JP3234544B2 - Image encoding apparatus and image encoding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding apparatus and an image encoding method, and more particularly to setting of an f-code value which is a parameter used when encoding a motion vector in MPEG2.

[0002]

2. Description of the Related Art Conventionally, in a system for transmitting a moving image signal, a technique for compressing and encoding the image signal has been developed in order to use a transmission line efficiently, and in particular, the temporal redundancy of the image signal can be reduced. Inter-picture prediction coding is widely used. The inter-picture predictive coding is a method of obtaining difference data between two temporally different images and a motion vector representing a motion in the image, and encoding and transmitting the difference data and the motion vector.

FIG. 12 schematically shows a method of calculating a motion vector used for predictive coding. In the figure, (a) is a processed image to be predictively coded, and (b)
Is a reference image serving as a reference. The processed image is divided into a plurality of macroblocks (16 × 16 pixels), a macroblock having a high degree of similarity to an arbitrary processed macroblock 100 is searched for in the reference image, and the displacement between the two is determined by the motion vector
Calculated as 2.

FIG. 13 schematically shows a block matching method used when calculating the motion vector 102. In the figure, (a) is a processed image, (b)-
(D) is a reference image. First, as shown in (b),
An appropriate detection range 104 is determined in the reference image, and the processing macroblock 100 is shifted rightward by one pixel within this detection range 104. After the search in one row is completed, the search in the left and right direction is performed again by shifting one pixel downward as shown in FIG. The above process is repeated, and all searches within the detection range 104 are finally performed (d). In each search, the difference between the processing macroblock 100 and the block of the reference image is evaluated using a predetermined evaluation function,
By calculating the position where the evaluation value becomes minimum, the motion vector of the processing macro block 102 is calculated.

In such a block matching method, in consideration of the fact that the distance between the reference image and the processed image generally increases, that is, as the time difference between the two images increases, the motion vector also increases. The detection range 104 is changed according to the distance between images.

FIG. 14 exemplifies a detection range 104 set for a reference image when the distance between images (distance between frames) is 1, 2, or 3. In the figure, (a) is a detection range when the inter-frame distance is 1, and ± 15 pixels at the top, bottom, left, and right around the reference image position corresponding to the position of the processing macroblock. (B) shows the detection range when the inter-frame distance is 2, and the detection range is ± 31 pixels centered on the position of the processing macroblock.
(C) shows a case where the distance between frames is 3, and ± 47 pixels are the detection range. That is, the detection range of the motion vector is set to be larger as the inter-image distance increases.

Of course, such a detection range can be set according to the nature of the image. For example, when predictive coding is performed on an image having a large motion, the detection range is set smaller than that of an image having a small motion. Can be set larger.

FIG. 15 shows the relationship between the inter-frame distance between the reference image and the processed image and the detection range 104. The relationship between the general motion image and the large motion image is shown. is there. For example, if the distance between frames is 2,
The detection range 104 is ± 31 pixels for a general motion image, and the detection range 104 for a large motion image.
Is ± 47 pixels.

When the motion vector is calculated by setting the detection range 104 as described above, motion vector data is obtained for each macroblock of the processed image, but the motions of adjacent macroblocks are generally considered to be similar. Therefore, instead of encoding the motion vector value itself, a difference from the motion vector of an adjacent macroblock is obtained, and the difference motion vector is encoded, so that the compression can be efficiently performed.

FIG. 16 shows adjacent macroblocks of a processed image and motion vector components calculated for each macroblock. In the figure, macroblocks MBi-1, MBi, MBi + 1 are adjacent macroblocks, and the motion vector components of MBi-1 are (xi-1,
yi-1), the motion vector components of MBi are (xi, yi), M
Assuming that the motion vector component of Bi + 1 is (xi + 1, yi + 1), the differential motion vector of MBi, which is the i-th macroblock, is expressed by the following equation.

[0011]

## EQU1 ## ΔMVi = MVi−MVi−1 (Δxi = xi−xi−
1, Δyi = yi−yi−1) When the differential motion vector is obtained as described above, the differential motion vector is then subjected to variable length coding (VLC).

The VLC of the differential motion vector is a fixed VL
Part coded by C table (upper bit)
And a portion (lower-order bit) whose code length is determined by the f-code. That is, the f code is
This parameter is used when encoding a motion vector (difference motion vector).
By dividing C into a fixed VLC table portion and an additional bit portion, the redundancy of the encoded data is reduced.

