JP3526258B2 - Image encoding device and image decoding device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention belongs to the field of digital image processing, and is an image coding apparatus for coding image data with high efficiency and an image decoding apparatus for decoding coded data created by this image coding apparatus. It relates to the device.

[0002]

2. Description of the Related Art In image coding, a method of synthesizing different moving image sequences has been studied. Document "Image Coding Using Hierarchical Representation and Multiple Templates"
In IE94-159, pp99-106 (1995)), a new moving image sequence that is the background and a moving image sequence of the component moving image that is the foreground (for example, a person image or a video of a fish cut out by chroma key technology) is synthesized and added. Techniques for creating simple sequences are described.

In addition, the document "Temporal Scalability based on image conte"
nt ", ISO / IEC / JTC1 / SC29 / WG11 MPEG95 / 211 (1995)) creates a new sequence by synthesizing a moving image sequence with a high frame rate and a moving image sequence with a high frame rate. The method is described.

In this system, as shown in FIG. 16, predictive coding is performed in a lower layer at a low frame rate.
In the upper layer, predictive coding is performed at a high frame rate only for the selected area (hatched portion). However, the frame encoded in the lower layer is not encoded in the upper layer, and the decoded image of the lower layer is directly copied and used. Further, it is assumed that a portion such as a person portion where the viewer's attention is focused is selected as the selection area.

FIG. 8 shows a block diagram of a conventional method. First, on the encoding side of the conventional method, the input moving image is frame-decimated by the first frame dropping unit 801 and the second frame dropping unit 802 so that the frame rate is equal to or lower than the frame rate of the input image, and then the upper layer encoding is performed. It is input to the unit 803 and the lower layer encoding unit 804. Here, the frame rate of the upper layer is equal to or higher than the frame rate of the lower layer.

In the lower layer encoding unit 804, the entire input moving image is encoded. As a coding method, for example, a moving picture coding international standardization method such as MPEG or H.261 is used. Further, in the lower layer encoding unit 804, a decoded image of the lower layer is created and used for predictive encoding, and at the same time, input to the combining unit 805.

FIG. 9 is a block diagram showing a code amount control unit in a conventional coding apparatus. In FIG. 9, the encoding unit 902 performs moving image encoding using motion compensation prediction, orthogonal transformation, quantization, variable length encoding, or the like.

The quantization width calculation unit 901 calculates the quantization width used by the encoding unit 902, and the generated code amount calculation unit 9
03 calculates the accumulation of encoded data. Generally, when the generated code amount becomes large, the quantization width is increased to suppress it, and conversely, when the generated code amount becomes small, the quantization width is controlled to be small.

In the upper layer coding unit 803 of FIG. 8, only the selected area of the input moving image is coded. Also in this case, the international standardization method of moving picture coding such as MPEG or H.261 is used, but only the selected area is coded based on the area information. However, the frame encoded in the lower layer is not encoded in the upper layer.

The area information is information indicating a selected area such as a person portion, and is, for example, a binary image having a value of 1 at the position of the selected area and a value of 0 at other positions. Further, in the upper layer encoding unit 803, only the selected area of the moving image is decoded,
It is input to the synthesis unit 805.

In the area information coding unit 806, the area information is coded using an eight-direction quantized code. As shown in FIG. 17, the 8-direction quantized code is a numerical value indicating the direction to the next point, and is generally used when expressing a digital figure.

The synthesizing unit 805 outputs the decoded image of the lower layer when the lower layer frame is encoded in the synthesizing target frame. When the lower layer frame is not encoded in the synthesis target frame, a moving image is output using the decoded images of the two encoded lower layers before and after the synthesis target frame and one upper layer decoded image. .

Here, the frames of the two images of the lower layer are before and after the frame of the upper layer. Also,
The moving image created by the combining unit 805 is input to the upper layer encoding unit 803 and used for predictive encoding. Synthesis part 8
The image creating method in 05 is as follows.

