JP2009284058A - Moving image encoding device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a moving image encoding device that determines the search range of a motion vector and searches for the motion vector without performing many arithmetic operations. <P>SOLUTION: The moving image encoding device includes a motion prediction unit 9 which evaluates the level ¾PMV¾ of a prediction vector calculated when encoded data are generated by a variable-length encoding unit 4 to determine the search range of the motion vector MV, and collates an input image signal with a reference image signal stored in a frame memory 8 to search for the suitable motion vector MV from the search range, and a motion compensation unit 10 uses the motion vector MV searched by the motion prediction unit 9 to generate a prediction image signal from the reference image signal stored in the frame memory 8. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a moving picture coding apparatus having a function of detecting a motion vector necessary for motion compensated interframe prediction for reducing temporal redundancy when coding a picture signal.

For example, in an international standard video encoding method such as “MPEG” or “ITU-T H.26x”, a luminance signal of 16 × 16 pixels and a corresponding color difference signal of 8 × 8 pixels are provided for each frame of an image signal. A method of compressing by performing motion compensation processing and orthogonal transform / transform coefficient quantization processing using the combined block data (hereinafter referred to as macroblock) as a unit is adopted.
When detecting a motion vector used for the motion compensation process, a block matching method is mainly used in which an input image signal to be encoded and a reference image signal are matched in units of 16 × 16 pixels.

In the video encoding method, the amount of calculation required for the motion prediction process for searching for a motion vector is generally larger than that for other processes.
Ideally, the search should be performed for all the pixels in the block in all the regions in the image, but such a method requires a large amount of computation. Therefore, a technique for reducing the amount of calculation while maintaining performance has been proposed.
For example, in Patent Document 1 below, motion vector information obtained at the time of encoding a past image is held, and a motion vector obtained by a macroblock located at the same position as the currently encoded image is distributed. Thus, the motion vector search range is estimated.
Thereby, since the search range of the motion vector is limited, the processing amount of the motion prediction process is reduced.

JP-A-11-308617 (paragraph numbers [0015] to [0016], FIG. 2)

  Since the conventional moving image encoding apparatus is configured as described above, the search range of motion vectors is limited, and the amount of motion prediction processing is reduced. However, when estimating the search range of the motion vector, there is a problem that it is necessary to prepare many memories and perform many operations (for example, by using a processor such as a DSP and performing software processing) When performing the encoding process, there is a limit to the amount of memory that can be used and the amount of calculation per unit time, so it is necessary to reduce the amount of memory and the amount of calculation as much as possible.

  The present invention has been made in order to solve the above-described problems, and provides a moving image encoding apparatus capable of determining a search range of a motion vector and searching for a motion vector without performing many operations. The purpose is to obtain.

  The moving image encoding apparatus according to the present invention evaluates the size of a prediction vector calculated when encoded data is generated by the encoded data generation means, determines a motion vector search range, and inputs an input image signal Motion vector search means for searching for an appropriate motion vector from the search range by comparing the reference image signal generated by the reference image signal generation means and the motion vector search means by the motion vector search means Is used to generate the predicted image signal from the reference image signal generated by the reference image signal generation means.

  According to the present invention, the motion vector search range is determined by evaluating the size of the prediction vector calculated when the encoded data is generated by the encoded data generation means, and the input image signal and the reference image signal are generated. A motion vector search means for collating the reference image signal generated by the means and searching for an appropriate motion vector from the search range is provided, and the motion compensation means refers to the motion vector searched by the motion vector search means. Since the prediction image signal is generated from the reference image signal generated by the image signal generation means, the motion vector search range can be determined and the motion vector can be searched without performing many operations. As a result, there is an effect that the motion vector can be efficiently searched without causing deterioration of the encoded image quality.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a moving picture coding apparatus according to Embodiment 1 of the present invention. In the figure, a differentiator 1 is an input picture signal which is a frame signal and a frame signal output from the motion compensation unit 10. The difference between the prediction image signals is obtained, and the prediction residual signal, which is the difference signal, is divided into macroblock units and output to the DCT conversion unit 2. The differentiator 1 constitutes an image signal difference means.

The DCT transform unit 2 performs a discrete cosine transform on the prediction residual signal in units of macroblocks output from the differentiator 1.
The quantization unit 3 performs a process of quantizing the prediction residual signal subjected to the discrete cosine transform by the DCT transform unit 2.
The DCT conversion unit 2 and the quantization unit 3 constitute a quantization means.

  The variable length encoding unit 4 generates encoded data by variable length encoding the prediction residual signal quantized by the quantization unit 3, encodes the motion vector output from the motion prediction unit 9, A process for outputting the encoded data and the encoded motion vector is performed. The variable length encoding unit 4 constitutes encoded data generating means.

The inverse quantization unit 5 performs a process of inversely quantizing the prediction residual signal quantized by the quantization unit 3.
The inverse DCT transformer 6 performs inverse discrete cosine transform on the prediction residual signal inversely quantized by the inverse quantizer 5.
The adder 7 performs a process of generating a reference image signal by adding the prediction residual signal subjected to inverse discrete cosine transform by the inverse DCT transform unit 6 and the predicted image signal output from the motion compensation unit 10.
The frame memory 8 is a memory for storing the reference image signal generated by the adder 7.
The inverse quantization unit 5, the inverse DCT conversion unit 6, the adder 7, and the frame memory 8 constitute reference image signal generation means.

