JP3877043B2 - Scene adaptive video encoder - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a scene adaptive video encoding apparatus, and more particularly to a scene adaptive video encoding apparatus that adaptively assigns picture types for performing interframe encoding based on scene characteristics.
[0002]
[Prior art]
The GOP structure in MPEG-2 is generally a structure as shown in FIG. That is, the GOP structure performs predictive coding from an I (intra) picture that performs orthogonal transform on a pixel value in a frame without using prediction, and an I or P picture located in the past in time. P picture to be performed and B picture to perform predictive coding from the screen in both directions using I or P picture located in the past and the future in time, and the combination of the three types of pictures is shown in the figure As shown, the order of I, B, B, P, B, B, P, B, B, P, B, B, P, B, B, and I pictures is generally used.
[0003]
[Problems to be solved by the invention]
However, in the GOP structure, a B picture is predicted by an I or P picture and a P (or I) picture three screens after the I or P picture. In this case, for example, as shown in FIG. 8, when an object that has been at the left end of the screen 50 has already moved to the right end of the screen 51 after three screens, the prediction efficiency of the P picture decreases. There has been a problem that the image quality is lowered due to the unexpected prediction.
[0004]
An object of the present invention is to provide a scene-adaptive moving image encoding apparatus that solves the problems of the prior art and has good image quality even in a moving image that moves quickly.
[0005]
[Means for Solving the Problems]
In order to achieve the above object, the present invention provides an input image signal and a locally decoded image prior to encoding in a video encoding apparatus that performs encoding with any one of I, P, and B picture types. A signal and a motion estimation signal are input to generate a motion compensated prediction image, an error index E indicating an error between the input image signal and the motion compensation prediction image, and motion vectors of all macroblocks of the motion compensation prediction image A motion compensation unit that obtains a motion index M that is a ratio occupied by a motion vector that is equal to or greater than a predetermined first threshold value, and a predetermined one of all macroblocks of the input image signal prior to encoding. A scene analysis unit for obtaining an interpolation prediction index B, which is a ratio of macroblocks having a luminance variance in units of macroblocks equal to or greater than a second threshold, and the error index E is predetermined. When the motion index M is greater than a predetermined fourth threshold value, the encoded picture type of the input image is determined to be a P picture, and the error index E is a predetermined third threshold value. When the motion index M is smaller or smaller than a predetermined fourth threshold, the picture type selection unit inserts a larger number of B pictures as the interpolation prediction index B is smaller. is there.
[0006]
According to this feature, since the picture type to be encoded can be determined according to the magnitude of the motion of the image, it is possible to perform encoding with good image quality even in a fast-moving moving image.
[0007]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a scene adaptive moving image encoding apparatus according to an embodiment of the present invention.
[0008]
In the figure, an image signal is input to a picture storage unit 1 that can store image signals for several frames and a scene analysis unit 3. The scene analysis unit 3 calculates an interpolation prediction index B of the input frame and sends it to the picture type selection unit 4 as will be apparent from the description below. The picture type selection unit 4 determines a picture type to be encoded based on an error index E, a motion index M, and the interpolation prediction index B from a motion compensation unit (MC unit) 6 described later. These operations will be described in detail later.
[0009]
The picture rearrangement unit 2 needs to perform encoding using screens that are temporally changed in the B picture, so that the picture rearrangement unit 2 matches the picture type (I, B, P picture) specified by the picture type selection unit 4. Rearrange the screen. The image signals rearranged by the picture rearrangement unit 2 are input to the subtraction unit 5. The switch units 7a and 7b are turned on and off respectively when the I picture is encoded, and are turned off and turned on when the P and B pictures are encoded.
[0010]
The DCT unit 8 orthogonally transforms the input image signal or prediction error signal. The orthogonally transformed signal is quantized by the quantizing unit 9, variable-length coded by the variable-length coding unit 10, temporarily stored in the transmission buffer 11, and then external video as a video bit stream at a predetermined rate. Is sent out. On the other hand, the inverse quantization unit 12 inversely quantizes the signal quantized by the quantization unit 9. The inversely quantized signal is subjected to inverse orthogonal transform by the inverse DCT unit 13 and is locally decoded. The signal is added to the predicted image from the motion compensation unit 6 by the adder 14 and accumulated in the frame memory 15. The motion estimation unit 16 performs motion estimation from the input image signal and the locally decoded image stored in the frame memory 15 and hands the motion vector to the motion compensation unit 6. The motion compensation unit 6 outputs the prediction image to the subtraction unit 5, calculates the error index E and the motion index M, and sends them to the picture type selection unit 4.
[0011]
Next, the operation of the picture type selection unit 4 will be described in detail with reference to the flowchart of FIG. In step S1, it is determined whether the picture to be encoded is an I picture. When this determination is affirmative, the process proceeds to step S2, an I picture is selected as the picture, and the process ends. On the other hand, when the determination in step S1 is negative, the process proceeds to step S3.
[0012]
In step S3, it is determined whether or not the picture is a reserved picture described later in steps S11 and S13. When this determination is affirmative, the process proceeds to step S4, and a reserved picture is selected as the picture. When a reserved picture is selected, the number of reserved pictures is reduced by one.
[0013]
On the other hand, when the determination in step S3 is negative, the process proceeds to step S5. In step S5, an interpolation prediction index B is acquired from the scene analysis unit 3. Next, in steps S6 and S7, an error index E and a motion index M of the determination picture are obtained from the motion compensation unit (MC unit), respectively. Subsequently, in step S8, it is determined whether or not the error index E and the motion index M are larger than the threshold values δ _E and δ _M set for the respective parameters. When this determination is affirmative, the process proceeds to step S9, and a P picture is selected as the picture. As a result, the P picture is selected for the fast moving image. Then, the process ends.
