectors exist outside the telop area. However, in a low-resolution video, the motion vector becomes relatively small, and the distribution of the encoding modes differs. In addition, a news video or the like often has a fixed camera, and when there is no motion vector in the background, an area other than a telop may be erroneously detected as a telop area.

SUMMARY OF THE INVENTION An object of the present invention is to solve the above-mentioned problems of the prior art, and to quickly and accurately detect the appearance of a telop from compressed coded data itself or information obtained by decoding only a part of the coded data. An object of the present invention is to provide a region detecting device and a telop region detecting device capable of extracting a telop position in a frame.

[0012]

SUMMARY OF THE INVENTION In order to achieve the above-mentioned object, the present invention provides a moving image which receives compressed moving image data as input, adds telop area information to the moving image data, and outputs the moving image data. In the telop area detection device within, a time change determination unit that detects the appearance of the telop from the temporal change of the motion prediction error information, a telop position determination unit that extracts the position of the telop based on the type of encoding mode information, The feature is that an appearance frame determination unit that detects an appearance frame of a telop from a transition of the reference direction of the motion prediction information is provided.

According to this feature, the process of detecting the telop area is made stepwise, so that the telop area can be detected at a high speed. In addition, since the temporal change of the motion prediction error information, the type of the encoding mode information, and the transition of the reference direction of the motion prediction information are used to detect the telop region, highly accurate telop region detection processing can be performed. Become like

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention. Note that this embodiment adopts an international standard of MPEG-1 video (ISO / IEC11172-
Although 2) is used, the present invention is not limited to this.

[0015] Encoded data of a compression-encoded moving image is provided as an input to the entire system. Only the necessary information is partially decoded from the coded data by the variable-length decoding unit 1, and the decoded information A, B, and C are used as a time shift determining unit 2, a telop position determining unit 3, and an appearance frame determining unit, respectively. Sent to the section 4. Here, the information A is (n−
N) The coding information of the intra-frame coded image (I picture) between the frame and the (n + mN) frame, B is the coding mode information of the P and B pictures between the n frame and the (n + mN) frame, and C is ( (n-N) motion prediction information of a B picture between a frame and an n-th frame. Here, the constant N represents the number of pictures in the GOP, and is a parameter indicating the appearance interval of the intra-frame coded image. Further, n and m are arbitrary positive integers.

The time transition judging section 2 examines the state of temporal transition based on the coded information A of a plurality of intra-frame coded images. Then, based on the examination, an area that is a telop area candidate is set on the I picture, and the position information D of this area is output to the telop position determination unit 3.

The telop position determination unit 3 determines a telop determination target block based on the position information D from the time transition determination unit 2. At the same time, the coding mode information B of the target block is input from the variable length decoding unit 1, and the selection status of the coding mode is grasped for each determination target block. Based on these results, a block having a coding mode suitable for the telop is extracted from the I picture, and the position information E of the block is sent to the appearance frame determination unit 4.

The appearance frame determination unit 4 determines a detection target block based on the position information E of the block input from the telop position determination unit 3, and simultaneously receives the motion prediction information C from the variable length decoding unit 1. Based on the motion prediction information C, frame determination processing is performed to determine from which frame the telop appears.

The determination result F is output to a detection result display unit or a recording unit (not shown). The detection result display unit or the recording unit outputs the detection result and, if requested, partially decodes and presents the video of the extraction area.

Next, the function of the time transition judging section 2 will be described in detail with reference to the flowchart of FIG. The time change determination unit 2 performs four processes. These four processes are
These are a line variation determination process in step S1, a block variation determination process in step S2, a convergence determination process in step S3, and a shape shaping process in step S4.