The motion prediction unit 9 evaluates the size of the prediction vector calculated when the encoded data is generated by the variable length encoding unit 4 to determine the search range of the motion vector, and stores the motion vector in the input image signal and the frame memory 8. The stored reference image signal is collated, and a process for searching for an appropriate motion vector from the search range is performed. Note that the motion prediction unit 9 constitutes a motion vector search means.
The motion compensation unit 10 generates a predicted image signal from the reference image signal stored in the frame memory 8 using the motion vector searched by the motion prediction unit 9, and sends the predicted image signal to the differencer 1 and the adder 7. Perform the output process. The motion compensation unit 10 constitutes motion compensation means.

FIG. 2 is a block diagram showing the motion prediction unit 9 of the moving picture encoding apparatus according to Embodiment 1 of the present invention. In the figure, the search range determination unit 11 generates encoded data by the variable length encoding unit 4. The motion vector search range is determined to be a wide range when the predicted vector size calculated is larger than the threshold δ, and the motion vector search range is set to the wide range when the predicted vector size is smaller than the threshold δ. A process for determining a narrower narrow range is performed.
The motion search unit 12 collates the input image signal with the reference image signal stored in the frame memory 8 and performs a process of searching for an appropriate motion vector from the search range determined by the search range determination unit 11.

Next, the operation will be described.
The differentiator 1 receives an input image signal that is a frame signal, and receives a predicted image signal that is a frame signal from a motion compensation unit 10 to be described later, obtains a difference between the input image signal and the predicted image signal, The prediction residual signal, which is a signal, is divided into macroblock units and output to the DCT transform unit 2.

When the DCT transform unit 2 receives the prediction residual signal in units of macroblocks from the differentiator 1, the DCT transform unit 2 performs discrete cosine transform on the prediction residual signal and outputs the prediction residual signal after the discrete cosine transform to the quantization unit 3. .
When the quantization unit 3 receives the prediction residual signal after the discrete cosine transform from the DCT conversion unit 2, the quantization unit 3 quantizes the prediction residual signal and converts the quantized prediction residual signal into the variable length encoding unit 4 and the inverse. It outputs to the quantization part 5.

When the variable length coding unit 4 receives the quantized prediction residual signal from the quantization unit 3, the variable length coding unit 4 performs variable length coding on each macroblock of the prediction residual signal, and generates coded data. The encoded data is output.
The variable length encoding unit 4 encodes a motion vector MV searched for by a motion prediction unit 9 described later, and outputs the encoded motion vector MV.

Note that the variable length coding unit 4 calculates the prediction vector PMV of the macroblock when generating the coded data of the macroblock.
Hereinafter, a method for calculating the prediction vector PMV will be described.
Here, a method determined by ISO / IEC 14496-2 (commonly known as MPEG-4 visual encoding system) will be described as an example.

For example, if the target macroblock for which the prediction vector PMV is calculated is a current macroblock (current MB) as shown in FIG. 3, the encoded macros located on the left side, top, and top right of the current macroblock, respectively. Using the motion vectors MV _a , MV _b , and MV _c of the macroblocks A, B, and C, the prediction vector PMV of the current macroblock is calculated.
That is, the motion vectors MV _a , MV _b , and MV _c of the encoded macroblocks A, B, and C are substituted into the following equation (1), and the median (center) of the motion vectors MV _a , MV _b , and MV _c is assigned. By calculating (value), the prediction vector PMV is calculated.
PMV = median (MV _a , MV _b , MV _c ) (1)

Here, an example in which the prediction vector PMV is calculated by a method determined by the MPEG-4 visual encoding method has been described, but this is only an example, for example, ITU-T H.264. The prediction vector PMV may be calculated by a method determined by an encoding method such as H.264.
In the example of FIG. 1, the variable length encoding unit 4 calculates the prediction vector PMV, but a processing unit other than the variable length encoding unit 4 (for example, a dedicated processing unit that calculates the prediction vector PMV) ) May calculate the prediction vector PMV.
Note that, as shown in FIG. 3, since there is a high possibility that the motion is continuous in the macro block adjacent to the encoding target macro block, the motion vector of the encoding target macro block is obtained by using the prediction vector PMV. Roughly predictable.

When the inverse quantization unit 5 receives the quantized prediction residual signal from the quantization unit 3, the inverse quantization unit 5 performs inverse quantization on the prediction residual signal, and converts the prediction residual signal after inverse quantization to the inverse DCT transform unit 6. Output to.
When the inverse DCT transform unit 6 receives the prediction residual signal after inverse quantization from the inverse quantization unit 5, the inverse DCT transform unit 6 performs inverse discrete cosine transform on the prediction residual signal, and converts the prediction residual signal after inverse discrete cosine transform to Output to adder 7.
When the adder 7 receives the prediction residual signal after the inverse discrete cosine transform from the inverse DCT transform unit 6 and receives the prediction image signal from the motion compensation unit 10 described later, the adder 7 adds the prediction residual signal and the prediction image signal. The reference image signal is generated, and the reference image signal is stored in the frame memory 8.

When the variable length encoding unit 4 calculates the prediction vector PMV when the variable length encoding unit 4 generates encoded data, the motion prediction unit 9 evaluates the size of the prediction vector PMV, determines the search range of the motion vector MV, and inputs The image signal and the reference image signal stored in the frame memory 8 are collated, and an appropriate motion vector MV is searched from the search range.
Hereinafter, the processing content of the motion estimation part 9 is demonstrated concretely.