FIGS. 17 and 18 show VLC examples of differential motion vectors in a table format. FIG. 17 shows VLC when the f-code is 1, and the vector difference value (Δ) is within ± 31, that is, the vector value itself is −16 to +15.
VLC in the case of Therefore, the VLC shown in FIG. 17 can be used when the detection range when the motion vector is detected is smaller than these values. Since the value to be encoded is in units of half pels (half pixels),
Expressed in integer pixel units, it is -8 to +7.5 pixels. In FIG. 17, for example, the case where the difference value is 1 and -31
Although the same VLC is used in the case of, the vector value is in the range of −16 to +15 as described above, so that it can be uniquely determined if the vector value of the immediately preceding adjacent macroblock is known. . For example, if the vector value of the immediately preceding adjacent macroblock is 14, the vector difference value is 1 instead of −31.

FIG. 18 shows the VL when the f code is 2.
C, and is a VLC when the vector difference value (Δ) is within ± 63. As is apparent from a comparison between FIG. 17 and FIG. 18, the upper bit portion of VLC in FIG.
Same as LC. That is, in FIG. 18, V in FIG.
There is a relationship that lower one bit is added to LC.
When motion detection is performed in a wider detection range, a large motion vector value may be detected, so that the lower-order bit portion becomes longer (that is, the f-code value increases). However, the longer the lower bit portion, the longer the entire code length, and the lower the coding efficiency. Therefore, the f-code value needs to be set in consideration of the balance between the improvement of the prediction efficiency and the decrease of the coding efficiency with the increase of the detection range.

FIG. 19 shows the relationship between the detection range and the prediction efficiency, and the relationship between the detection range and the coding efficiency of the motion vector. As the detection range becomes larger, the similarity can be made larger, so that the prediction efficiency is improved, but at the same time, the difference value becomes larger, so that a larger number of lower-order bits are required and the coding efficiency is reduced. Conventionally, it is considered that such a decrease in coding efficiency causes deterioration in image quality of a decoded image, and a minimum value that can be obtained according to a detection range is used as an f-code value at the time of coding.

FIG. 20 shows f-code values of the conventional MPEG standard defined as described above. In addition,
When encoding an interlaced image with a frame structure, there are two types of motion vectors, a frame vector and a field vector. These two types of motion vectors have different distance expressions in the vertical direction. In FIG. 20, two types of detection ranges exist for the same f-code value corresponding to the frame vector and the field vector. Specifically, for example, in the case of frame prediction coding, the f-code is 1 when the detection range is -8 to +7.5, and the f-code is 2 when the detection range is -16 to +15.5. .

[0017]

As described above, the conventional M
In the PEG standard, a predetermined f-code corresponding to the detection range is defined, and as the f-code value increases, the code length increases. Therefore, it is usual to use the smallest f-code value that is allowed. . Therefore, when encoding a moving image, the detection range is changed in accordance with the distance between the reference image and the processed image, and the VLC is set to the minimum f-code value defined in accordance with each detection range. Therefore, a different VLC is required for each detection range, and as a result, there is a problem that the configuration of the variable length coding circuit is complicated.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems of the related art, and has as its object to enable a common encoding circuit regardless of a motion vector detection range when encoding a motion vector. An object of the present invention is to provide an image encoding device and an image encoding method that can simplify the circuit configuration.

[0019]

According to a first aspect of the present invention, there is provided an image encoding apparatus for encoding a motion vector of an input image using an f-code. It is characterized by having encoding means for encoding a motion vector using a value larger than an f-code value predetermined according to a motion detection range. Here, "f code"
Is a parameter used when encoding a motion vector in MPEG2, and the “predetermined f-code value according to the motion detection range” means the minimum f-code value allowed in the MPEG standard. I do. MPEG
By setting a value larger than the minimum f-code allowed in the standard, it is possible to cope with a wider variety of motion vector detection ranges.

According to a second aspect of the present invention, in the first aspect, a value larger than the f-code value is 4 to 5. By fixing the f-code value to 4 or 5, the coding configuration can be simplified.

In a third aspect based on the first and second aspects, the motion detection range is three times the size of the input image.
% To 4%. By fixing the motion detection range to the above range, the calculation time can be reduced while maintaining the image quality of the decoded image.