The line fluctuation determination processing (step S1)
Performs processing on coded data of two input (Intra) coded images In-N and In. Here, the constant N represents the number of pictures in the GOP, and is a parameter indicating the appearance interval of the intra-frame coded image. In the line variation determination process, an area where a change has occurred with the passage of time is extracted for each line of a block. Step S1
The position information of the area determined to be a fluctuation area in the processing of (1) is input to the block fluctuation determination processing in step S2. In the block variation determination process, a variation region is determined in block units, and the extracted block is sent to the convergence determination process in step S3.

In the convergence determination processing, a region where the change converges in all frames In + mN (0 <m <q) after In with respect to the designated region is extracted, and the region is set as a telop region candidate. . Here, the constant q represents the number of frames for which convergence is to be determined, and takes a value of 1 or more. The shape shaping process in step S4 receives a telop area candidate and eliminates isolated small areas in consideration of the telop size. Next, the missing portion of the telop area candidate is compensated by the expansion / contraction processing, and the telop area candidate is shaped for the next and subsequent determination processing. With the above, the processing of the time change determination unit 2 ends.

The time transition determination unit 2 focuses on a region in which a change has been recognized by comparing the current intra-frame coded image with the immediately preceding intra-frame coded image, and only that region is included in the future frame. Consider whether the change converges in the encoded image. To determine the occurrence of a specific time change, D
The DC component and the AC component of the CT coefficient are individually used as criteria.

Hereinafter, the processing in steps S1 to S3 will be described in more detail with reference to FIGS.

First, with reference to FIG. 3, a description will be given of the line fluctuation determination processing in step S1 of FIG. FIG. 3 (a)
It is a flowchart which shows the detail of the line fluctuation determination processing by a DC component. In the line fluctuation determination processing, the frame I
n and DCT coefficient DC component information of the frame In-N one GOP before are input. First, assuming that the telop appears horizontally or vertically with respect to the screen, the DC component is
In step 10, the data is read in units of one line each in the vertical and horizontal directions. For example, as shown in FIG. 3B, the DC of each block of the frame In and the frame In-N one GOP before is shown.
The DC component of the T coefficient is read for each vertical and horizontal line.

In step S11, a coarsely quantized luminance histogram is generated in order to capture a rough change in the DC component. In step S12, the past frame In-
In N, the sum of absolute differences between the histogram of the same position line and the histogram is obtained. In step S13, a determination is made based on a threshold value, and in the line having a difference value equal to or greater than the threshold value, one line is determined as a telop area candidate in step S14. Otherwise, the entire one line is set as a non-telop area in step S15.

In step S16, it is determined whether or not processing for each line has been completed for all blocks. If not, the process returns to step S10, and a series of processing in steps S10 to S15 is repeated for the next line. If the determination in step S16 is affirmative, the process proceeds to step S17.

In step S17, the number of lines that have become telop area candidates is counted. If the telop area candidates occupy a large part of the frame, all the blocks are regarded as non-telop areas as luminance changes due to causes other than telop. The detection processing of the current frame ends. If not, the position information of the extracted block is output, and the time variation determination process based on the DC component ends.

Next, a description will be given, with reference to FIG. 4, of the block change determination processing in step S2. FIG.
9 shows a flowchart of a block variation determination process based on components. In the fluctuation determination process, the frame In and the frame In-
N DCT coefficient AC component information is input and processed in block units. DCT is the edge area where text and background are interwoven
A block in which a change in the partial sum of blocks exceeds a threshold both spatially and temporally is set as a telop area because it corresponds to the number of coefficient AC components.

In step S19, it is checked whether or not the target block is an area determined as a telop area candidate in the line variation determination processing in step S1. If it is a telop area candidate, the process is continued; if not, a non-telop is set, and the determination for the block is terminated. Step S20
Calculates the partial sum of the absolute value of the AC component for the telop area candidate.

FIG. 4B shows an example of a coefficient range used for determining fluctuation. Here, the DCT coefficient AC component has an AC low-frequency component 9 in a zigzag scan order.
Judgment using partial sums of absolute values is used.