When the variable length coding unit 4 calculates the prediction vector PMV, the search range determination unit 11 of the motion prediction unit 9 calculates the magnitude | PMV | of the prediction vector PMV.
As a method of calculating the magnitude | PMV | of the prediction vector PMV, for example, the absolute value of the horizontal component and the vertical component of the prediction vector PMV are calculated, and the larger absolute value is used as the magnitude of the prediction vector PMV. A method of setting PMV | is conceivable.
Alternatively, a method is conceivable in which the sum of squares of the horizontal and vertical components of the prediction vector PMV is calculated and the sum of the squares is used as the magnitude | PMV | of the prediction vector PMV.

The search range determination unit 11 of the motion prediction unit 9 compares the magnitude | PMV | of the prediction vector PMV with a threshold δ, and if the magnitude | PMV | of the prediction vector PMV is larger than the threshold δ, the search for the motion vector MV is performed. Determine the range extensively.
On the other hand, when the magnitude | PMV | of the prediction vector PMV is smaller than the threshold δ, the motion vector search range MV is determined to be a narrow range (wide range> narrow range).
| PMV | ≧ δ → Wide range | PMV | <δ → Narrow range

In general, the required search range differs depending on the magnitude of the motion, and if the motion is large, a wide range of search is required, but if the motion is small, a narrow range search is sufficient. On the other hand, if a wide range search is performed, useless calculation is performed.
For this reason, in the first embodiment, wasteful computation can be reduced by switching the search range.

The motion search unit 12 of the motion prediction unit 9 collates the input image signal with the reference image signal stored in the frame memory 8 and corrects the search range (wide range or narrow range) determined by the search range determination unit 11. Search for a motion vector MV.
That is, the motion search unit 12 selects an arbitrary motion vector from among a plurality of motion vectors within the search range (motion vector candidates of the macroblock), and encodes the input image signal by the amount of the motion vector. Move the target macroblock.
Then, the motion search unit 12 collates the moved macroblock with the reference image signal and evaluates whether or not they are similar.
As an evaluation method, for example, an absolute value difference sum (hereinafter sometimes abbreviated as “SAD”) can be used. Alternatively, the sum of squared differences can be used.

The motion search unit 12 sequentially selects all motion vectors within the search range, moves the macroblock to be encoded by the motion vector, and collates the moved macroblock with the reference image signal. Evaluate whether the two are similar.
The motion search unit 12 uses the motion vector having the highest evaluation result (for example, the motion vector having the smallest SAD) among the plurality of motion vectors in the search range as an appropriate motion vector MV, The data is output to the variable length coding unit 4.

When the motion prediction unit 9 searches for the motion vector MV, the motion compensation unit 10 generates a predicted image signal from the reference image signal stored in the frame memory 8 using the motion vector MV, and calculates the difference between the predicted image signal. Output to the device 1 and the adder 7.
In actual encoding, an optimal encoding mode is selected in consideration of not only the inter-frame prediction mode but also the intra-frame (intra) encoding mode, etc., but this point is out of the essence of the present invention. Description is omitted.

  As apparent from the above, according to the first embodiment, the motion vector is evaluated by evaluating the magnitude | PMV | of the prediction vector PMV calculated when the encoded data is generated by the variable length encoding unit 4. A motion prediction unit 9 is provided that determines an MV search range, collates an input image signal with a reference image signal stored in the frame memory 8, and searches for an appropriate motion vector MV from the search range. Is configured to generate the predicted image signal from the reference image signal stored in the frame memory 8 using the motion vector MV searched by the motion prediction unit 9, so that the motion vector can be obtained without performing many operations. As a result, it is possible to efficiently search for a motion vector without degrading the encoded image quality. Achieve the kill effect.

That is, if the search range determination unit 11 of the motion prediction unit 9 determines a search range required according to the magnitude of the motion of the prediction vector PMV, and the motion search unit 12 searches only the required search range. Therefore, useless computation is reduced, and as a result, encoded data can be created with a small amount of computation while maintaining equivalent image quality.
In particular, in the first embodiment, the search range of the motion vector MV is switched by using the prediction vector PMV that is always calculated when the encoded data is generated. It is possible to determine the search range of the motion vector MV. Therefore, it is possible to reduce the amount of calculation required for the motion vector search and to encode with a smaller amount of calculation than the conventional encoding device.

Needless to say, the method of Embodiment 1 can be applied to a method of switching between intra-frame prediction and inter-frame prediction in units of macroblocks as employed in the above international standard video coding method.
This method is also disclosed in MPEG-4, H.264. It goes without saying that the present invention can be applied to all international standard video encoding schemes such as H.264 / AVC and VC-1.

Embodiment 2. FIG.
4 is a block diagram showing a moving picture coding apparatus according to Embodiment 2 of the present invention. In the figure, the same reference numerals as those in FIG.
The motion prediction unit 20 evaluates the size of the prediction vector calculated when the encoded data is generated by the variable length encoding unit 4, determines the motion vector search range and the motion vector search density, and inputs the input image. The signal and the reference image signal stored in the frame memory 8 are collated, and processing for searching an appropriate motion vector from the search range is performed. Note that the motion prediction unit 20 constitutes a motion vector search means.

