JP2008098734A - Motion image encoding device, program, and motion image encoding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve the encoding efficiency of a frame image that has a scene change. <P>SOLUTION: A motion image encoding device includes a frame-field converter which converts one frame image into two field images, a scene change detector which detects whether a scene change as a switching point of scenes is included between the two field images, and an image information encoder which encodes the two field images by using an in-picture prediction mode or an inter-picture prediction mode according to the detection result of the scene change detector. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a moving image encoding device, a program, and a moving image encoding method, and more particularly, to a moving image encoding device, a program, and a moving image encoding method that perform encoding using an intra prediction mode or an inter prediction mode. .

Currently, in various applications such as communication, broadcasting, and storage, a moving picture encoding method that compresses digitized moving picture information with high efficiency is widely used as a standard. Non-Patent Document 1 and Non-Patent Document 2 describe examples of standardized moving image encoding methods.
In encoding methods including Non-Patent Document 1 and Non-Patent Document 2, an image that is a unit of encoding is a frame image or a field image, and the frame image may be composed of two field images.

FIG. 13 is a block diagram showing a configuration of a conventional moving image re-encoding device 500. As shown in FIG. The moving image re-encoding device 500 includes a picture type determining unit 501, an encoded information analyzing unit 502, an encoded information decoding unit 503, a thinning unit 504, a video memory 505, an image information encoding unit 506, a motion vector synthesizing unit 507, A motion vector detection unit 508, an information buffer 509, a complexity calculation unit 510, a scene change detection unit 511, and a GOP structure determination unit 512 are provided.
The moving image re-encoding apparatus 500 performs the second encoded moving image information after thinning out the number of pixels in the horizontal and vertical directions and reducing the frame rate of the input first encoded moving image information. Re-encode and output.

The picture type determination unit 501 determines the picture type and discards the B picture for the first encoded moving picture information. The encoded information analysis unit 502 analyzes the first encoded moving image, acquires a motion vector, a quantization width, and a code amount and outputs them.
The encoded information decoding unit 503 decodes the first encoded moving image information and outputs a frame image. The thinning unit 504 thins and reduces the pixels of the frame image. The video memory 505 stores the input image.
The image information encoding unit 506 encodes an image using the second encoding method. The complexity described later is used to control the code amount. The motion vector synthesis unit 507 synthesizes a motion vector used for re-encoding from the motion vector used in the first encoded video information.
The motion vector detection unit 508 detects a more accurate motion vector based on the given motion vector. The information buffer 509 stores the quantization width and the code amount. The complexity calculation unit 510 calculates and estimates the complexity that is the product of the code amount and the quantization width.
The scene change detection unit 511 detects a scene change between frame images based on a variation amount of complexity. The GOP structure determination unit 512 selects a picture type for each frame and determines the GOP structure.
In this moving image re-encoding device 500, the encoded information decoding unit 503 decodes the first encoded moving image information input to the picture type determining unit 501, and configures the first encoded moving image information. Each frame image is re-encoded by the image information encoding unit 506 into the second encoded moving image information.

  Since the moving picture re-encoding apparatus 500 discards the B picture at the time of re-encoding, the use frequency of the I picture is relatively increased in the moving picture after re-encoding, and the encoding efficiency is lowered. End up. For this reason, the GOP structure determination unit 512 changes some frame images, which are I pictures, to P pictures with better encoding efficiency than the I pictures in order to prevent a decrease in encoding efficiency. However, when a scene change is detected immediately before the frame image to be re-encoded, it is determined not to change to a P picture, and image quality deterioration is reduced.

As described above, the image re-encoding device 500 according to the conventional technique described in Patent Document 1 reduces the image quality deterioration caused by the scene change between the frame images, and reduces the second from the first encoded moving image information. Re-encoding into the encoded moving image information.
Patent Document 2 discloses a technique for estimating a motion vector of a picture to be recoded using a motion vector in a neighboring picture of the picture to be recoded in recoding. Non-Patent Document 3 discloses a technique of a scene change detection method based on a dissimilarity between image differences.
JP 2002-152759 A JP 2000-165887 A ISO / IEC 13818-2 ISO / IEC 14496-10 JPO Standard Technology Collection “Video Shot Switching Detection Method” (http://www.jpo.go.jp/shiryou/s_sonota/hyoujun_gijutsu/bidirectional_video/111_15.htm)

  In the technique described in Patent Document 1, when a frame image is composed of two field images, a scene change may occur because there is a time difference between the field images.

FIG. 14A and FIG. 14B are diagrams for explaining a case where a scene change occurs in a moving image.
In FIG. 14A, the frame images change to 41, 42, and 43 with the passage of time. The frame image 42 includes a field image 421 composed of an odd line image of the frame image 42 and a field image 422 composed of an even line image of the frame image 42. The frame image 41 belongs to the scene A, and the frame images 42 and 43 belong to the scene B. A scene change that changes from the scene A to the scene B occurs between the frame image 41 and the frame image 42. Such a scene change is hereinafter referred to as an inter-frame scene change.

In FIG. 14B, the frame images change to 44, 45, and 46 with the passage of time. The frame image 45 includes a field image 451 composed of an odd-numbered image of the frame image 45 and a field image 452 composed of an even-numbered image of the frame image 45. The frame image 44 belongs to the scene A, the frame image 45 belongs to the scene A + B, and the frame image 46 belongs to the scene B. A scene change has occurred in the scene A + B of the frame image 45. Such a scene change is hereinafter referred to as an intra-frame scene change.
The frame images shown in FIGS. 14A and 14B are a part of the first encoded moving image information, and both the frame image 42 and the frame image 45 are I pictures.