In a fourth aspect based on the first to third aspects, the motion detection range is 15 × 15 pixels or 3 pixels.
It is one of 1 × 15 pixels. By limiting to this range, it is possible to balance the improvement of the image quality and the reduction of the calculation time as in the third invention.

According to a fifth aspect of the present invention, there is provided an image encoding method for encoding a motion vector of an input image using an f-code, wherein the minimum value is determined among the motion vector detection ranges in the input image. Is characterized in that the motion vector is encoded using a value larger than the f-code value. MPEG
By setting a value larger than the minimum f-code allowed in the standard, it is possible to cope with a wider variety of motion vector detection ranges.

In a sixth aspect based on the fifth aspect, the value larger than the f-code value is 4 to 5. By fixing the f-code value to 4 or 5, it is possible to encode a motion vector more easily than before.

In a seventh aspect based on the fifth and sixth aspects, the motion detection range is fixed at 3% to 4% of the size of the input image. By fixing the motion detection range to the above range, the calculation time can be reduced while maintaining the image quality of the decoded image.

In an eighth aspect based on the fifth to seventh aspects, the motion detection range is 15 × 15 pixels or 3 pixels.
It is characterized by 1 pixel × 15 pixels. By fixing the motion detection range to the above range, the calculation time can be reduced while maintaining the image quality of the decoded image, as in the seventh aspect.

[0027]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows MPEG2 in the present embodiment.
1 shows the overall configuration of an image encoding device. The image rearranging circuit 110 is a circuit that rearranges the screen order of the input moving image signal. In other words, when a temporally later image is used as a reference image, the latter image needs to be stored in the memory first, so that the images are rearranged so that the latter image is processed first. Done. An image to be inter-frame predictively coded with reference to a previous image is a P picture, and an image to be inter-frame predictively coded with reference to a previous image or a subsequent image is a B picture.
An image that is not subjected to inter-frame prediction coding using a picture and a reference image is defined as an I picture.

The scan conversion macroblock conversion circuit 112
A circuit that divides each image into blocks of 8 × 8 pixels. Motion vectors and the like are calculated in units of four macroblocks, and DCT and quantization are performed in units of blocks.

The subtractor 114 is a circuit for subtracting each pixel data of the reference image from each pixel data of the current image supplied from the scan conversion macroblocking circuit 112. The signal is supplied to a circuit 116 and subjected to DCT (discrete cosine transform).

The DCT circuit 116 has a block unit (8 ×
8 × 8 × 8 × 8 × 8 × 8 × 8 × 8 × 8 × 8 × 8
Each block of pixels is converted into a coefficient matrix Cij of 8 rows and 8 columns of low frequency terms to high frequency terms. Each coefficient is represented by a quantization circuit 11
8 is supplied.

The weighted quantization circuit 118 is a circuit that divides each coefficient Cij by Qij and rounds the remainder to quantize and reduce the data amount. The quantization step width is supplied from the bit rate control circuit 124.

After the quantization, the coefficient data output in the order of the low frequency term to the high frequency term is subjected to variable length coding (VLC) by the variable length coding circuit 120, and the data amount is further reduced. The data after the variable length encoding is stored in the buffer memory 1
After being temporarily stored in the memory 22, the data is read out at a predetermined bit rate and output as a bit stream.

Inverse quantization circuit 126, inverse DCT circuit 128
Is a local encoder for reproducing the image data of the previous image and the subsequent image to serve as a reference image,
The reproduced image is stored in the image memory 132 and output to the subtractor 114. When the image data decoded by the local decoders 126 and 128 is difference data, the adder 130 is a circuit for adding reference image data with motion compensation to the difference data to complete an image. .

The image memory 132 is a memory for storing image data for at least two screens, and image data for reference is output from the terminal 132b in macroblock units. Further, image data for calculating a motion vector is supplied to the motion detection circuit 138 from the terminal 132a.

The motion detection circuit 138 detects a reference macroblock most similar to the current macroblock in the current image, and detects the displacement as a motion vector. The detected motion vector is stored in a variable length coding (VLC) circuit 1
20 and encoded with the determined f-code value.