FIG. 5 is a block diagram showing the motion prediction unit 20 of the moving picture coding apparatus according to Embodiment 2 of the present invention. In the figure, the search range / density determination unit 21 is the same as the search range determination unit 11 of FIG. When the size of the prediction vector calculated when the encoded data is generated by the variable length encoding unit 4 is larger than the threshold δ, the motion vector search range is determined in a wide range, and the size of the prediction vector is If it is smaller than the threshold δ, the motion vector search range is determined to be a narrow range.
However, unlike the search range determination unit 11 in FIG. 2, the search range / density determination unit 21 determines the search density of the motion vector when the search range of the motion vector is determined. That is, the motion vector search density is determined according to the size of the motion vector search range.

Similar to the motion search unit 12 in FIG. 2, the motion search unit 22 collates the input image signal with the reference image signal stored in the frame memory 8, and determines the appropriate value from the search range determined by the search range / density determination unit 21. A process for searching for a motion vector is performed.
However, unlike the motion search unit 12 in FIG. 2, the motion search unit 22 is determined by the search range / density determination unit 21 when searching for an appropriate motion vector from the search range determined by the search range / density determination unit 21. The motion vector is searched with the search density.

Next, the operation will be described.
The second embodiment is the same as the first embodiment except that the motion prediction unit 20 is mounted instead of the motion prediction unit 9 in FIG. 1, and therefore only the operation of the motion prediction unit 20 is performed. explain.
The motion prediction unit 20 calculates the prediction vector PMV when the variable length encoding unit 4 generates encoded data, evaluates the size of the prediction vector PMV, and determines the search range of the motion vector MV and the motion vector MV. The search density is determined, the input image signal and the reference image signal stored in the frame memory 8 are collated, and an appropriate motion vector MV is searched from the search range.
Hereinafter, the processing content of the motion estimation part 20 is demonstrated concretely.

When the variable length coding unit 4 calculates the prediction vector PMV, the search range / density determination unit 21 of the motion prediction unit 20 calculates the magnitude | PMV | of the prediction vector PMV.
As a method of calculating the magnitude | PMV | of the prediction vector PMV, for example, the absolute value of the horizontal component and the vertical component of the prediction vector PMV are calculated, and the larger absolute value is used as the magnitude of the prediction vector PMV. A method of setting PMV | is conceivable.
Alternatively, a method is conceivable in which the sum of squares of the horizontal and vertical components of the prediction vector PMV is calculated and the sum of the squares is used as the magnitude | PMV | of the prediction vector PMV.

The search range / density determination unit 21 compares the magnitude | PMV | of the prediction vector PMV with a threshold δ, and if the magnitude | PMV | of the prediction vector PMV is larger than the threshold δ, the search range / density determination unit 21 widens the search range of the motion vector MV. To decide.
On the other hand, when the magnitude | PMV | of the prediction vector PMV is smaller than the threshold δ, the motion vector search range MV is determined to be a narrow range (wide range> narrow range).
| PMV | ≧ δ → Wide range | PMV | <δ → Narrow range

When the search range / density determination unit 21 determines the search range of the motion vector MV, the search range / density determination unit 21 determines the search density of the motion vector MV according to the size of the search range.
That is, the search range / density determination unit 21 determines to decrease the search density of the motion vector MV as the search range of the motion vector MV is wider.
Here, the relationship between the search density of the motion vector MV and the search range of the motion vector MV is as follows.
D = C / R (2)
Here, D is the search density, R is the search range, and C is the number of motion vector search points (the number of motion vector candidates).

As is clear from equation (2), when the search range R of the motion vector MV is wide and the search density D of the motion vector MV is high, the number of search points C of the motion vector MV increases, and conversely, the search for the motion vector MV is performed. It can be seen that when the range R is narrow and the search density D of the motion vector MV is low, the number of search points C of the motion vector MV decreases.
Therefore, it is understood that the search density D may be determined in accordance with the ratio of the search range R of the motion vector MV in order to prevent the calculation amount from being increased by switching the search point number C of the motion vector MV.

When the search range R of the motion vector MV is wide, for example, ± 16 pixels around the search origin (0,0) are searched, and when the search range R is narrow, for example, ± 8 around the search origin (0,0). If a pixel is searched, the ratio of the wide search range to the narrow search range is 4: 1.
In such a case, the search density D is determined so that the ratio before and after switching of the search density D of the motion vector MV is opposite to the ratio before and after switching of the search range R.
For example, if the search density D for searching in a wide range is determined to be a density for searching every 4 pixels, and the search density D for searching in a narrow range is determined to be a density for searching every 2 pixels, the search density D is switched. The ratio before and after becomes 1: 2, and the search point C of the motion vector MV is the same before and after switching.

When the search range R of the motion vector MV is a narrow range, the processing time is reduced when the same search method is used as compared with the case where the search range R is a wide range.
Generally, increasing the search density D is expected to improve the accuracy of motion prediction and improve the image quality.
As described above, when the search density D is determined according to the search range R, the number C of search points does not change, so that the accuracy of motion prediction can be improved and the image quality can be improved without increasing the processing time.

The motion search unit 22 of the motion prediction unit 20 collates the input image signal with the reference image signal stored in the frame memory 8, and starts from the search range (wide range or narrow range) determined by the search range / density determination unit 21. Search for an appropriate motion vector MV.
That is, the motion search unit 22 selects an arbitrary motion vector from among a plurality of motion vectors within the search range (candidate motion vectors of the macroblock), and encodes the input image signal by the amount of the motion vector. Move the target macroblock.
However, when an arbitrary motion vector is selected from a plurality of motion vectors within the search range, the motion vector is selected according to the search density D determined by the search range / density determination unit 21.