First, two lower layer interpolated images are created. The decoded image of the lower layer at time t is B (x, y,
t) (where x and y are coordinates representing the pixel position in space), and the times of the two lower layers are t1, t2, and
When the time of the upper layer is t3 (where t1 <t3 <t2 ), the interpolated image I (x, y, t3) at time t3 is I (x, y, t3).
y, t3) = [(t2-t3) B (x, y, t1) + (t3-t1) B (x, y, t2)
] / (t2-t1) Calculated by (1).

Next, the interpolated image I obtained above is combined with the decoded image E of the upper layer. For this, the region information M (x,
Create weight information W (x, y, t) for composition from y, t),
The synthetic image S is obtained by the following equation. S (x, y, t) = [1-W (x, y, t)] I (x, y, t) + E (x, y, t) W (x, y, t) (2) domain Information M (x, y, t) is 1 inside the selected area and 0 outside the selected area
Is a binary image having a value of, and weight information W (x, y, t) can be obtained by applying a low-pass filter to this image a plurality of times.

That is, the weight information W (x, y, t) takes a value of 1 inside the selected area, 0 outside the selected area, and 0 to 1 at the boundary of the selected area. The above is the description of the image creating method in the combining unit 805.

The coded data coded by the lower layer coding section 804, the upper layer coding section 803, and the area information coding section 806 are integrated by a coded data integration section (not shown) and transmitted or stored.

Next, on the decoding side of the conventional method, the coded data is decomposed by a coded data decomposing unit (not shown) into coded data of the lower layer, coded data of the upper layer, and coded data of the area information. . As shown in FIG. 8, these encoded data are decoded by the lower layer decoding unit 808, the upper layer decoding unit 807 and the area information decoding unit 809.

The decoding side synthesizing unit 810 is composed of the same device as the coding side synthesizing unit 805, uses the lower layer decoded image and the upper layer decoded image, and is the same as that described in the explanation of the coding side. The method synthesizes the images. The moving image combined here is displayed on the display,
It is input to the upper layer decoding unit 807 and used for prediction of the upper layer.

Here, the decoding apparatus for decoding both the lower layer and the upper layer has been described. However, if the decoding apparatus includes only the decoding section for the lower layer, the upper layer coding section 807 and the synthesizing section 810 are unnecessary. Therefore, a part of encoded data can be reproduced with a small hardware scale.

[0021]

(1) In the prior art, when an output image is obtained from two lower layer decoded images and one upper layer decoded image as in the above equation (1), Since the two lower layers are interpolated, when the position of the selected area changes with time, there is a problem that a large distortion occurs around the selected area and the image quality is greatly deteriorated.

FIG. 18 illustrates this problem.
In FIG. 18A, images A and C are two decoded images of the lower layer, image B is a decoded image of the upper layer, and the display time order is A, B, and C. However, the selected area is indicated by diagonal lines. Further, since only the selected area is encoded in the upper layer, the area outside the selected area is indicated by a broken line.

Since the selected area is moving, image A and image C
The interpolated image obtained from the above is such that two selected areas overlap, like the halftone dot portion in FIG. Furthermore, when the image B is combined using the weight information, the output image becomes an image in which three selected areas overlap, as shown in FIG.

In particular, around the selected area of the upper layer (outside)
The selected area of the lower layer appears as an afterimage in the image, and the image quality is greatly deteriorated. The entire moving image does not have the above distortion when only the lower layer is displayed, and the above distortion appears when the composite image of the upper layer and the lower layer is displayed, which causes flicker distortion. However, the image quality is extremely deteriorated.

(2) In the conventional technique, the eight-direction quantization code (FIG. 17) is used to encode the area information.
When applied to a low bit rate or when the shape of a region becomes complicated, the ratio of the data amount of the region information to the total encoded data amount becomes large, which causes a problem of image quality deterioration.

(3) In the prior art, the weight information is obtained by applying the low-pass filter to the area information a plurality of times. However, since the filter operation is performed a plurality of times, there is a problem that the processing amount increases. is there.

(4) In the conventional technique, the predictive coding is used, but the predictive coding may be used even when there is a scene change in the lower layer, which causes a large distortion. Since the distortion in the lower layer spreads to the upper layer, there is a problem that the distortion is sustained for a long time.