In the prior art, when the frame image 42 (see FIG. 14A) and the frame image 45 (see FIG. 14B) for scene change are re-encoded, the entire frame image is used as an I picture frame image. Encoding. At this time, when viewed in units of field images, the field image 422 uses the intra-screen prediction mode without using the inter-screen prediction mode even though the correlation with the field image 421 is high.
On the other hand, the field image 451 uses the intra-screen prediction mode without using the inter-screen prediction mode even though the correlation with the frame image 44 is high. In either case, the conventional technique has a problem that the encoding efficiency deteriorates when re-encoding a moving image in which a scene change exists.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a moving image encoding apparatus, a program, and a moving image encoding method capable of improving the encoding efficiency of a frame image including a scene change. It is to provide.

The present invention has been made to solve the above problems, and the moving image encoding apparatus of the present invention includes a frame-field conversion unit that converts one frame image into two field images, and the two images. A scene change detection unit that detects whether or not a scene change that is a scene switching point is included between the field images, and each of the two field images is displayed on the screen based on the detection result of the scene change detection unit. And an image information encoding unit that performs encoding using the intra prediction mode or the inter-screen prediction mode.
In the present invention, the frame-field conversion unit converts one frame image into two field images, and determines whether or not a scene change that is a scene switching point is included between the two field images. Since the detection unit detects and the image information encoding unit encodes each of the two field images based on the detection result of the scene change detection unit using the intra prediction mode or the inter prediction mode, By encoding using the intra-screen prediction mode or the inter-screen prediction mode suitable for each of the two field images, the information amount can be reduced and encoded as compared with one frame image.

The image information encoding unit of the moving image encoding apparatus of the present invention encodes the frame image as two field images when the scene change detection unit detects a scene change.
In the present invention, when there is a correlation between two field images, the amount of information can be reduced and encoded compared to a single frame image.

Further, the moving picture encoding apparatus of the present invention has a GOP structure that determines a GOP structure by selecting an intra-screen prediction mode or an inter-screen prediction mode for each field image in accordance with a scene change detection result of the scene change detection unit. A determination unit, and the image information encoding unit encodes each field image using the prediction mode selected by the GOP structure determination unit.
In the present invention, since the intra-frame prediction mode or the inter-screen prediction mode is selected and encoded for each field image constituting the GOP, the information amount of the GOP is reduced by using the correlation between the field images constituting the GOP. Can be encoded.

Further, the GOP structure determination unit of the video encoding device of the present invention is configured such that the prediction mode used for the frame image converted by the frame-field conversion unit when the scene change detection unit detects a scene change. The combination of prediction modes used for the two field images is selected according to the above.
In the present invention, encoding is performed by reducing the amount of information compared to a frame image by selecting an optimal combination of prediction modes used for two field images in accordance with the prediction mode used for the frame image. Can do.

Further, the GOP structure determination unit of the moving picture coding apparatus according to the present invention, when the scene change detection unit detects a scene change, changes the prediction mode used for the field image before the scene change to the inter-screen prediction mode. And
In the present invention, by setting the prediction mode used for the field image before the scene change to the inter-screen prediction mode, the field image can be predicted from the other field images, and the information on the field image before the scene change is obtained. The amount can be reduced and encoded.

Further, the frame-field conversion unit of the moving image encoding device of the present invention, when the scene change detection unit detects a scene change, converts one field image of the two field images to the other. Conversion is performed so that the difference from the field image decreases.
In the present invention, one of the two field images is converted so as to reduce the difference from the other field image, so that one field image can be easily predicted from the other field image. Thus, it is possible to reduce the amount of information when encoding a frame image.

Further, the frame-field conversion unit of the moving image encoding device of the present invention, when the scene change detection unit detects a scene change, converts one field image of the two field images to the other. Convert to the same field image as the field image.
In the present invention, one of the two field images is converted into the same field image as the other field image, so that the information amount of one field image can be reduced to zero. The amount of information when encoding a frame image can be reduced.

In addition, the image information encoding unit of the moving image encoding device of the present invention, when the scene change detection unit detects a scene change, configures a frame image to be encoded and belongs to a field before the scene change. The average quantization width used for image encoding is increased more than the average quantization width used for encoding the previous field image in the same prediction mode.
In the present invention, the frame image is encoded by reducing the amount of information by configuring the frame image to be encoded and increasing the average quantization width used for encoding the field image belonging to the scene before the scene change. The amount of information at the time can be reduced.

In addition, the image information encoding unit of the moving image encoding device of the present invention includes a field that constitutes a frame image to be encoded and belongs to a scene after a scene change when the scene change detection unit detects a scene change. The average quantization width used for encoding the image is made smaller than the average quantization width used for encoding the field image that forms the frame image to be encoded and belongs to the scene before the scene change.
In the present invention, since the frame image to be encoded is configured and the average quantization width used for encoding the field image belonging to the scene after the scene change is reduced, the field image belonging to the scene before the scene change On the other hand, the information amount of the field image belonging to the scene after the scene change can be increased.

  In the program of the present invention, the computer includes a first step of converting one frame image into two field images, and a scene change that is a scene switching point between the two field images. A second step of detecting whether or not each of the two field images is encoded using an intra-screen prediction mode or an inter-screen prediction mode based on a detection result of the scene change detection unit; The steps are executed.

  The moving image encoding method of the present invention includes a frame-field conversion process for converting one frame image into two field images, and a scene change that is a scene switching point between the two field images. Each of the two field images is encoded using an intra-screen prediction mode or an inter-screen prediction mode based on a scene change detection process for detecting whether or not it is included and a detection result of the scene change detection unit. And an image information encoding process.

  In the present invention, it is possible to improve the encoding efficiency of a frame image including a scene change.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram showing an example in which two field images fld1 and fld2 constitute one frame image frm. Two field images fld1 and field image fld2 having a height h / 2 that is half the height h of the frame image frm are alternately stored so as to occupy the odd and even lines of the frame image frm by interlaced scanning. Has been. When taking such a structure, the time of the video of the field image fld1 and the field image fld2 is different. Note that the field image fld1 in FIG. 1 is an image ahead of the field image fld2 in the video time order and the coding order.