When the motion search unit 22 moves the macro block to be encoded of the input image signal, the motion search unit 22 collates the moved macro block with the reference image signal and evaluates whether or not they are similar.
As an evaluation method, for example, SAD which is a sum of absolute difference can be used. Alternatively, the sum of squared differences can be used.

The motion search unit 22 sequentially selects all motion vectors within the search range, moves the macroblock to be encoded by the motion vector, and collates the moved macroblock with the reference image signal. Evaluate whether the two are similar.
The motion search unit 22 uses the motion vector having the highest evaluation result (for example, the motion vector having the smallest SAD) among the plurality of motion vectors in the search range as an appropriate motion vector MV, The data is output to the variable length coding unit 4.

  As is apparent from the above, according to the second embodiment, the search range / density determination unit 21 determines the search density D of the motion vector MV according to the size of the search range R of the motion vector MV. Thus, for example, when the search range R of the motion vector MV is switched to a narrow range, the search density D can be increased without increasing the amount of calculation, and compared with a conventional encoding device, There is an effect that the accuracy of the motion vector can be improved and the image quality can be improved.

Embodiment 3 FIG.
FIG. 6 is a block diagram showing a moving picture coding apparatus according to Embodiment 3 of the present invention. In the figure, the same reference numerals as those in FIG.
Similar to the motion prediction unit 9 in FIG. 1, the motion prediction unit 30 evaluates the size of the prediction vector calculated when the encoded data is generated by the variable length encoding unit 4 and determines the motion vector search range. The input image signal and the reference image signal stored in the frame memory 8 are collated, and processing for searching for an appropriate motion vector from the search range is performed.
However, unlike the motion prediction unit 9 in FIG. 1, the motion prediction unit 30 determines a motion vector search range in consideration of the frame rate and resolution of the input image signal. The motion prediction unit 30 constitutes a motion vector search means.
In the third embodiment, the motion prediction unit 30 determines the motion vector search range, but the motion vector search density is also determined in the same manner as the motion prediction unit 20 of FIG. Also good.

FIG. 7 is a block diagram showing the motion prediction unit 30 of the moving picture coding apparatus according to Embodiment 3 of the present invention. In the figure, the same reference numerals as those in FIG.
Similar to the search range determination unit 11 in FIG. 2, the search range determination unit 31 performs motion when the size of the prediction vector calculated when the encoded data is generated by the variable length encoding unit 4 is larger than the threshold δ. A vector search range is determined over a wide range, and if the size of the prediction vector is smaller than the threshold δ, the motion vector search range is determined as a narrow range.
However, unlike the search range determination unit 11 in FIG. 2, the search range determination unit 31 determines the search range of the motion vector in consideration of the frame rate and resolution of the input image signal.

Next, the operation will be described.
The third embodiment is the same as the first embodiment except that the motion prediction unit 30 is mounted instead of the motion prediction unit 9 in FIG. 1, and therefore only the operation of the motion prediction unit 30 is performed. explain.
The motion prediction unit 30 calculates the prediction vector PMV when the variable length encoding unit 4 generates encoded data, evaluates the magnitude of the prediction vector PMV, determines the search range of the motion vector MV, and inputs The image signal and the reference image signal stored in the frame memory 8 are collated, and an appropriate motion vector MV is searched from the search range.
However, the search range determination unit 31 determines the search range of the motion vector MV in consideration of the frame rate and resolution of the input image signal.
Hereinafter, the processing content of the motion estimation part 30 is demonstrated concretely.

When the frame rate of the input image signal is small or when the resolution is large, the magnitude of the motion vector between frames is expected to be large.
Conversely, when the frame rate of the input image signal is large or the resolution is small, it is expected that the size of the motion vector between frames will be small.
Therefore, in the third embodiment, the search range of motion vectors is switched according to the frame rate and resolution.

When the search range determination unit 31 of the motion prediction unit 30 inputs information indicating the frame rate and resolution of the input image signal, an index for determining the search range from the frame rate and resolution as shown in the following equation (3). M is calculated.
M = (R _a × R _b ) / F (3)
Here, F is the frame rate, R _a × R _b is the resolution (R _a is the number of macroblocks in the vertical direction in the screen, and R _b is the number of macroblocks in the horizontal direction in the screen).
The value of the index M for determining the search range increases as the frame rate F decreases, and increases as the resolution R _a × R _b increases.

When the search range determination unit 31 calculates the index M for determining the search range, when the value of the index M for determining the search range is large, for example, as shown in the following formula (4), a wide range of search ranges Is set to a small value so that is easily selected.
On the other hand, when the value of the index M is small, for example, as shown in the following formula (4), the threshold δ is set to a large value so that a narrow search range can be easily selected.
δ = a / M (4)
However, a is a preset constant.

When the threshold δ is set as described above, the search range determination unit 31 calculates the magnitude | PMV | of the prediction vector PMV calculated by the variable length encoding unit 4.
As a method of calculating the magnitude | PMV | of the prediction vector PMV, for example, the absolute value of the horizontal component and the vertical component of the prediction vector PMV are calculated, and the larger absolute value is used as the magnitude of the prediction vector PMV. A method of setting PMV | is conceivable.
Alternatively, a method is conceivable in which the sum of squares of the horizontal and vertical components of the prediction vector PMV is calculated and the sum of the squares is used as the magnitude | PMV | of the prediction vector PMV.