(5) In the prior art, since the moving image coding international standardization method such as MPEG or H.261 is used in the lower layer, there is not much difference in image quality between the selected area and other areas. . On the other hand, in the upper layer, only the selected area is coded with high image quality, so that the image quality in the selected area changes with time, which is detected as flicker distortion.

The object of the present invention is to solve these problems,
An object of the present invention is to provide an image encoding device and an image decoding device that reduce the amount of data after encoding and do not deteriorate the quality of a decoded image.

[0030]

An image coding apparatus according to the invention described in claim 1 of the present application hierarchically codes a specific component area in a moving image sequence to form a lower layer and an upper layer.
An image coding apparatus including moving image coding means for generating coded data of 1. and area information coding means for coding area information indicating the area shape of the specific component area, wherein the area information code Region information that approximates the region information
Approximation means, inter-frame difference data calculation means for obtaining an inter-frame difference of the approximated area information, and change information code for encoding change information indicating whether the approximated area shape has changed based on the difference data. Means and output data
Have a selection means for selecting the sign of the area information is not approximate
The encoded data of the region information approximated with the encoded data
It is characterized by outputting to.

The image decoding apparatus according to the second aspect of the present invention is a moving image for decoding encoded data of a lower layer and an upper layer in which a specific component area in a moving image sequence is hierarchically encoded. An image decoding device comprising image decoding means and area information decoding means for decoding coded data obtained by coding area information indicating the area shape of the specific component area, wherein the area information decoding means does not approximate. Area information encoding data
Data, the encoded data of the region information approximated to
A change means for inputting by replacement and a change information decoding means for decoding change information indicating presence / absence of a change between frames of the approximated area shape, the change information being changed between frames of the approximated area shape. the, it is approximate to indicate that there is no
The area information of the previous frame is used as the area information.

[0032]

BEST MODE FOR CARRYING OUT THE INVENTION First , a synthesis section 805 in FIG.
An example of the first image synthesizer that solves the problem that occurs in
Stomach you described. That is, the present invention relates to an image synthesizing apparatus that does not cause distortion such as an afterimage in the periphery of a selected region of an upper layer when synthesizing an image from two lower layer decoded images. Figure 1 is a block diagram showing the first images synthesizer.

In FIG. 1, the first area extraction unit 101
Is the first area information of the lower layer and the second area of the lower layer.
A region that is the first region and is not the second region is extracted from the region information of. In FIG. 10A, if the first area information is represented by a dotted line (the inside of the dotted line has a value of 0 and the outside of the dotted line has a value of 1), and similarly, the second area information is represented by a broken line, The area extracted by the first area extracting unit 101 is a shaded area in FIG.

The second area extracting unit 102 extracts the first area information of the lower layer and the second area information of the lower layer from the first area information of the lower layer.
A region that is the second region and not the first region is extracted. In the case of FIG. 10A, the halftone dot portion is extracted.

The controller 103 is a part for controlling the switch 104 by the outputs of the first area extracting unit 101 and the second area extracting unit 102. That is, when the target pixel position is only in the first area, the switch 104 is connected to the second decoded image side, and when the target pixel position is only in the second area, the switch 104 is connected to the first decoded image. Otherwise, the switch 104 is connected to the output from the interpolation image creating unit 105 in all other cases.

The interpolated image generation unit 105 calculates an interpolated image of the first decoded image of the lower layer and the second decoded image of the lower layer according to the equation (1) described as the above prior art. However, in equation (1), B (x, y, t1) is the first decoded image,
B (x, y, t2) is the second decoded image, I (x, y, t3) is the interpolated image, and t1, t2, t3 are the first decoded image, the second decoded image and the interpolated image, respectively. It's time.

Since the image is created as described above,
For example, in the case of FIG. 10A, since the second decoded image is used in the shaded area, background pixels outside the selected area appear, and the first decoded image is used in the halftone area, and thus outside the selected area. Background pixel appears, and in other portions, an interpolated image of the first decoded image and the second decoded image appears.

Since the weighted averaging unit 106 of FIG. 1 superimposes the decoded image of the upper layer on the image created in this way, the combined image has the selected region (as shown in FIG. 10B). There is no afterimage around the shaded area, and an image with less distortion can be obtained. The weighted average unit 106 in FIG. 1 synthesizes the interpolated image and the decoded image of the upper layer by weighted average. Since the synthesizing method has been described in the above-mentioned conventional technique, the description thereof is omitted here.