Frame images and field images as encoding targets are collectively called pictures, and pictures are encoded by being divided into small areas called macroblocks. At the time of encoding, it is general that a prediction encoding method, that is, a prediction mode and a quantization width that is a value for adjusting image quality and code amount can be selected for each macroblock. There are two types of prediction modes: an intra-screen prediction mode and an inter-screen prediction mode.
The intra prediction mode is a prediction mode that uses the correlation of pixel values in the spatial direction. The inter-screen prediction mode is a prediction mode that uses the correlation of pixel values in the time direction. A combination of prediction modes that can be selected for each macroblock is selected as a picture type for each picture. There are basically three types of picture types: I, P, and B.

A picture whose picture type is I is called an I picture, and only the intra prediction mode is used. A picture whose picture type is P is called a P picture, and in addition to the same intra prediction mode as that of an I picture, an inter prediction mode with one reference picture serving as prediction source data can be used. A picture whose picture type is B is called a B picture, and in addition to the same prediction mode as a P picture, an inter-screen prediction mode using two reference pictures can be used.
The P picture and the B picture that can use the inter-screen prediction mode encode only the difference from the previously coded picture. Therefore, the higher the correlation between the picture to be coded and the reference picture, the more efficient the picture. Is possible.
In the first to third embodiments described later, a case will be described in which a P picture is used as an inter-screen prediction mode and an I picture is used as an intra-screen prediction mode.

FIG. 2 is a diagram illustrating an example of correlation between pictures. FIG. 2 shows a case where the frame image changes to frm1, frm2, frm3, and frm4 with the passage of time.
Between pictures belonging to the same scene, there is a high correlation as long as the video of the scene does not change significantly, but when the reference picture and the picture to be encoded belong to different scenes, the correlation is low. More specifically, the frame image frm1 and the frame image frm2 are similar in that the circle pattern m1 moves within the frame image, and the correlation is high. Further, the frame image frm3 and the frame image frm4 are similar in that the square pattern m2 moves in the frame image, and the correlation is high. On the other hand, the frame image frm2 and the frame image frm3 have different patterns included in the frame image and have a low correlation.
Switching from one scene to another is called a scene change. Use of the inter-screen prediction mode between pictures with a scene change is generally low in coding efficiency and leads to deterioration of image quality, so use the intra-screen prediction mode. Is desirable. In FIG. 2, a scene change occurs between the frame image frm2 and the frame image frm3.

  On the other hand, the I picture that uses only the intra prediction mode is used for coding the picture immediately after the scene change for the above reasons, and realizes random access to a group of pictures (GOP) that is a collection of a series of pictures. Therefore, it is also used in encoding the first picture of the GOP. In the case of the conventional encoding method, when encoding is performed in units of field images, since GOP is configured in units of frame images, random access is also performed in units of frame images. However, if the field image encoded earlier in the frame image is not an I picture, it does not become the first picture of the GOP. Even if the field image encoded later is encoded as an I picture, the frame image Random access is not possible.

  By the way, a process for converting the first moving image information encoded by the predetermined first encoding method into the second moving image information encoded by the predetermined second encoding method is generally a moving image. This is called image re-encoding. Re-encoding is performed for the purpose of reducing the amount of code by using an encoding method with higher encoding efficiency, as well as converting the resolution and conforming to the specifications of the communication / broadcast / storage device. For any purpose, it is preferable to avoid image quality degradation due to re-encoding as much as possible.

(First embodiment)
FIG. 3 is a block diagram illustrating a configuration of the moving image re-encoding device 100 (also referred to as a moving image encoding device) according to the first embodiment of the present invention. The moving image re-encoding device 100 includes a motion vector synthesis unit 101, a frame-field conversion unit 102, an image information encoding unit 103, an intra-frame scene change detection unit 104, a GOP structure determination unit 105, a picture type determination unit 201, The encoded information analysis unit 202, the encoded information decoding unit 203, the video memory 205, the motion vector detection unit 208, the information buffer 209, and the complexity calculation unit 210 are provided.

The moving image re-encoding device 100 receives, as an input, first encoded moving image information having a frame structure according to the first encoding method, and an appropriate encoding parameter corresponding to a frame image including an intra-frame scene change. And the second encoded moving picture information having the field structure by the second encoding method is output.
The first encoding method is a method in which the intra-screen encoding mode and the inter-screen encoding mode can be selected for each picture. For example, a method specified in ISO / IEC 13818-2 is used. it can.
The second encoding method is a method that enables encoding in units of field images, and for example, a method defined in ISO / IEC 14496-10 can be used.
Note that the same encoding method may be used for the first encoding method and the second encoding method.

The motion vector synthesis unit 101 synthesizes the motion vector of the second encoding method from the motion vector in the first encoding method output from the encoding information analysis unit 202 and outputs the synthesized motion vector to the motion vector detection unit 208. The motion vector refers to a spatial shift when a frame image to be encoded next is predicted by referring to the encoded frame image.
The frame-field conversion unit 102 converts frame-structured image information output from the video memory into a field structure, and outputs the field structure to the frame-field conversion unit 102.
The image information encoding unit 103 encodes the field image output from the frame-field conversion unit 102 using the second encoding method, and outputs the encoded image to the outside of the moving image re-encoding device 100.
The intra-frame scene change detection unit 104 detects the presence / absence of an intra-frame scene change based on the field image output from the frame-field conversion unit 102, and sends a scene change flag as a detection result to the GOP structure determination unit 105. Output.
The GOP structure determination unit 105 selects a picture type based on the scene change flag output from the intra-frame scene change detection unit 104, determines the GOP structure, and outputs the picture type to the image information encoding unit 103.