The search range determination unit 31 compares the magnitude | PMV | of the prediction vector PMV with the threshold δ, and if the magnitude | PMV | of the prediction vector PMV is larger than the threshold δ, the search range of the motion vector MV is determined over a wide range. To do.
On the other hand, when the magnitude | PMV | of the prediction vector PMV is smaller than the threshold δ, the motion vector search range MV is determined to be a narrow range (wide range> narrow range).
| PMV | ≧ δ → Wide range | PMV | <δ → Narrow range

  In this way, the search range determination unit 31 determines the search range of the motion vector MV in consideration of the frame rate and the resolution. For example, in the case of VGA 30 fps, assuming that the surrounding ± 16 pixels is the search range, QCIF 30 fps, etc. When the resolution is small as described above, it is determined that the surrounding ± 8 pixels are set as the search range. In addition, when the frame rate is small, such as VGA 7.5 fps, it is determined that the surrounding ± 32 pixels are set as the search range.

  Similar to the first embodiment, the motion search unit 12 of the motion prediction unit 30 collates the input image signal with the reference image signal stored in the frame memory 8, and determines the search range determined by the search range determination unit 31. An appropriate motion vector MV is searched from (wide range or narrow range).

  As apparent from the above, according to the third embodiment, the motion prediction unit 30 is configured to determine the motion vector search range in consideration of the frame rate and resolution of the input image signal. There is an effect that a more appropriate search range can be determined than in the first form.

  In the third embodiment, the threshold value δ is set according to the frame rate and resolution. However, as in the first embodiment, the size of the prediction vector is evaluated to determine the motion vector search range. After the determination, the search range may be corrected according to the frame rate and resolution, and the same effect can be obtained.

Embodiment 4 FIG.
FIG. 8 is a block diagram showing a moving picture coding apparatus according to Embodiment 4 of the present invention. In the figure, the same reference numerals as those in FIG.
Similar to the motion prediction unit 9 in FIG. 1, the motion prediction unit 40 evaluates the size of the prediction vector calculated when the encoded data is generated by the variable length encoding unit 4 and determines the motion vector search range. The input image signal and the reference image signal stored in the frame memory 8 are collated, and processing for searching for an appropriate motion vector from the search range is performed.
However, unlike the motion prediction unit 9 in FIG. 1, the motion prediction unit 40 has a plurality of search range candidates, and the prediction vector calculated when the encoded data is generated by the variable length encoding unit 4. The size is compared with a plurality of thresholds, a search range candidate corresponding to the comparison result is selected from a plurality of search range candidates, and the search range candidate is determined as a search range of a motion vector. Note that the motion prediction unit 40 constitutes a motion vector search means.
In the fourth embodiment, the motion prediction unit 40 determines the motion vector search range, but the motion vector search density is also determined in the same manner as the motion prediction unit 20 of FIG. Also good.

FIG. 9 is a block diagram showing a motion prediction unit 40 of a moving picture coding apparatus according to Embodiment 4 of the present invention. In the figure, the same reference numerals as those in FIG.
Search range determining unit 41 variable-length coding unit 4 by a plurality of threshold values [delta] ₁ the magnitude of the prediction vector encoded data is calculated when it is generated, [delta] _2, as compared ..., and [delta] _N , R _N , a search range candidate corresponding to the comparison result is selected from the search range candidates R ₁ , R ₂ ,.

Next, the operation will be described.
Since the fourth embodiment is the same as the first embodiment except that the motion prediction unit 40 is mounted instead of the motion prediction unit 9 in FIG. 1, only the operation of the motion prediction unit 40 is performed. explain.
When the motion prediction unit 40 calculates the prediction vector PMV when the variable length encoding unit 4 generates encoded data, the motion prediction unit 40 compares the magnitude of the prediction vector PMV with a plurality of thresholds, and calculates a plurality of search range candidates. A search range candidate corresponding to the comparison result is selected, and the search range candidate is determined as a motion vector search range. Then, the input image signal and the reference image signal stored in the frame memory 8 are collated, and an appropriate motion vector MV is searched from the search range.
Hereinafter, the processing content of the motion estimation part 40 is demonstrated concretely.

When the variable length coding unit 4 calculates the prediction vector PMV, the search range determination unit 41 of the motion prediction unit 40 calculates the magnitude | PMV | of the prediction vector PMV.
As a method of calculating the magnitude | PMV | of the prediction vector PMV, for example, the absolute value of the horizontal component and the vertical component of the prediction vector PMV are calculated, and the larger absolute value is used as the magnitude of the prediction vector PMV. A method of setting PMV | is conceivable.
Alternatively, a method is conceivable in which the sum of squares of the horizontal and vertical components of the prediction vector PMV is calculated and the sum of the squares is used as the magnitude | PMV | of the prediction vector PMV.

Search range determining unit 41 holds a plurality of search ranges candidate R _1, R _2, · · ·, and R _N, a plurality of threshold values δ _1, δ _2, ···, and [delta] _N.
Search range determining section 41, the magnitude of the prediction vector PMV | calculating the, its magnitude | | PMV comparing multiple thresholds δ _1, δ _2, ···, and [delta] _N | PMV.
Search range determining section 41, as shown below, a plurality of search ranges candidate R _1, R _2, · · ·, among R _N, and select the search range candidate R _n corresponding to the comparison result, the motion search range candidate R _n to determine the search range of the vector.

| PMV | <δ ₁ → R ₁
δ ₁ ≦ | PMV | <δ ₂ → R ₂
δ ₂ ≦ | PMV | <δ ₃ → R ₃
:
:
δ _n-1 ≦ | PMV | <δ _n → R _n
:
:
_{δ N-1 ≦ | PMV |} <δ N → R N
However, δ ₁ <δ _2, a ··· <δ _N, is _{R 1 <R 2 <··· <} R N.