[0014]
Next, when the determination in step S8 is negative, the process proceeds to step S10, and the number N of _B pictures to be continuously inserted is determined based on the magnitude of the interpolation prediction index B acquired in step S5. The As described later, in the present embodiment, the inner interpolation predictor B, since the luminance dispersion A macroblock is proportional to a predetermined threshold value [delta] _A more macro blocks, the number of B pictures to be inserted the succession N _B becomes larger as the interpolation prediction index B is smaller.
[0015]
In step S11, the subsequent N _B sheets of pictures including the picture book the B picture. In step S12, the (N _B +1) th picture counted from the picture is reserved as a P picture. Next, the process proceeds to step S13, where a B picture is selected as the picture and the process ends. When the (N _B +1) th picture overlaps with the I picture, the reservation of the P picture may be canceled.
[0016]
As a result, in the case of an image with slow motion, the number of B pictures corresponding to the slowness is inserted, and the coding efficiency can be improved. On the other hand, in the case of a fast-moving image, only the P picture is used, so that it is possible to prevent deterioration in image quality.
[0017]
Next, details of step S5 will be described with reference to the flowchart of FIG. In step S31, the scene analysis unit 3 acquires an input frame. In step S32, the input frame is divided into macro blocks. In step S33, the luminance variance A is calculated for each macroblock. In step S34, counting the number of macro blocks equal to or greater than the threshold value [delta] _A there is luminance dispersion A, and the number and C _A. In step S35, the interpolation prediction index B is obtained from C _A / MB _cnt (where MB _cnt is the total number of macroblocks in the frame). Note that the scene analysis unit 3 may obtain the interpolation prediction index B based on the ratio of the high-frequency component of the input frame instead of the above.
[0018]
Next, details of step S6 will be described with reference to the flowchart of FIG. In step S41, the motion compensation unit 6 receives the motion compensated prediction image (MC image) and the original image, and in step S42, calculates and stores the mean square error E _pred from both these images. In step S43, the mean square error E _pred is set as the error index E. Instead of the mean square error, a sum of error squares or a sum of error absolute values may be used.
[0019]
Next, details of step S7 will be described with reference to the flowchart of FIG. In step S51, the motion compensation unit 6 acquires a motion vector MV _F for frame prediction motion compensation for all macroblocks. In step S52, when the threshold for MV _F is δ _MV , the number of macroblocks satisfying _{MV F} > δ _MV is counted, and the count number _CM is stored. In step S53, the motion index M is obtained from C _M / MB _cnt (where MB _cnt is the total number of macroblocks in the frame). Note that an average of motion vectors of all macroblocks may be used as the motion index M.
[0020]
As is clear from the above description, both the error index E and the motion index M increase as the image moves faster. In addition, the interpolation prediction index B increases as the ratio of the high frequency component to the image increases. Therefore, according to the present embodiment, as shown in FIG. 6, when the motion of the image is large, P pictures for forward prediction are continuously selected. Accordingly, one or a plurality of B pictures are selected between P pictures.
[0021]
In the embodiment, the determination in step S8 is performed using the error index E and the motion index M. However, the present invention is not limited to this, and the determination may be performed using only one of them. May be a determination by using in combination with the interpolation prediction index B.
[0022]
【The invention's effect】
As is apparent from the above description, according to the present invention, it becomes possible to select a picture type for encoding according to the magnitude of motion of an image and the effect of motion compensation prediction , and adaptively select a picture type. become able to. That is, when the image motion is large, the P picture is selected. On the other hand, when the image motion is small, a plurality of B pictures are continuously selected. And the encoding efficiency can be improved.
[Brief description of the drawings]
FIG. 1 is a block diagram of an encoding apparatus including a configuration according to an embodiment of the present invention.
FIG. 2 is a flowchart showing the operation of one embodiment of the present invention.
FIG. 3 is a flowchart showing details of the operation in step S5 of FIG.
FIG. 4 is a flowchart showing details of the operation in step S6 of FIG.
FIG. 5 is a flowchart showing details of the operation in step S7 in FIG. 2;
FIG. 6 is a conceptual diagram of a picture type arrangement obtained by an embodiment of the present invention.
FIG. 7 is an explanatory diagram of a conventional GOP structure.
FIG. 8 is an explanatory diagram of an image having a fast-moving object.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Picture storage part, 2 ... Picture rearrangement part, 3 ... Scene analysis part, 4 ... Picture type selection part, 6 ... Motion compensation part

Claims (1)

In a video encoding apparatus that performs encoding with any one of I, P, and B picture types,
Prior to encoding, an input image signal, a locally decoded image signal, and a motion estimation signal are input to generate a motion compensated predicted image, and an error index E indicating an error between the input image signal and the motion compensated predicted image; A motion compensation unit that obtains a motion index M that is a ratio occupied by a motion vector equal to or greater than a predetermined first threshold among motion vectors of all macroblocks of the motion compensated prediction image;
Prior to encoding, a scene analysis unit that obtains an interpolation prediction index B that is a ratio of macroblocks having luminance dispersion in units of macroblocks equal to or greater than a predetermined second threshold among all macroblocks of the input image signal When,
When the error index E is greater than a predetermined third threshold and the motion index M is greater than a predetermined fourth threshold, the encoded picture type of the input image is determined as a P picture, and the error When the index E is smaller than a predetermined third threshold or the motion index M is smaller than a predetermined fourth threshold, the number of B pictures determined based on the size of the interpolation prediction index B is A scene-adaptive moving image encoding apparatus comprising a picture type selection unit to be inserted.

Images