In the first image synthesizing apparatus described above, the interpolated image creating unit 105 shown in FIG. 1 is provided. Instead, however, the first decoded image B (x, y, t1) and the second decoded image B B (x, y,
Of t2), the pixel value of the decoded image that is temporally close to t3, which is the time of the upper layer, may be used.

In that case, using the frame number of each image, when t3-t1 <t1-t2, I (x, y, t3) = B (x, y, t1), and otherwise. Is I (x, y, t3) = B (x, y, t2).

However, t1, t2, and t3 are times of the first decoded image, the second decoded image, and the decoded image of the upper layer, respectively.

Next , an example of the second image synthesizing device will be described. This example, in the first images synthesizer, taking into account the motion information of the decoded image of the lower layer, and an image synthesizing apparatus for synthesizing a more accurate image. FIG. 2 is a block diagram showing an apparatus for estimating motion parameters and transforming two decoded images and two pieces of area information corresponding to them.

In FIG. 2, the motion parameter estimation unit 20
In 1, the motion information from the first decoded image to the second decoded image in the lower layer is estimated. For example, a motion vector in block units is calculated, or motion of the entire image (parallel movement, rotation, enlargement / reduction, etc.) is calculated and used as a motion parameter.

In the transformation unit 202, the first decoded image and the second decoded image
The decoded image, the first area information, and the second area information are transformed by the estimated motion parameters based on the temporal position of the synthesis target frame. For example, it is assumed that a motion vector (MVx, MVy) from the first decoded image to the second decoded image is obtained as a motion parameter. here,
MVx is the horizontal component of the motion vector, and MVy is the vertical component of the motion vector.

At this time, the motion vector from the first decoded image to the interpolated image is calculated by (t3-t1) / (t2-t1) (MVx, MVy), and the first decoded image is converted to this motion vector. Shift. When rotation, enlargement / reduction, or the like is used as a motion parameter, deformation is involved rather than simple shift.

In FIG. 2, the transformed data are represented by a, b, c and d, respectively, where a is the first in FIG.
Image, b is the second decoded image in FIG. 1, c is FIG.
1 is input as the first area information in FIG. 1 and the second area information in FIG. 1 in d is input to the image combining apparatus illustrated in FIG. 1 to create a combined image.

In the second image synthesizing device , the motion parameter is estimated from the two decoded images, but in the predictive coding, the motion vector of each block of the image is included in the coded data. Since these are general, these motion vectors may be used.

For example, the average value of the decoded motion vectors is set as the motion vector of the entire image from the first decoded image to the second decoded image, or the frequency distribution of the decoded motion vectors is obtained, A vector with high frequency may be used as a motion vector of the entire image from the first decoded image to the second decoded image. The above process is performed independently in the horizontal and vertical directions.

[0049] Next, a description will be given of the actual施例of the present invention. The present embodiment relates to an area information encoding device that efficiently encodes area information. 3 and 4 are block diagrams of this embodiment. FIG. 3 shows the encoding side and FIG. 4 shows the decoding side.

The area information approximation unit 301 in FIG. 3 approximates the area information with a plurality of figures. FIG. 11 shows an example of approximation. In this example, a rectangle is used as the figure, and the region information (hatched portion) of the person is approximated by two rectangles. A rectangle 1 represents a person's head, and a rectangle 2 represents a person's chest.

The area approximation information coding section 302 is a section for coding the above-mentioned approximated area information. As shown in FIG. 11, in the case of approximation with a rectangle, the upper left coordinate value of each rectangle and the size of the rectangle may be encoded with a fixed length.
Alternatively, when approximation is performed with an ellipse, the coordinates of the center point of the ellipse, the length of the long axis, and the length of the short axis may be encoded with a fixed length. The approximated area information and encoded data are
It is sent to the selection unit 304.

The area information coding unit 303, similar to the area information coding unit 806 described in the above-mentioned prior art, codes the area information using the 8-direction quantization code without approximating. The area information and the encoded data are sent to the selection unit 304.