Accordingly, for example, the search range determination unit 41 selects the search range candidate R ₂ if δ ₁ ≦ | PMV | <δ ₂ is satisfied, and searches for the search range if δ ₂ ≦ | PMV | <δ ₃ is satisfied. If candidate R ₃ is selected and δ _n-1 ≦ | PMV | <δ _n is satisfied, search range candidate R _n is selected, and the selected search range candidate is determined as the search range of the motion vector.

  Similar to the first embodiment, the motion search unit 12 of the motion prediction unit 40 collates the input image signal with the reference image signal stored in the frame memory 8 and determines the search range determined by the search range determination unit 41. To search for an appropriate motion vector MV.

As is apparent from the above, according to the fourth embodiment, the size of the prediction vector calculated when the motion prediction unit 40 generates encoded data by the variable length encoding unit 4 is set to a plurality of threshold values δ. _1, [delta] _2, · · ·, as compared to the [delta] _N, a plurality of search ranges candidate R _1, R _2, · · ·, among R _N, and select the search range candidate corresponding to the comparison result, Since the search range candidate is determined to be the search range of the motion vector, there is an effect that a more appropriate search range can be determined than in the first embodiment.

Embodiment 5 FIG.
10 is a block diagram showing a moving picture coding apparatus according to Embodiment 5 of the present invention. In the figure, the same reference numerals as those in FIG.
The motion prediction unit 50 calculates the target processing time T _b from the frame rate and resolution of the input image signal, calculates the actual processing time T _c from the encoding processing start time to the current macrobook processing start time, and the target processing A process for setting a threshold value δ from the time T _b and the actual measurement processing time T _c is performed, and is calculated when the encoded data is generated by the variable length encoding unit 4 as in the motion prediction unit 9 of FIG. The motion vector search range is determined by evaluating the size of the predicted vector, the input image signal and the reference image signal stored in the frame memory 8 are collated, and an appropriate motion vector is searched from the search range. Perform the process. Note that the motion prediction unit 50 constitutes a motion vector search means.
In the fifth embodiment, the motion prediction unit 50 determines the motion vector search range. However, the motion vector search density is also determined in the same manner as the motion prediction unit 20 in FIG. Also good.

FIG. 11 is a block diagram showing a motion prediction unit 50 of a moving picture coding apparatus according to Embodiment 5 of the present invention. In the figure, the same reference numerals as those in FIG.
The search range determination unit 51 calculates the target processing time T _b from the frame rate and resolution of the input image signal, calculates the actual processing time T _c from the encoding processing start time to the current macrobook processing start time, and sets the target processing time T _c. When the process of setting the threshold value δ from the processing time T _b and the actually measured processing time T _c is performed, and the encoded data is generated by the variable length encoding unit 4 as in the search range determination unit 11 of FIG. If the size of the predicted vector calculated in the above is larger than the threshold δ, the motion vector search range is determined over a wide range. If the size of the predicted vector is smaller than the threshold δ, the motion vector search range is narrower than the wide range. A process for determining a narrow range is performed.

Next, the operation will be described.
Since the fifth embodiment is the same as the first embodiment except that the motion prediction unit 50 is mounted instead of the motion prediction unit 9 of FIG. 1, only the operation of the motion prediction unit 50 is performed. explain.
The motion prediction unit 50 calculates the target processing time T _b from the frame rate and resolution of the input image signal, calculates the actual processing time T _c from the encoding processing start time to the current macrobook processing start time, and sets the target processing time T _c. A threshold δ is set from the processing time T _b and the measured processing time T _c . Also, the search range of the motion vector MV is determined by evaluating the size of the prediction vector PMV calculated by the variable length encoding unit 4, and the input image signal and the reference image signal stored in the frame memory 8 are collated. The appropriate motion vector MV is searched from the search range.
Hereinafter, the processing content of the motion estimation part 50 is demonstrated concretely.

When the search range determination unit 51 of the motion prediction unit 50 receives information indicating the frame rate and resolution of the input image signal, one macroblock is calculated from the frame rate and resolution as shown in the following equation (5). It calculates the processing time T _a to be subjected to.
T _a = 1000 / (F × R _a × R _b ) (5)
Here, F is the frame rate, R _a × R _b is the resolution (R _a is the number of macroblocks in the vertical direction in the screen, and R _b is the number of macroblocks in the horizontal direction in the screen).

Search range determining section 51 calculates calculating the processing time T _a to be applied to one macroblock, as shown in the following formula (6), a target processing time T _b which is subjected to the N macroblocks.
T _b = T _a × N (6)
When the search range determination unit 51 calculates the target processing time T _b to be applied to N macroblocks, if the target processing time T _b exceeds the actual processing time T _c measured by a timer, for example, Therefore, the preset threshold value δ is changed to a large value so that a narrow search range with a short processing time can be easily selected.
On the other hand, if the target processing time T _b does not exceed the actual measurement processing time T _c , the possibility that the processing will not be in time is small. Therefore, the threshold δ is set to a small value so that a wide search range with a long processing time can be easily selected. change. Alternatively, the preset threshold value δ is maintained without being changed.