Further, the motion compensation circuit 134 outputs an area (reference macroblock area) specified by the motion vector information from the terminal 132b of the image memory 132. As a result, the image data of the reference macroblock is
4 and the difference between the current macroblock and the reference macroblock is calculated as described above. Note that the processing of the motion compensation circuit 134 is executed with reference to the macroblock type information MBT supplied from the mode determination circuit 136. That is, the selection of outputting the previous image, outputting the subsequent image, outputting the previous image and the subsequent image, or not outputting the image from the image memory 132 is performed based on the MBT.

In the configuration of the encoding circuit as described above,
Four types of images having different motions are prepared, and these images are coded by changing the detection range of the motion vector and the f-code value in various ways. Further, the coded images are decoded by a decoding device similar to the conventional one. The image quality was evaluated. The images used for the evaluation are as follows.

Image A: Image with movement of human and car passing each other Image B: Image of tunnel exit seen from running vehicle Image C: Image with fine object movement Image D: Multiple people at high speed The image moves randomly, and particularly the horizontal motion is large. Comparing the amounts of motion of these images, it can be roughly found that the image C <
Image B <Image A <Image D The size of each image is 704 × 480 (pixels), and the bit rate is 3M.
bps, the picture structure is a frame structure, and the prediction method is frame motion compensation.

FIG. 2 shows the evaluation result of the image A. Note that the detection range is simply expressed as “7 × 7 pixels” for −8 to +7 pixels, “15 × 15 pixels” for −16 to +15 pixels, and “31 × 31 pixels” for −32 to +31 pixels. Is represented in In the figure, the horizontal axis is the f-code value, and the vertical axis is the S / N of the luminance of the decoded image. The broken line a has a detection range of 7 × 7 pixels S / N, the broken line b has a detection range of 15 × 15 pixels S / N, the broken line c has 31 × 31 pixels S / N, and the broken line d has 63 × 63. S / N of pixel, broken line e is 1
The S / N of 27 × 127 pixels is shown. Focusing on the polygonal line a (7 × 7 pixels), the minimum f-code value conventionally used in the MPEG standard is 1, which is equal to or greater than the S / N in the case where the f-code value is 2 to 5, Is obtained. Also, the folded line b (15 × 1
5 pixels), the minimum f-code value used in the MPEG standard is 2, and when the f-code value is 3 to 5, an S / N equal to or higher than that in this case is obtained. ing.

FIG. 3 shows an evaluation result of the image B. As in FIG. 2, the broken line a indicates that the detection range is 7 × 7 pixels.
/ N, broken line b shows the S / N of the detection range of 15 × 15 pixels, and broken line c shows the S / N of the detection range of 31 × 31 pixels. Also in this image B, focusing on the polygonal line a, the minimum f-code value 1 conventionally used in the MPEG standard
When the f-code value is 2 to 5, the same or higher S / N is obtained than the S / N in the case of. Also at the broken line b, when the f-code value is 3 to 5, an S / N equal to or higher than that when the minimum f-code value is 2 conventionally used in the MPEG standard is 2 is obtained. Have been.

FIG. 4 shows the evaluation result of the image C. The broken line a shows the S / N of the detection range of 7 × 7 pixels, the broken line b shows the S / N of the detection range of 15 × 15 pixels, and the broken line c shows the S / N of the detection range of 31 × 31 pixels. Also in this image C, the S / N of the broken line a is equal to or higher than that of the conventional f code value of 1 to 2 when the f code value is 2 to 5, and the f code value of the broken line b is the same as the conventional f code value. Compared with the case of 2, the S / N equal to or higher than that of the case of 3 to 5 was obtained.

FIG. 5 shows the evaluation result of the image D. The broken line a has a detection range of 7 × 7 pixels S / N, the broken line b has a detection range of 15 × 15 pixels S / N, the broken line c has a detection range of 31 × 31 pixels S / N, and the broken line d has a detection range. Is 63 ×
S / N of 63 pixels, broken line e has a detection range of 127 × 127
The S / N of each pixel is shown. In this image D as well, the S / N of the broken line a is equal to or higher than that of the conventional case where the f code value is 2 to 5 when the f code value is 2 to 5, and the f code value is the same as that of the broken line b in the conventional case. As compared with the case of 3, the S / N equal to or higher than that of the case of 3 to 5 is obtained. further,
Even in the case of the broken line c, an S / N equal to or higher than that of the case where the minimum f-code value conventionally used in the MPEG standard is 3 or 4 is 5 or 5 is obtained.