In step S21, the past frame (I
The difference between the absolute value partial sum of nine AC low frequency components of the same position block in the (nN frame) is calculated. In step S22, a determination is made based on a threshold, and blocks having a difference value equal to or greater than the threshold are determined as telop area candidates in step S23. Otherwise, it is determined as a non-telop area candidate in step S24. A step S25 decides whether or not the processing has been completed for all the blocks, and if not completed, returns to the step S19 and repeats a series of the processing of the steps S19 to S24. If it is determined in step S25 that the processing of all blocks has been completed, the fluctuation determination processing based on the AC component ends.

Next, details of the convergence determination processing in step S3 will be described with reference to FIG. FIG. 5A shows a flowchart of the convergence determination process. It is highly probable that the fluctuating area is the appearance of a telop, but it cannot be denied that it is fluctuating due to a moving object or camera work. The telop has a large temporal fluctuation in luminance during the transition period, but in the steady state, the luminance hardly changes. Therefore, in the convergence determination process, the stationarity with respect to the position of the telop is used in order to remove a fluctuation region due to a factor other than the appearance of the telop among the telop region candidates extracted in steps S1 and S2.

More specifically, after confirming that there is no scene change or the like based on the temporal change of the DC component for the entire screen, the direction and position of the edge of the stationary telop with respect to the telop area candidate are the same. Take advantage of that. For the coincidence of edges, a class classification using a partial sum of AC components is used. For example, when the nine AC components shown in FIG. 4B are further divided into three parts of vertical (vertical), horizontal (horizontal), and diagonal elements, as shown in FIG. Thus, a total of 12 classes can be formed.

In the convergence determination process in step S3, DCT coefficient AC component information from frame In to frame In + qN is input. In step S26, it is checked whether or not the target block is a telop area candidate. If it is a telop area candidate, the processing is continued. Otherwise, the convergence determination processing for the block is terminated after setting a non-telop area. A step S27 decides the edge class of the target block in the frame In. For example, an edge class having the maximum partial sum is obtained from the twelve classes shown in FIG. In step S28, the frame In + mN (0 <
The same edge class of the same position block in m <q) is determined. Generally, intra-frame coded images are often arranged at intervals of 12 to 15 frames, so if 30 fps, they are arranged at intervals of about 0.5 seconds.
Assuming that the telop is presented for 2 seconds or more, the value of q can be set to about 4.

In step S29, the process proceeds to step S27.
And whether the partial sums of the edge classes obtained in S28 match. If they match, the block is set as a telop area candidate in step S30. Otherwise, it is determined as a non-telop area candidate in step S31. Step S
32 judges whether the processing has been completed for all blocks, and if not completed, returns to step S26 and repeats the series of processing of steps S26 to S31. If the processing of all blocks has been completed, the convergence determination processing based on the class classification ends.

By the above-described time transition determination processing, an area serving as a telop area candidate has been set on the I picture.

Next, referring to FIG. 6, the function of the telop position judging unit 3 (see FIG. 1) based on the encoding mode information will be described. The telop area determination unit 3 receives the telop area candidate information extracted by the time change determination unit 2.
At the same time, the frame In and the frame I
The coding mode information of the frames (P, B pictures) existing between n + mN is input. Over multiple frames,
Processing is performed with the block group located at the same position as one unit.
Verification of the telop based on the encoding mode information is limited to the telop area candidates extracted in the above-described detection processing.

In step S33, it is determined whether or not the target block has been determined as a telop area candidate by the time transition determining unit 2. If it is a telop area candidate, the process is continued; if not, a non-telop is set, and the determination for the block is terminated. In step S34, a count counter is generated from the encoding mode information and the inter-frame distance referred to by the motion prediction information. In step S35, a determination based on a threshold value is performed on the counter counted in step S34. The block group forming the counter having the threshold value or more is determined in step S
At 36, a telop area candidate is set. Otherwise, it is determined as a non-telop area candidate in step S37. Step S
38 judges whether the processing has been completed for all the blocks, and if not completed, returns to step S33 and repeats a series of processing of steps S33 to S37. If the processing for all blocks has been completed, the processing moves to step S39. A step S39 performs a shape shaping process on the telop area candidate. The processing content is the same as that of the shape shaping processing section (Step S4). Thus, the determination process based on the encoding mode information ends.