When the threshold value δ is changed or maintained as described above, the search range determination unit 51 calculates the magnitude | PMV | of the prediction vector PMV calculated by the variable length encoding unit 4.
As a method of calculating the magnitude | PMV | of the prediction vector PMV, for example, the absolute value of the horizontal component and the vertical component of the prediction vector PMV are calculated, and the larger absolute value is used as the magnitude of the prediction vector PMV. A method of setting PMV | is conceivable.
Alternatively, a method is conceivable in which the sum of squares of the horizontal and vertical components of the prediction vector PMV is calculated and the sum of the squares is used as the magnitude | PMV | of the prediction vector PMV.

The search range determination unit 51 compares the magnitude | PMV | of the prediction vector PMV with a threshold δ, and if the magnitude | PMV | of the prediction vector PMV is larger than the threshold δ, the search range of the motion vector MV is determined over a wide range. To do.
On the other hand, when the magnitude | PMV | of the prediction vector PMV is smaller than the threshold δ, the motion vector search range MV is determined to be a narrow range (wide range> narrow range).
| PMV | ≧ δ → Wide range | PMV | <δ → Narrow range

As can be seen from the above description, according to the fifth embodiment, with the motion prediction unit 50 calculates the target processing time T _b from the frame rate and resolution of the input image signal, the current macroblock processing from the encoding processing start time Since the actual measurement processing time T _c up to the start time is calculated and the threshold value δ is set from the target processing time T _b and the actual measurement processing time T _c , it is possible to encode with a calculation amount corresponding to the processing time. There is an effect that can be done.

It is a block diagram which shows the moving image encoder by Embodiment 1 of this invention. It is a block diagram which shows the motion estimation part 9 of the moving image encoder by Embodiment 1 of this invention. It is explanatory drawing which shows the prediction vector of a current macroblock, and the motion vector of an adjacent macroblock. It is a block diagram which shows the moving image encoder by Embodiment 2 of this invention. It is a block diagram which shows the motion estimation part 20 of the moving image encoder by Embodiment 2 of this invention. It is a block diagram which shows the moving image encoder by Embodiment 3 of this invention. It is a block diagram which shows the motion estimation part 30 of the moving image encoder by Embodiment 3 of this invention. It is a block diagram which shows the moving image encoder by Embodiment 4 of this invention. It is a block diagram which shows the motion estimation part 40 of the moving image encoder by Embodiment 4 of this invention. It is a block diagram which shows the moving image encoder by Embodiment 5 of this invention. It is a block diagram which shows the motion estimation part 50 of the moving image encoder by Embodiment 5 of this invention.

Explanation of symbols

  1 differencer (image signal difference unit), 2 DCT transform unit (quantization unit), 3 quantization unit (quantization unit), 4 variable length encoding unit (encoded data generation unit), 5 inverse quantization unit ( Reference image signal generation means), 6 Inverse DCT conversion section (reference image signal generation means), 7 Adder (reference image signal generation means), 8 Frame memory (reference image signal generation means), 9 Motion prediction section (motion vector search) Means), 10 motion compensation unit (motion compensation unit), 11 search range determination unit, 12 motion search unit, 20 motion prediction unit (motion vector search unit), 21 search range / density determination unit, 22 motion search unit, 30 motion Prediction unit (motion vector search unit), 31 search range determination unit, 40 motion prediction unit (motion vector search unit), 41 search range determination unit, 50 motion prediction unit (motion vector search unit), 51 search Range determining unit.

Claims (6)

  Image signal difference means for obtaining a difference between the input image signal and the predicted image signal, and outputting a prediction residual signal that is a signal of the difference, and quantization for quantizing the prediction residual signal output from the image signal difference means Means, variable length coding of the prediction residual signal quantized by the quantization means to generate encoded data, and the prediction residual signal quantized by the quantization means is inverted. Reference image signal generation means for generating a reference image signal by adding the prediction residual signal after quantization and inverse quantization and the prediction image signal, and when the encoded data is generated by the encoded data generation means A motion vector search range is determined by evaluating the size of the calculated prediction vector, the input image signal and the reference image signal generated by the reference image signal generation means are collated, and the search range is calculated. Using a motion vector search means for searching for a positive motion vector, and a motion vector searched by the motion vector search means, a predicted image signal is generated from the reference image signal generated by the reference image signal generation means, A moving image encoding apparatus comprising: a motion compensation unit that outputs a predicted image signal to the image signal difference unit and the reference image signal generation unit.   The motion vector search means determines a motion vector search range over a wide range when the size of the prediction vector calculated when the encoded data is generated by the encoded data generation means is larger than a predetermined threshold, 2. The moving picture coding apparatus according to claim 1, wherein when the vector size is smaller than a predetermined threshold, the motion vector search range is determined to be narrower than the wide range.   3. The moving picture coding apparatus according to claim 1, wherein the motion vector search means determines the search density of the motion vector according to the size of the search range of the motion vector.   4. The moving image according to claim 1, wherein the motion vector search means determines a motion vector search range in consideration of a frame rate and resolution of an input image signal. Encoding device.   When a plurality of search range candidates are prepared, the motion vector search unit compares the size of the prediction vector calculated when the encoded data is generated by the encoded data generation unit with a plurality of thresholds, 2. The moving picture coding apparatus according to claim 1, wherein a search range candidate corresponding to the comparison result is selected from a plurality of search range candidates, and the search range candidate is determined as a search range of a motion vector. .   The motion vector search means calculates the target processing time from the frame rate and resolution of the input image signal, calculates the actual processing time from the encoding processing start time to the current macrobook processing start time, and calculates the target processing time and the above 3. The moving picture encoding apparatus according to claim 2, wherein a threshold value is set from an actual measurement processing time.

Images