FIG. 6 shows another evaluation result of the image D. In the figure, a broken line a is an S / N of a detection range of 7 × 3 pixels (−8 to +7 in the horizontal direction, −4 to +3 in the vertical direction), and a broken line b.
Indicates the S / N of the detection range of 15 × 7 pixels, and the broken line c indicates the S / N of the detection range of 31 × 15 pixels. The detection range is set in consideration of the property of the image D, that is, the property that the horizontal movement is particularly large. Even when such a detection range is set, by setting an f-code value larger than the minimum f-code value used in the conventional MPEG standard, the same or higher S-code value can be obtained.
/ N is obtained. That is, in the case of the polygonal line a, an S / N equal to or higher than that in the case where the f-code value is 2 is obtained when the f-code value is 2 to 5 as compared with the case where the f-code value is 1. S / S equal to or greater than 3/5
N has been obtained. Further, in the case of the broken line c, the S / N is particularly large as compared with the other detection ranges, and the S / N equal to or higher than that in the case where the f code value is 3 is 4 or 5 as compared with the case where the f code value is 3. .

As is apparent from the above evaluation results, when the f-code value larger than the minimum f-code value used in the conventional MPEG standard is set, the image quality is improved without deterioration. Especially when the f-code value is 4
It can be seen that the tendency is remarkable in the cases of Nos. 1 to 5.

It is not true that the conventional idea that if the f-code value increases, the redundancy of data increases and the coding efficiency decreases, so that the image quality of the decoded image also deteriorates, is rather unreasonable. Means that it is preferable to set a value larger than the minimum value that can be taken according to the detection range (see FIG. 20). By setting the f-code value larger than the minimum value used in the conventional MPEG standard, the following advantages are obtained. That is, in the related art, since the minimum f-code value corresponding to the detection range is used, different VL values vary according to the detection range.
C (the length of the lower bit length is different), but f
By setting the code value to a relatively large value, particularly to 4 to 5, the detection range becomes 7 × 7 pixels, 15 × 15 pixels, 31 ×
Even if the number of pixels changes to 31, the common VLC can be performed using the common f-code value.

The reason why the quality of the decoded image is improved by setting the f-code value (particularly, 4 or 5) larger than the f-code value of the conventional MPEG standard is considered as follows. That is, the value to be actually coded is not the motion vector value as described above, but the difference between the motion vector value of the macroblock coded immediately before and the motion vector value of the macroblock to be coded. If the motion of the object or the camera is simple and the directions of the detected motion vectors are the same, the difference value becomes a value close to zero. On the other hand, in a sequence in which a plurality of objects move randomly, the absolute value of the difference increases. Therefore, the code length of the motion vector data is determined by the magnitude of the difference value and the f-code, and by setting the f-code value adapted to the difference value distribution, the coding efficiency is improved, and the image quality of the decoded image is also considered to be improved. Can be

FIGS. 7 to 12 show the images A, B,
The difference value distribution of C and D is shown. 7A and 7B show the difference value distribution of the image A, wherein FIG. 7A shows a difference value distribution having a detection range of 7 × 7 pixels, FIG. 7B shows a difference value distribution having a detection range of 15 × 15 pixels, and FIG. Is a difference value distribution of 31 × 31 pixels. In each figure, the horizontal axis is the difference value Δ, and the vertical axis is the number of distributions. As can be seen from the figure, there are many cases where the difference value is zero, and the difference value is concentrated on a place where the absolute value is small.

FIG. 8 shows the difference value distribution of the image B.
Is a difference value distribution having a detection range of 7 × 7 pixels, (b) is a difference value distribution having a detection range of 15 × 15 pixels, and (C) is a difference value distribution having a detection range of 31 × 31 pixels. As in the case of the image A, the image B often has a difference value of zero, and tends to concentrate on a place where the absolute value is small.

FIG. 9 shows the difference value distribution of the image C.
In the figure, (b) shows a difference value distribution having a detection range of 15 × 15 pixels, and (c) shows a difference value distribution having a detection range of 31 × 31 pixels. The difference value is less likely to be zero than in the case of the image A or B, and the difference value is also distributed where the absolute value is large.