As the coding mode information, which is one of the input information, frame coding information and block coding information are used. There are the following three types of frame coding information. (1) Intra-frame coded image (I picture) (2) Forward predicted image (P picture) (3) Bidirectional predicted image (B picture)

In the block coding mode, there are an intra-frame coding block (Intra) and an inter-frame coding block (Inter). Further, there are the following four types of inter-frame coded blocks based on the presence or absence of motion compensation and coding. (1) Motion compensation coding block (MC coded) (2) Frame difference coding block (no MC coded) (3) Motion compensation block (MC no coded) (4) Skipped block (Skip) There are three types of forward, backward, and both directions.

At this time, there is a high correlation between the feature of the still telop and the coding mode of the corresponding block. For example, there is no motion prediction information in a static telop, or its size is close to zero. Also, as described in the character string constituting the telop, the edge of the boundary portion is cited as a feature of the character string, the motion prediction error information is rarely omitted due to the presence of a complicated texture.

Therefore, no MC coded having the above characteristics
The encoding mode is most likely to be a telop. However, when the encoding process of the actual video is considered, the motion prediction information is not provided to the still region as the accuracy of the encoder is improved. That is, a mode that does not have motion prediction information is often selected regardless of the presence or absence of a telop. Also, the closer the reference frame for motion prediction is, the smaller the apparent motion is, even in a moving area, so that motion prediction information may not be assigned.

As a method for solving this problem, a concept of a temporal distance is introduced when making a determination based on an encoding mode.
It is used as encoding mode reliability information. If the reference frame is close, even if motion prediction information does not exist, the reliability of the encoding mode information is set low because it does not guarantee the possibility of being a telop. Conversely, if motion prediction information exists, it is assumed that a clear moving object exists, and the reliability as a non-telop area candidate is increased. On the other hand, when the reference frame is far away, the reverse setting is prepared. That is, when there is no motion prediction information, it can be determined that the area is completely still, so that the reliability of the encoding mode information as a telop area candidate is set high. Even if motion prediction information exists, the reliability of the encoding mode information as a non-telop area candidate is set low.

An example of the counting method based on the coding mode information using the temporal distance as the reliability information (step S34) will be described with reference to the flowchart of FIG. The input information is the encoding mode information of the block group at the same position.

In step S40, the distance to the frame referred to by the motion prediction information is calculated. However, when the prediction method is bidirectional prediction, the one with the shorter distance is adopted. A step S41 makes a decision using the coding mode information of the block. MC whose block coding type has motion prediction information
If it is coded, MC no coded, in step S42, a number inversely proportional to the temporal distance to the reference frame obtained in step S40 is subtracted. On the other hand, does not have motion prediction information
Intra, no MC coded where the size of motion prediction information is 0
, Or Skip, a number proportional to the temporal distance is added in step S43. A step S44 decides whether or not the processing has been completed for all the blocks located at the same position. If not, a series of operations from the step S40 is repeated for the next block. If not, the counting of the counter ends.

By the above telop position determination processing, a block having an encoding mode suitable for a telop can be extracted from an I picture.

Next, the operation of the appearance frame determination section 4 will be described. The verification of the telop based on the motion prediction information is performed only for the telop area candidates extracted in the above-described detection processing. Here, the coding mode of the block and the temporal reference direction of the motion vector are used. When a telop appears, the following properties appear in the telop area of a continuous B picture (hereinafter referred to as a B picture group) divided into I or P pictures. (1) There is no bidirectional prediction in the B picture group. (2) When the appearance frame is an I or P picture, only a forward motion vector exists in the B picture group. (3) When the appearing frame is a B picture, a backward motion vector also exists in the B picture group. (4) When the appearing frame is a B picture, there is only one switching in the B picture group in the forward / reverse direction. (5) After the telop appears, it has no motion vector.