  The picture type determination unit 201 determines the picture type of the frame image included in the first encoded moving image information input to the moving image re-encoding device 100, and discards the B picture. The picture type determination unit 201 outputs the first encoded moving image information to the encoding information analysis unit 202 and outputs the determined picture type to the GOP structure determination unit 105. Note that the picture type determination unit 201 does not discard the B picture when frame rate conversion is not performed. In this case, the encoded information decoding unit 203 is provided with a B picture decoding function.

The motion vector synthesis unit 101 converts the motion vector for the first encoding method output from the encoding information analysis unit 202 so as to match the motion vector in the second encoding method, and outputs the motion vector to the motion vector detection unit 208. To do. The motion vector synthesis unit 101 converts a motion vector for a frame image into a motion vector for a field image.
The frame-field conversion unit 102 interlaces and scans the frame image based on the relationship between the frame image and the field image, converts the frame image into two field images including only odd lines and even lines of the frame image, and encodes image information. Output to the unit 103.

  The image information encoding unit 103 encodes one frame image encoded using the first encoded moving image information as two field images using the second encoding method, and generates a moving image. Output to the outside of the re-encoding device 100. The picture types applied to the two field images are the picture type pt1 and the picture type pt2 selected by the GOP structure determination unit 105, respectively. In order to encode two field images, the amount of code is controlled for each field image. The complexity described later is used to control the code amount.

The in-frame scene change detection unit 104 detects whether or not a scene change is included in the two field images in the frame image by using a scene change detection method based on the dissimilarity of the inter-image difference. . The in-frame scene change detection unit 104 detects whether a scene change is included in the two field images using a method other than the scene change detection method based on the dissimilarity of the difference between images. May be.
Specifically, the in-frame scene change detection unit 104 determines that there is a scene change between two field images when the dissimilarity between the two field images is greater than a predetermined threshold, and the true The scene change flag sc is output to the GOP structure determination unit 105. On the other hand, if the dissimilarity between the two field images is equal to or less than a predetermined threshold, it is determined that there is no scene change between the two field images, and a false scene change flag sc is sent to the GOP structure determination unit 105. Output.

  The GOP structure determination unit 105 selects one of the two field images based on the picture type pt output from the picture type determination unit 201 and the scene change flag sc output from the intra-frame scene change detection unit 104. The picture type pt1 to be applied to and the picture type pt2 to be applied to the other field image are selected and output to the image information encoding unit 103.

FIG. 4 is a flowchart showing a process of the GOP structure determination unit 105 according to the embodiment of the present invention. First, the GOP structure determination unit 105 determines whether the picture type output by the picture type determination unit 201 (FIG. 3) is an I picture, that is, whether pt = I (step S10). .
When the picture type is I picture in step S10 (pt = I), the GOP structure determination unit 105 (FIG. 3) determines whether the scene change flag output from the in-frame scene change detection unit 104 is true or false. It is determined whether there is, that is, whether sc = true (step S11).

When the intra-frame scene change detection unit 104 detects a scene change (sc = true), the GOP structure determination unit 105 sets 2 depending on the picture type used for the frame image converted by the frame-field conversion unit 102. A combination of picture types used for one field image is selected. Specifically, when the scene change detection unit 102 detects a scene change, the GOP structure determination unit 105 sets the picture type used for the field image before the scene change as a P picture.
That is, when the scene change flag is true in step S11 (sc = true), the scene image is included in the frame image frm to be re-encoded, and there are two scenes in the frame image. is doing. In this case, when the inter-picture prediction mode is used for the field image fld1 using the temporally forward picture belonging to the same scene as a reference picture, the encoding efficiency can be improved. On the other hand, the field image fld2, which is the first image of the scene after the scene change, uses the intra-screen prediction mode because it is difficult to predict from the temporally previous picture.
That is, the GOP structure determination unit 105 selects a P picture as a picture type to be applied to the field image fld1, and selects an I picture as a picture type to be applied to the field image fld2 (pt1 = P, pt2 = I) (step S12). ).

In step S12, a P picture may be selected as a picture type to be applied to the field image fld2, and an intra prediction mode may be selected for each macroblock.
Here, since the frame image frm to be re-encoded is an I picture and does not have a motion vector, a new motion vector for the field image fld1 is required. For a motion vector that is insufficient, the motion vector detection unit 208 may newly detect a motion vector, or another method may be used. That is, the motion vector may be estimated from the motion vector of another picture.

  When the frame image frm that can be randomly accessed in the first encoded moving image information becomes an image that cannot be randomly accessed after re-encoding when the process of step S12 is performed, the field image fld1 By selecting the I picture as the picture type to be applied to, random access to the frame image frm is possible even after re-encoding. In this case, a third embodiment described later may be used in order to avoid a decrease in encoding efficiency due to the fact that both of the picture types applied to the field images fld1 and fld2 are I pictures.

If the scene change flag is false in step S11 (sc ≠ true), no scene change is included between the field image fld1 and the field image fld2, and a high correlation is present between the field image fld1 and the field image fld2. Therefore, if the field image fld2 is predicted from the field image fld1, the encoding efficiency can be increased.
Therefore, the field image fld1 is an I picture, and the field image fld2 is a P picture without inheriting the picture type I in the first encoded moving picture information (pt1 = I, pt2 = P). That is, the GOP structure determination unit 105 selects an I picture as a picture type to be applied to the field image fld1, and selects a P picture as a picture type to be applied to the field image fld2 (step S13).

  If the picture type is not an I picture in step S10 (pt ≠ I), the GOP structure determination unit 105 continues the picture type in the first encoded moving picture information as the picture type applied to the field image fld1. The picture type in the first encoded moving picture information is continuously selected as the picture type to be applied to the field picture fld2 (pt1 = pt, pt2 = pt) (step S14).