FIG. 10 shows the difference value distribution of the image D.
(A) is a difference value distribution having a detection range of 7 × 7 pixels, (b) is a difference value distribution having a detection range of 15 × 15 pixels, and (c) is a difference value distribution having a detection range of 31 × 31 pixels. This image D
However, the difference value is less likely to be zero as compared with the images A and B, and the distribution increases near the area where the absolute value is large.

On the other hand, FIG. 11 shows the code length of a motion vector (actually, a difference motion vector) when the f-code values are 1 and 2. (A) shows the case where the f-code value is 1, the horizontal axis represents the difference value (Δ), and the vertical axis represents the code length. As shown in the figure, in order to reduce the code amount of the motion vector, the absolute value of the difference value needs to be 10 to 8 or less, or 25 or more. In particular, when the difference value becomes zero, it is very advantageous because at least two bits can be reduced as compared with other cases.

(B) shows a case where the f-code value is 2, and in order to reduce the amount of code of the motion vector, the absolute value of the difference value should be 20 or less, or 50 or more. .

Considering the relationship between the difference value and the code length and the difference value distribution shown in FIGS. 7 to 10, the following becomes clear. That is, in the image A, the difference value is extremely zero in many cases as compared with other images, so that the code length is small and the coding efficiency is good. On the other hand, the difference value distribution of the image D is flatter than the image A, and the number of cases where the difference value becomes zero is small. However, since the number of parts having a large absolute value of the difference value is large, this is useful for reducing the code amount.

Further, the number of the images C tends to decrease as the absolute value of the difference value increases, but the number when the difference value is zero is also small. For this reason, the code amount cannot be sufficiently reduced with the conventional f-code value. Therefore, by setting the f-code value to be larger than before, the code efficiency can be improved, and the S / N can be improved.

The image B is similar to the image A in that the number of cases where the difference value is zero is large, and is similar to the image C in other tendencies. Therefore, image B is composed of image A and image C.
By setting an f-code value larger than the f-code value in the conventional MPEG standard as in the case of the image C, the coding efficiency is improved.

As described above, by setting an f-code value larger than the f-code value of the conventional MPEG standard and performing predictive coding, a motion vector can be coded without deteriorating the image quality of a decoded image. It becomes possible. Preferably,
By setting the f-code value to 4 or 5, it is possible to encode a motion vector calculated from an arbitrary detection range while maintaining the S / N at or above that of the related art.

As a specific circuit configuration, the variable length encoding circuit 120 in FIG.
F code value (MP
Instead of performing VLC with the minimum value conventionally used in the EG standard), VLC is performed with a fixed f-code value. Therefore, the conventional variable-length encoding circuit 120 requires an arithmetic circuit for processing the f-code value because it is a variable, but the variable-length encoding circuit 120 according to the present embodiment requires such an arithmetic circuit. An arithmetic circuit is not required, and the configuration is simplified.

Although the detection range can be set according to the inter-frame distance, as is apparent from FIGS. 2 to 6, even if the detection range is limited to 15 × 15 pixels, a sufficient S / S ratio is obtained.
Since N is obtained, it is also preferable to fix the detection range to this value regardless of the inter-frame distance. Generally, when the area of the motion detection range is doubled, the amount of calculation for calculating the motion vector is doubled in proportion to the area. Therefore, by fixing the motion detection range to 15 × 15 pixels, it is possible to reduce the calculation processing time.

In particular, for an image having a large horizontal movement such as the image D, it is preferable to fix it to a horizontally long detection range such as 31 × 15 pixels from the viewpoint of S / N and calculation time. These 15 × 15 pixels or 31 × 15 pixels have a size corresponding to about 3 to 4% of the frame size. Generally, by restricting the motion detection range to this range, the calculation is performed without deteriorating the image quality of the decoded image. Time can be reduced.

As a specific circuit configuration, in FIG. 1, the motion detection circuit 138 inputs the motion detection range instead of detecting the motion vector by changing the motion detection range according to the inter-frame distance as in the related art. About 3% of the image size
The processing is fixed at a range corresponding to 4%. Therefore, the conventional motion detection circuit 138 requires a processing circuit for setting a motion detection range according to the inter-frame distance and a memory capacity for storing image data to be processed. In the case of 138, such a processing circuit is not required, and a fixed small-capacity memory is required, so that the configuration is simplified.