FIG. 9 is shown to facilitate understanding of the above properties. From FIGS. 7A and 7B, when a telop appears in a B picture, an I picture, or a P picture, it is clear that there is no bidirectional prediction in the B picture group as in (1). Further, when the appearance frame of the telop is an I or P picture, from FIG.
It is obvious that only the forward motion vector exists in the B picture group as shown in (2). Also, when the telop appearing frame is a B picture, it is clear from FIG. 10A that the B picture group has the above-mentioned properties (3) and (4). As shown in FIG. 3C, when bidirectional prediction exists in the B picture group, no telop appears in any of the I, P, and B picture groups.

Therefore, the appearance frame of the telop is narrowed down to the frame satisfying the above conditions (1) to (5) from the motion prediction information of the B picture group. Note that both sides of the B picture group in FIG. 9 may be both P pictures.

More specifically, when the time transition judging unit 2 detects the appearance of a telop in a plurality of intra-frame encoded images, the appearing frame judging unit 4 sets the above-mentioned characteristic (property) for the group of B pictures in the GOP. Verify First, a temporal reference direction of motion prediction information is checked for each block for each B picture group. When the group of B pictures is composed only of forward prediction (the property 2), it is determined that a telop has appeared in the immediately following I or P picture. When the group of B pictures includes the backward direction, or when there is only one change from the forward direction to the backward direction (the above properties 3 and 4),
It is determined that a telop has appeared in the B picture in which the reverse direction has started. In other cases, it is determined that no telop appears in this continuous B picture group, and the determination process is continued for the next B picture group. This makes it possible to detect the appearance frame of the telop in frame units.

However, since a stationary telop is assumed, the length of the motion prediction information itself is limited to a block having almost zero. Blocks in which the motion prediction information has a significant length are excluded from the telop area candidates regardless of the reference direction. Similarly, for the GOP after the telop appearance determination,
The length of the motion prediction information is verified for each block at the same position. If the length is not sufficiently close to 0, the block is excluded from the telop area candidates.

FIG. 8 is a flowchart showing the operation of the appearance frame determination unit 4 based on the motion prediction information. The telop position determination unit 3 receives position information of a telop area candidate. At the same time, the motion prediction information of the frame between the frame In-N and the frame In is input from the variable length decoding unit 1. The determination is made for each block of the B picture group.

In step S45, it is determined whether or not the target block has been determined by the telop position determination unit 3 as a telop area candidate in the frame In. If it is a telop area candidate, the process is continued; otherwise, the determination for the block group is terminated. In step S46, the characteristics of the B picture group accompanying the appearance of the telop described above are verified (see the characteristics (1) to (5) and FIG. 9). Step S47
Then, it is determined whether or not the output of step S46 satisfies the above property. If so, the B picture frame in which the backward vector appears or the I or P picture frame in step S48 is determined. Output as the appearance frame of the telop. Otherwise, in step S49, the block is set as a non-telop, and it is determined that no telop appears in the B picture group. A step S50 decides whether or not the decision processing has been completed for all blocks of the B picture group. If the processing has not been completed, the process returns to step S45, and the series of steps S45 to S49 is repeated for the next block. Otherwise, the determination process is continued.

In step S51, all Bs in the GOP are
It is determined whether the determination processing has been completed for the picture group.
If the processing has not been completed, a series of steps S45 to S50 is repeated for the next B picture group. Otherwise, the process ends. If no telop start frame is detected in step S48, all blocks are set as non-telop areas, and the process of the appearance frame determination unit 4 ends. In this case, the appearance frame determination unit 4 then performs the determination processing of FIG. 8 again on all the B pictures in the next GOP.