Next, an operation example of the GOP structure determination unit 105 will be described with reference to FIGS. 5 and 6.
FIG. 5 is a schematic diagram of first encoded moving image information according to the first embodiment of the present invention. In FIG. 5, the frame images change as I0, P1, P2, I3, P4, P5, I6, P7, and P8 with the passage of time. The frame images I0, I3, and I6 have an I picture applied as a picture type, and can be randomly accessed. In addition, a P picture is applied as a picture type to the frame images P1, P2, P4, P5, P7, and P8.
The frame images I0, P1, and P2 constitute GOP · g1, which is a collection of frame images. Similarly, the frame images I3, P4, and P5 constitute GOP · g2, and the frame images I6, P7, and P8 constitute GOP · g3.

FIGS. 6A to 6C are schematic diagrams of second encoded moving image information according to the first embodiment of the present invention.
In FIG. 6A, pictures corresponding to the field image fld1 are denoted by I0t, P1t, P2t, I3t, P4t, P5t, I6t, P7t, and P8t, and pictures corresponding to the field image fld2 are denoted by P0b, P1b, and P2b. , P3b, P4b, P5b, P6b, P7b, P8b.
In FIG. 6B, pictures corresponding to the field image fld1 are denoted by I0t, P1t, P2t, P3t, P4t, P5t, I6t, P7t, and P8t, and pictures corresponding to the field image fld2 are denoted by P0b and P1b. , P2b, I3b, P4b, P5b, P6b, P7b, P8b.
In FIG. 6C, pictures corresponding to the field image fld1 are indicated by I0t, P1t, P2t, I3t, P4t, P5t, I6t, P7t, and P8t, and pictures corresponding to the field image fld2 are indicated by P0b and P1b. , P2b, I3b, P4b, P5b, P6b, P7b, P8b.
In FIG. 6A to FIG. 6C, the GOP is configured in units of frame images obtained by combining the field image fld1 and the field image fld2.

  FIG. 6A shows an example of a GOP structure after re-encoding when there is no frame image including an intra-frame scene change in the moving image shown in FIG. g1a, g2a, and g3a are GOPs after re-encoding, and correspond to GOP · g1, g2, and g3 in FIG. 5, respectively. The frame image that was an I picture in the first encoded moving image information becomes a field image of an I picture and a field image of a P picture after re-encoding. For example, the frame image I0 (FIG. 5) of the first encoded moving image information is re-encoded as two field images (field image I0t and field image P0b) as shown in FIG. 6 (a). At this time, the reference picture of the field image P0b is the field image I0t. Similarly, the frame images I3 and I6 in FIG. 5 are separated and re-encoded into two field images of an I picture and a P picture. For images other than I pictures, the same picture type as before re-encoding is applied during re-encoding.

  FIG. 6B shows an example in which the frame image I3 of the first encoded moving image information shown in FIG. 5 includes an intra-frame scene change. Since the scene change flag sc becomes true in the re-encoding of the frame image I3, the frame image I3 is re-encoded separately into a field image P3t that is a P picture and a field image I3b that is an I picture. In FIG. 6B, the reference picture of the field image P3t is the field image P2t, but a picture other than the field image P2t may be used as long as it belongs to the same scene as the field image P3t. In any case, unless the field image P3t is an I picture, it cannot be the head of the GOP, and the pictures from the field image I0t to the field image P5b constitute one GOP. G1b in FIG. 6B is a GOP corresponding to GOP · g1 and GOP · g2 in FIG.

  FIG. 6C is an example when the frame image I3 shown in FIG. 5 includes an intra-frame scene change, as in FIG. 6B. However, unlike FIG. 6B, the frame image I3 is re-encoded as an I picture so that random access is possible even after re-encoding. GOP · g1c, GOP · g2c, and GOP · g3c are GOPs after re-encoding corresponding to GOP · g1, GOP · g2, and GOP · g3 before re-encoding (see FIG. 5).

FIGS. 6A to 6C show examples in which the frame image is composed only of an I picture and a P picture. However, the first encoded moving picture information includes the I picture and the P picture. , B picture may be included.
As described above, the GOP structure determination unit 105 selects the picture type pt1 and the picture type pt2 for each frame image and determines the GOP structure. As a selection method of the picture type pt1 and the picture type pt2, a method of selecting a picture type that can use an inter-screen prediction mode for the field image fld1 and an intra-screen prediction mode for the field image fld2 is used. be able to.
Further, in addition to the configuration shown in FIG. 3, a function for detecting an inter-frame scene change may be provided, and the picture type may be controlled based on the presence / absence of an intra-frame scene change and the presence / absence of an inter-frame scene change. For example, when neither an inter-frame scene change nor an intra-frame scene change exists, the picture type pt1 and the picture type pt2 may be set to pt1 = P and pt2 = P.

Next, processing when the moving image re-encoding device 100 according to the first embodiment of the present invention performs re-encoding will be described. The encoded information decoding unit 203 decodes the frame image that is the first encoded moving image information input to the picture type determining unit 201.
The frame-field conversion unit 102 converts one frame image decoded by the encoding information decoding unit 203 into two field images.
The in-frame scene change detection unit 104 detects whether or not a scene change that is a scene switching point is included between the two field images converted by the frame-field conversion unit 102.
The GOP structure determination unit 105 selects a picture type (I picture or P picture) for each field image in accordance with the scene change detection result of the in-frame scene change detection unit 104 and determines the GOP structure.
Based on the detection result of the in-frame scene change detection unit 104, the image information encoding unit 103 encodes each of the two field images using the picture type selected by the GOP structure determination unit 105.