[0062]

As described above, according to the image coding apparatus and the image coding method of the present invention, the f-code value larger than that of the conventional MPEG standard, preferably 4
By fixing to 5 or less, a common encoding circuit can be used regardless of the motion detection range, and the device configuration can be simplified.

Further, by limiting the motion detection range to 3 to 4% of the frame size, it is possible to reduce the calculation processing time while maintaining the image quality of the decoded image.

[Brief description of the drawings]

FIG. 1 is an overall configuration diagram of an embodiment of the present invention.

FIG. 2 is a graph illustrating a relationship between an f-code and S / N of an image A according to the embodiment.

FIG. 3 is a graph showing the relationship between the f-code and S / N of an image B according to the embodiment.

FIG. 4 is a graph illustrating the relationship between the f-code and S / N of an image C according to the embodiment.

FIG. 5 is a graph illustrating a relationship between an f-code and an S / N of an image D according to the embodiment.

FIG. 6 is a graph showing the relationship between f code and S / N in another detection range of the image D according to the embodiment.

FIG. 7 is a graph illustrating a difference value distribution of an image A according to the embodiment.

FIG. 8 is a graph illustrating a difference value distribution of an image B according to the embodiment.

FIG. 9 is a graph showing a difference value distribution of an image C according to the embodiment.

FIG. 10 is a graph illustrating a difference value distribution of an image D according to the embodiment.

FIG. 11 is a graph showing a motion vector code length when an f-code value is 1 in the embodiment.

FIG. 12 is an explanatory diagram showing a motion vector calculation process.

FIG. 13 is an explanatory diagram of a block matching method.

FIG. 14 is an explanatory diagram showing a relationship between an inter-frame distance and a detection range.

FIG. 15 is a table showing the relationship between the inter-frame distance and the detection range.

FIG. 16 is an explanatory diagram of calculating a differential motion vector.

FIG. 17 is an explanatory diagram of VLC of a motion vector (f code =
1).

FIG. 18 is an explanatory diagram of VLC of a motion vector (f code =
2).

FIG. 19 is a graph showing a relationship between a detection range, prediction efficiency, and coding efficiency.

FIG. 20 shows a detection range and f in the conventional MPEG standard.
FIG. 4 is an explanatory diagram showing a relationship with a code value.

[Explanation of symbols]

110 image rearrangement, 112 scan conversion macroblock conversion, 116 DCT, 118 weighted quantization, 120
Variable length coding, 122 buffer, 124 bit rate control, 126 inverse quantization, 128 inverse DCT, 132
Image memory, 134 motion compensation, 136 mode determination, 138 motion detection.

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-6-334993 (JP, A) JP-A-7-274174 (JP, A) US Pat. No. 5,731,850 (US, A) Television Society, “MPEG” , First Edition, Ohmsha, April 1996, p. 94 -96 (58) Field surveyed (Int.Cl. ⁷ , DB name) H04N ^7/ 24-7/68 JICST file (JOIS)

Claims (10)

(57) [Claims] 1. An image encoding apparatus for encoding a motion vector of an input image using an f-code, comprising a plurality of f-code values allowed according to a motion vector value of a motion detection range in the input image. An image coding apparatus comprising: coding means for coding a motion vector using an f-code value larger than a minimum value. 2. The image coding apparatus according to claim 1, wherein the f-code value is a value that is 1 or 2 larger than a minimum value. 3. The image coding apparatus according to claim 1, wherein said f-code value is 4 to 5. 4. The image encoding apparatus according to claim 1, wherein the motion detection range is fixed to 3% to 4% of the size of the input image. 5. The image encoding apparatus according to claim 1, wherein the motion detection range is any one of 15 × 15 pixels and 31 × 15 pixels. 6. An image encoding method for encoding a motion vector of an input image using an f-code, comprising a plurality of f-code values allowed according to a motion vector value of a motion detection range in the input image. An image coding method characterized by coding a motion vector using an f-code value larger than a minimum value. 7. The image encoding method according to claim 6, wherein the f-code value is a value that is one or two times larger than a minimum value. 8. The image encoding method according to claim 6, wherein the f-code value is 4 to 5. 9. The image encoding method according to claim 6, wherein the motion detection range is fixed at 3% to 4% of the size of the input image. 10. The image encoding method according to claim 6, wherein the motion detection range is one of 15 × 15 pixels and 31 × 15 pixels.