As apparent from the above description, according to the present invention, the following features (1) to (5) can be provided. (1) Since the step of detecting the telop area is made stepwise, both high-speed processing and high-accuracy processing can be achieved. (2) Since the temporal variation determination and the subsequent convergence determination are performed, unnecessary variation regions can be eliminated, and the telop region detection processing can be performed. (3) Blocks having significant motion prediction information can be excluded from detection targets. (4) In consideration of the reliability of the encoding mode information, it is possible to perform the detection determination by the weighting counting. (5) The telop detection resolution in one frame unit can be achieved using the motion prediction information.

[0057]

As is apparent from the above description, according to the present invention, in addition to partially decoding the compression-encoded moving image data, in addition to the frames at intervals of ten or more frames (for example, The detection process of the telop start frame is made hierarchical such that detection is first performed for the intra-frame coded image, and then detection is performed in units of one frame. Needless to say, the first detection method can reduce the processing cost as compared with the detection method in the code data area (the second and third detection methods). That is, according to the present invention, the applicable range of the telop detection determination can be suppressed to a necessary minimum, so that it is possible to further reduce the processing amount and the speed of the telop region extraction method on the compression-encoded data. Become.

Also, in the present invention, the two properties of the precursor (fluctuation) and the stationarity (convergence) after the appearance of the telop are determined by different discriminants, respectively. It is possible to achieve much better detection accuracy than the detection method.

Further, according to the present invention, the accuracy of a block serving as a telop area candidate is improved by using the encoding mode information, and the telop start frame is obtained by using the motion prediction information. It becomes possible.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of an embodiment of the present invention.

FIG. 2 is a flowchart illustrating an operation of a time transition determining unit in FIG. 1;

FIG. 3 is a line variation determination process of FIG. 2 (step S1);
FIG. 4 is a flowchart and an explanatory diagram showing details of FIG.

FIG. 4 is a block change determination process (step S) of FIG.
It is a flowchart and explanatory drawing which show the detail of 2).

FIG. 5 is a flowchart and an explanatory diagram showing details of a convergence determination process (step S3) of FIG. 2;

FIG. 6 is a flowchart showing an operation of a telop position determination unit in FIG. 1;

FIG. 7 is a flowchart illustrating details of a weighted encoding mode counting process (step S34) in FIG. 6;

FIG. 8 is a flowchart showing an operation of an appearance frame determination unit in FIG. 1;

FIG. 9 shows an appearance sign verification process (step S4) in FIG.
It is explanatory drawing of 6).

FIG. 10 is a block diagram showing a configuration of a conventional first detection method.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Variable length decoding part, 2 ... Time transition judgment part, 3 ... Telop position judgment part, 4 ... Appearance frame judgment part.

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5C023 AA18 BA01 CA05 5C059 KK00 MA04 MA05 MA23 MC32 MC34 NN21 PP05 PP06 PP07 PP20 PP24 PP26 UA05 5L096 BA18 FA26 FA34 FA35 FA44 HA03

Claims (20)