  As described above, the intra-frame scene change detection unit 104 detects the presence or absence of an intra-frame scene change for the frame image that is an I picture in the first encoded moving image information, and the GOP structure determination unit 105 Based on the result, the picture type applied to the field image at the time of re-encoding is controlled. As a result, the moving image re-encoding device 100 can re-encode the first encoded moving image information into the second encoded moving image information with good encoding efficiency even if there is an intra-frame scene change. .

(Second Embodiment)
FIG. 7 is a block diagram illustrating a configuration of the moving image re-encoding device 110 according to the second embodiment of the present invention. Parts in which the moving image re-encoding device 110 according to the present embodiment has the same configuration as that of the moving image re-encoding device 100 (FIG. 3) according to the first embodiment are denoted by the same reference numerals and description thereof is provided. Omitted.

The moving image re-encoding device 110 includes a frame-field conversion unit 111 and a GOP structure determination unit 112.
The frame-field conversion unit 111 converts the frame structure image information output from the video memory 205 into a field structure and outputs the field structure to the image information encoding unit 103.
The GOP structure determination unit 112 determines the GOP structure and picture type based on the picture type output from the picture type determination unit 201.

  The frame-field conversion unit 111 operates in the same manner as the frame-field conversion unit 102 except when the scene change flag sc is false and the field image fld1 is output. When the scene change flag sc is true, the same image as the field image fld2 belonging to the scene after the scene change is output instead of the field image fld1 belonging to the scene before the scene change. The output of the field image fld2 is the same as that of the frame-field conversion unit 102.

FIG. 8 is a diagram for explaining conversion processing from one frame image to two field images. As shown in FIG. 8, when the in-frame scene change detection unit 104 detects a scene change, the frame-field conversion unit 111 (FIG. 7) selects one of the two field images fld1. One field image fld2 is converted into the same field image. Conversely, the image of the field image fld2 can be output in the same manner as the image of the field image fld1, but if this is done with the first frame image of the GOP, only the first frame image of the GOP belongs to the scene before the scene change. It becomes an image. Therefore, it is desirable to output the same image as the field image fld2 as the field image fld1.
Similar to the GOP structure determining unit 105, the GOP structure determining unit 112 selects the picture type pt1 and the picture type pt2 for the two field images fld1 and field image fld2, respectively, and determines the GOP structure.

FIG. 9 is a flowchart showing a flow of selection processing of the picture type pt1 and the picture type pt2 in the GOP structure determination unit 112. Hereinafter, the steps of the picture type selection process in the GOP structure determination unit 112 will be described with reference to the flowchart of FIG.
First, in step S20, it is determined whether or not the picture type pt of the frame image frm to be re-encoded is I, that is, whether or not pt = I.
Step S21 is a process performed when the picture type pt = I in step S20. Here, when the intra-frame scene change is not detected, the field image fld1 and the field image fld2 have a high correlation. Therefore, the field image fld2 is encoded by inter-screen prediction using the field image fld1 as a reference picture (pt1 = I, pt2 = P), the encoding efficiency can be improved.
On the other hand, when an intra-frame scene change is detected, the same image is output from the frame-field conversion unit 111 as the field image fld1 and the field image fld2. Therefore, the field image fld2 can be encoded with almost no code amount by inter-screen prediction using the field image fld1 as a reference picture. Eventually, regardless of whether or not an intra-frame scene change is detected, the picture type pt1 and the picture type pt2 are set to pt1 = I and pt2 = P.

On the other hand, step S22 is a process performed when the picture type pt is not equal to I. Here, the picture type pt1 and the picture type pt2 inherit the picture type pt in the first encoded moving picture information (pt1 = pt, pt2 = pt).
An example of the operation of the GOP structure determination unit 112 will be described with reference to FIGS. FIG. 10 is a schematic diagram of the second encoded moving image information after re-encoding as in FIG. 6, and the frame image I3 (FIG. 5) of the first encoded moving image information has an intra-frame scene change. This is an example. The frame image I3 is re-encoded separately for the field image I3t and the field image P3b by the process of step S21 in FIG. At this time, the field image I3t is the same image as the field image P3b by the processing of the frame-field conversion unit 111, and the reference picture of the field image P3b is the field image I3t. Since the field image I3t is the first picture of the GOP, random access is possible.

As described above, the GOP structure determination unit 112 selects the picture type pt1 and the picture type pt2 and determines the GOP structure. Note that the selection method of the picture type pt1 and the picture type pt2 is a method of selecting a picture type that can use the intra prediction mode for the field image fld1 and the inter prediction mode for the field image fld2. That's fine.
Furthermore, as in the first embodiment, in addition to the configuration of the present embodiment, a function for detecting an inter-frame scene change is provided, and the picture type is controlled based on the presence / absence of an intra-frame scene change and the presence / absence of an inter-frame scene change. You may make it do.

  As described above, when the intra-frame scene change detection unit 104 detects the intra-frame scene change in the first encoded moving image information, the frame-field conversion unit 111 is the same as the field image fld2 as the field image fld1. Is output so that the frame image frm has no intra-frame scene change. Thereby, the moving image re-encoding device 110 can re-encode the first encoded moving image information into the second encoded moving image information with good encoding efficiency even if there is an intra-frame scene change. .

  In the second embodiment, when the in-frame scene change detection unit 104 detects a scene change, the frame-field conversion unit 111 converts one field image fld1 of the two field images to the other field image. Although the case of converting to the same field image as the field image fld2 has been described, the present invention is not limited to this. For example, when the in-frame scene change detection unit 104 detects a scene change, the frame-field conversion unit 111 converts one field image fld1 of the two field images fld1 and fld2 with the other field image fld2. The conversion may be performed so that the difference decreases, that is, the field image fld1 and the field image fld2 are similar. By performing such processing, one field image fld1 can be easily predicted from the other field image fld2, and the amount of information when the frame image frm is encoded can be reduced. .

(Third embodiment)
FIG. 11 is a block diagram showing a configuration of a moving image re-encoding device 120 according to the third embodiment of the present invention. Parts in which the moving image re-encoding device 120 according to the present embodiment has the same configuration as that of the moving image re-encoding device 100 (FIG. 3) according to the first embodiment are denoted by the same reference numerals and description thereof will be given. Omitted.

The image information encoding unit 121 encodes the field image output from the frame-field conversion unit 102 using the second encoding method, and outputs the encoded image to the outside of the moving image re-encoding device 120.
The image information encoding unit 121 is the same as the image information encoding unit 103 in that the image information is encoded using the second encoding method. The difference from the image information encoding unit 103 is that the picture type inherits the picture type pt of the frame image in the first encoded moving image information, and the picture type pt1 and the picture type of the field image fld1 and the field image fld2 The point is pt2, and the point where the quantization width is determined with reference to the scene change flag sc during the code amount control.

  If the frame image frm to be re-encoded has an intra-frame scene change, the field image fld1 before the scene change is the last image of a scene and can be temporally referenced from the subsequent pictures. Therefore, assigning a large amount of code to the field image fld1 leads to a decrease in encoding efficiency. On the other hand, the field image fld2 after the scene change is the first image of another scene and has a high possibility of affecting the image quality of subsequent pictures in time, and a large amount of code is assigned to the field image fld2. Therefore, it is considered that the encoding efficiency is improved.

  Considering this, the image information encoding unit 121 sets the quantization widths for the field image fld1 and the field image fld2 as Q1 and Q2, and the quantization width calculated for each field as initial values Q1 ′ and Q2 ′. Using the quantization width correction amounts ΔQ1 and ΔQ2 for each field image, Q1 = Q1 ′ + ΔQ1 and Q2 = Q2 ′ + ΔQ2 are calculated.

FIG. 12 is a flowchart showing a flow of processing for determining quantization width correction amounts ΔQ1 and ΔQ2 in the image information encoding unit 121.
First, in step S30, the value of the scene change flag sc is determined.
In step S31, when the intra-frame scene change detection unit 104 detects a scene change sc, the image information encoding unit 121 configures an encoding target frame image and encodes a field image belonging to the scene before the scene change. The average quantization width ΔQ1 used in the above is increased more than the average quantization width ΔQ2 used for encoding the immediately preceding field image in the same prediction mode. Specifically, ΔQ1 and ΔQ2 are determined so that the quantization width Q1 in the field image fld1 before the scene change is larger than Q1 ′ and the quantization width Q2 in the field image fld2 after the scene change is smaller than Q2 ′. . In the present embodiment, the correction amount of the quantization width is calculated by ΔQ1 = + a and ΔQ2 = −a. Here, a is a variable. In addition, you may use another method as a calculation method of (DELTA) Q1 and (DELTA) Q2.
If a positive value is given to the variable a, the quantization width Q1 of the field image fld1 becomes larger, and the quantization width Q2 of the field image fld2 becomes smaller. That is, the assigned code amount of the field image fld1 before the scene change decreases, and the assigned code amount of the field image fld2 after the scene change increases. The degree of change in image quality can be adjusted by changing the absolute value of a according to the encoding method used.

Step S32 is a process performed when the scene change flag sc is false, and ΔQ1 = 0 and ΔQ2 = 0 are set.
In this way, the image information encoding unit 121 performs appropriate code amount control even when the frame image frm has an intra-frame scene change. Note that the quantization width can be calculated by relatively reducing the code amount assigned to the field belonging to the scene before the scene change or relatively increasing the code amount assigned to the field belonging to the scene after the scene change. The method is not limited to the method of this embodiment.
For example, when the intra-frame scene change detection unit 104 detects a scene change sc, the image information encoding unit 121 configures a frame image to be encoded and is used for encoding a field image belonging to the scene after the scene change. The average quantization width that is formed may be smaller than the average quantization width that is used to encode the field image that forms the frame image to be encoded and belongs to the scene before the scene change. By performing such processing, the information amount of the field image belonging to the scene after the scene change can be increased with respect to the field image belonging to the scene before the scene change.

  As a method for controlling the code amount, in the above description, the quantization width for each field image is uniform for each macroblock. However, the quantization width for each macroblock is set based on the calculated quantization widths Q1 and Q2. You may adjust.

  As described above, when the intra-frame scene change detection unit 104 detects the intra-frame scene change in the first encoded moving image information, the image information encoding unit 121 applies the field image before and after the scene change. Correct the quantization width. As a result, the moving image re-encoding device 120 can re-encode the first encoded moving image information into the second encoded moving image information with good encoding efficiency even if there is an intra-frame scene change. . Furthermore, the present embodiment can be implemented in combination with the first embodiment or the second embodiment, and the encoding efficiency is further improved.

  The moving image re-encoding device according to the first to third embodiments described above uses only the field structure for re-encoding, but the frame structure for each frame image in one encoded moving image. As long as it is an encoding scheme capable of frame-field adaptive encoding that adaptively selects the field structure, it may be used.

The moving image re-encoding device according to the embodiment of the present invention detects the presence or absence of an intra-frame scene change in the first encoded moving image information, and controls the re-encoding process based on the result, thereby Even if there is an inner scene change, it can be re-encoded to the second encoded moving image information with good encoding efficiency. In the case of the first embodiment, when the in-frame scene change is detected, the above-described effect can be obtained by encoding the field image belonging to the scene before the scene change as a P picture with higher encoding efficiency.
In the case of the second embodiment, the above effect can be obtained by substituting the field image immediately before the scene change within the frame with the field image immediately after the scene change to eliminate the scene change within the frame.
In the case of the third embodiment, the amount of code to be assigned can be adjusted by changing the quantization width of the field image constituting the frame image in which the intra-frame scene change has occurred, thereby preventing a decrease in encoding efficiency. .