[Claims] 1. Compressed moving image data is input,
A telop area detecting device for adding a telop area information to the data of the moving image, and outputting the telop area information in the moving image; A telop position determining unit for extracting the position of the telop based on the type of the telop, and an appearance frame determining unit for detecting the appearance frame of the telop from the transition of the reference direction of the motion prediction information. Telop detection device. 2. The apparatus for detecting a telop area in a moving image according to claim 1, wherein the time change determination unit uses a difference value of motion prediction error information between frames as a criterion as a telop appearance determination. A telop area detection device characterized by the above-mentioned. 3. The apparatus for detecting a telop area in a moving image according to claim 2, wherein the time transition determining unit determines a histogram based on a DCT coefficient DC component of motion prediction error information as a difference value between frames used for a variation determination. A telop area detection device using a difference. 4. The apparatus for detecting a telop area in a moving image according to claim 2, wherein the time transition determining unit uses a histogram difference of the motion prediction error information for each line of the block in each of the vertical and horizontal directions. Telop area detecting device. 5. The apparatus for detecting a telop region in a moving image according to claim 2, wherein the time transition determining unit determines a DCT coefficient AC component of motion prediction error information as a difference value between frames used for a variation determination. A telop area detection device using a difference based on a sum of absolute values. 6. The telop area detection device according to claim 1, wherein the time change determination unit includes a unit that checks a steady state after the appearance of the telop. 7. The telop area detection apparatus according to claim 6, wherein the time change determination unit uses identity determination based on class classification to confirm a steady state. apparatus. 8. The apparatus for detecting a telop area in a moving image according to claim 6, wherein the time transition determining unit determines whether a class used for grasping a steady state is unevenly distributed in a coefficient distribution of motion prediction error information. And a telop area detecting device. 9. The apparatus for detecting a telop area in a moving image according to claim 6, wherein the time transition determination unit performs the motion prediction error information DCT coefficient AC to form a class used for convergence determination. A telop area detection apparatus characterized by using an edge direction in which a component is identified from three elements of vertical, horizontal and diagonal. 10. The apparatus for detecting a telop area in a moving image according to claim 1, wherein the telop position determination unit uses 0-approximation determination based on the size of motion prediction information for determination of a telop. A telop area detection device characterized by the above-mentioned. 11. The apparatus for detecting a telop area in a moving image according to claim 1, wherein the telop position determination unit uses a correlation between coding mode information for each block and a feature as a telop for telop position determination. A telop area detection device, characterized in that: 12. The apparatus for detecting a telop area in a moving image according to claim 1, wherein the telop position determination unit uses a correlation between coding mode information for each block and a feature as a non-telop for telop position determination. A telop area detecting device. 13. The apparatus for detecting a telop area in a moving image according to claim 11, wherein the telop position determination unit determines a reliability of the encoding mode information by using a time interval from a reference frame of the motion prediction information to a reference frame. A telop area detection device using a distance. 14. The apparatus for detecting a telop area in a moving image according to claim 11, wherein the telop position determination unit includes a weighting counter for determining a continuity of the encoding mode information. A telop area detection device characterized by using. 15. The method of claim 11, 12, 13, and 14.
In the apparatus for detecting a telop area in a moving image according to any one of the above, the telop position determination unit is characterized in that a weight coefficient of a weighting counter is proportional to a temporal distance to a frame referred to by the motion prediction information. Telop area detection device. 16. The apparatus for detecting a telop area in a moving image according to claim 1, wherein the appearance frame determination unit uses a temporal change in the direction of motion prediction information to determine an appearance frame of the telop. Characteristic telop area detection device. 17. The apparatus for detecting a telop area in a moving image according to claim 16, wherein the appearance frame determination unit determines the appearance frame of the telop by adding a bidirectional motion to a telop area in a continuous bidirectional prediction image. A telop area detecting device, wherein the absence of prediction information is used as a criterion. 18. The apparatus for detecting a telop area in a moving image according to claim 16, wherein the appearance frame determination unit determines that the appearance frame of the telop is such that the telop appears in the intra-frame encoded image or the unidirectional prediction image. A telop area detection device that uses, as a criterion, the presence of only forward motion prediction information in a telop area in a continuous bidirectional prediction image before that. 19. The apparatus for detecting a telop area in a moving image according to claim 16, wherein the appearance frame determination unit determines the appearance frame of the telop when the telop appears in the bidirectional prediction image. A telop area detection device, which uses, as a criterion, the presence of backward motion prediction information in a telop area in a direction prediction image. 20. The apparatus for detecting a telop area in a moving image according to claim 16, wherein the appearance frame determination unit determines the appearance frame of the telop when the telop appears in the bidirectional prediction image. A telop area detection device characterized in that a change of the direction of motion prediction information to a telop area in a direction prediction image is switched only once, as a determination criterion.