  In the embodiment described above, the motion vector synthesis unit 101, the frame-field conversion unit 102, the image information encoding unit 103, the intra-frame scene change detection unit 104, and the GOP structure determination unit shown in FIGS. 105, frame-field conversion unit 111, GOP structure determination unit 112, image information encoding unit 121, picture type determination unit 201, encoded information analysis unit 202, encoded information decoding unit 203, motion vector detection unit 208, complex By recording the function of the city calculation unit 210 or a program for realizing a part of these functions on a computer-readable recording medium, and causing the computer system to read and execute the program recorded on the recording medium The moving image re-encoding device may be controlled. Here, the “computer system” includes an OS and hardware such as peripheral devices.

  The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Further, the “computer-readable recording medium” dynamically holds a program for a short time, like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, it is also assumed that a server that holds a program for a certain time, such as a volatile memory inside a computer system that serves as a server or client. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes designs and the like that do not depart from the gist of the present invention.

FIG. 5 is a diagram illustrating an example in which two field images fld1 and fld2 constitute one frame image frm. It is a figure which shows the example of the correlation between pictures. It is a block diagram which shows the structure of the moving image re-encoding apparatus 100 in the 1st Embodiment of this invention. It is a flowchart which shows the process of the GOP structure determination part 105 by embodiment of this invention. It is a schematic diagram of the 1st encoding moving image information by the 1st Embodiment of this invention. It is a schematic diagram of the 2nd encoding moving image information by the 1st Embodiment of this invention. It is a block diagram which shows the structure of the moving image re-encoding apparatus 110 in the 2nd Embodiment of this invention. It is a figure for demonstrating the conversion process from one frame image to two field images. It is a flowchart which shows the flow of the selection process of the picture type pt1 in the GOP structure determination part 112, and the picture type pt2. It is a schematic diagram of the 2nd encoding moving image information by the 2nd Embodiment of this invention. It is a block diagram which shows the structure of the moving image re-encoding apparatus 120 by the 3rd Embodiment of this invention. 12 is a flowchart showing a flow of processing for determining quantization width correction amounts ΔQ1 and ΔQ2 in the image information encoding unit 121. FIG. 10 is a block diagram showing a configuration of a conventional moving image re-encoding device 500. It is a figure for demonstrating the case where a scene change occurs in a moving image.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Moving image re-encoding apparatus, 101 ... Motion vector synthetic | combination part, 102 ... Frame-field conversion part, 103 ... Image information encoding part, 104 ... In-frame scene change detection part , 105... GOP structure determination unit, 110... Video re-encoding device, 111... Frame-field conversion unit, 112... GOP structure determination unit, 120. , 121 ... Image information encoding unit, 201 ... Picture type discrimination unit, 202 ... Encoding information analysis unit, 203 ... Encoding information decoding unit, 205 ... Video memory, 208 ...・ Motion vector detection unit, 209... Information buffer, 210... Complexity calculation unit

Claims (11)

A frame-field conversion unit for converting one frame image into two field images;
A scene change detection unit for detecting whether or not a scene change that is a scene switching point is included between the two field images;
An image information encoding unit that encodes each of the two field images based on a detection result of the scene change detection unit using an intra prediction mode or an inter prediction mode;
A moving picture encoding apparatus comprising:   The moving image encoding apparatus according to claim 1, wherein the image information encoding unit encodes the frame image as two field images when the scene change detection unit detects a scene change. . A GOP structure determining unit for selecting a GOP structure by selecting an intra-screen prediction mode or an inter-screen prediction mode for each field image according to a scene change detection result of the scene change detection unit;
The moving image encoding apparatus according to claim 2, wherein the image information encoding unit encodes each field image using the prediction mode selected by the GOP structure determination unit.   The GOP structure determination unit is used for the two field images according to the prediction mode used for the frame image converted by the frame-field conversion unit when the scene change detection unit detects a scene change. 4. The moving picture encoding apparatus according to claim 3, wherein a combination of prediction modes is selected.   5. The GOP structure determination unit according to claim 4, wherein, when the scene change detection unit detects a scene change, the prediction mode used for the field image before the scene change is set as an inter-screen prediction mode. The moving image encoding apparatus described.   The frame-field conversion unit converts one field image of the two field images so that a difference from the other field image is reduced when the scene change detection unit detects a scene change. The moving picture encoding apparatus according to claim 2, wherein:   The frame-field conversion unit converts one field image of the two field images into the same field image as the other field image when the scene change detection unit detects a scene change. The moving picture coding apparatus according to claim 6.   When the scene change detection unit detects a scene change, the image information encoding unit forms an encoding target frame image and is used for encoding a field image belonging to the scene before the scene change. The moving picture coding apparatus according to claim 2, characterized in that is increased more than an average quantization width used for coding the immediately preceding field picture in the same prediction mode.   When the scene change detection unit detects a scene change, the image information encoding unit forms an encoding target frame image and is used for encoding a field image belonging to the scene after the scene change. 3. The moving picture coding apparatus according to claim 2, wherein the video image coding apparatus is made smaller than an average quantization width used for coding a field image constituting a frame image to be coded and belonging to a scene before a scene change. On the computer,
A first step of converting one frame image into two field images;
A second step of detecting whether a scene change that is a scene switching point is included between the two field images;
A third step of encoding each of the two field images based on a detection result of the scene change detection unit using an intra prediction mode or an inter prediction mode;
A program characterized in that is executed. A frame-field conversion process for converting one frame image into two field images;
A scene change detection process for detecting whether a scene change that is a scene switching point is included between the two field images;
An image information encoding process for encoding each of the two field images based on a detection result of the scene change detection unit using an intra prediction mode or an inter prediction mode;
A moving picture encoding method comprising:

Images