The selecting unit 304 is a region approximation information coding unit 3
02 or the output of the area information coding unit 303 is selected. When the output of the region approximation information coding unit 302 is selected, the coded data of the region approximation information is sent together with 1-bit selection information (for example, 0) to a coded data integration unit (not shown) to display the region approximation information. No Send to the synthesizer.

When the output of the area information encoding unit 303 is selected, the encoded data of the area information that is not approximated is set to 1
Along with the bit selection information (for example, 1), it is sent to a coded data integration unit (not shown), and non-approximate area information is sent to the image synthesis unit. Image combining unit are those described in each example above the first image synthesizing device and the second image synthesizer.

As the selection method in the selection unit 304,
For example, a method of selecting a smaller encoded data amount, or an output of the region information encoding unit 303 when the encoded data amount of non-approximated region information is within a certain threshold, and an area approximate information encoding when the amount of encoded data exceeds the threshold The output of the unit 302 is selected. By making such a selection, it is possible to reduce the amount of encoded data while suppressing the encoding distortion of the area information.

Next, a description will be given decoding side real施例of the present invention (FIG. 4). In FIG. 4, the selection unit 401 selects whether the encoded data is the area approximation information or the area information, based on the 1-bit selection information included in the encoded data.

The area approximation information decoding unit 402 decodes the area approximation information. The area information decoding unit 403 decodes area information that is not approximate. The switch 404 is the selection unit 40.
The region approximation information or the region information that is not approximated is selected as the output to the synthesis unit, which is controlled by the signal from 1.

As described above, since the region approximation information and the original region information that is not approximated are adaptively selected and encoded / decoded, when the region information is complicated and has a huge amount of data, The coding of the region approximation information is selected, and the region information can be coded with a small amount of information.

In the above example, the area information that does not approximate is 8
Although the coding is performed by the direction quantization code, the prediction coding may be further combined for efficient coding. The eight-direction quantized code has values of 0 to 7 as shown in FIG. 17, but if the difference is obtained by predictive coding, it becomes -7 to 7.

However, the difference value can be suppressed to -3 to 4 by adding 8 when the difference value is -4 or less and subtracting 8 when the difference value is larger than 4. When decoding, add the difference value to the previous value, add 8 if the result is negative,
If it exceeds 7, the original 8-direction quantized value can be obtained by subtracting 8.

An example is shown below. 8-direction quantized value 1, 6, 2, 1, 3, ... Difference value 5, -4, -1, -2, ... Transformed value -3, 4, -1, 2, ... Decoding Value 1, 6, 2, 1, 3, ... For example, the difference between the value 6 and the previous value is 5, but from now on it is 8
Becomes −3 by subtracting, and −2 is obtained by adding the previous value 1 to the decoded value −3 at the time of decoding, but since the value is negative, 8 is added to this and a decoded value 6 is obtained. Such predictive coding utilizes the property that the 8-direction quantized code is cyclic.

In the present embodiment, the coding of the approximated area information is carried out independently for each image. However, in general, since moving images have a high correlation between frames, the previous coding result is used. The coding efficiency may be increased by using the above method.

That is, when the coding of the approximated area information is continuous between frames, only the difference of the approximated area information is coded. For example, the region is approximated by a rectangle, and the rectangle of the previous frame is the upper left point: (10, 20), size:
Represented by (100, 150), the rectangle of the current frame is the upper left point: (1
3, 18) and size: (100, 152), the upper left difference value in the current frame: (3, 2), size difference value: (0, 2)
Is encoded.

When the shape change of the area is small, all the difference values are concentrated in the vicinity of 0. Therefore, if entropy coding such as Huffman coding is used, the code amount of the area information can be greatly reduced. Furthermore, when the rectangle does not change in many cases, 1-bit information in the current frame may be encoded as the rectangle change information.

That is, when the rectangle does not change, only 1-bit information (for example, 0) representing this is encoded, and when the rectangle changes, 1-bit information (for example, 1) and the difference information are encoded. To do.

Next , an example of the fourth image synthesizing device will be described. The fourth image synthesizing apparatus relates to a weight information creating apparatus that creates multivalued weight information from area information. FIG. 5 is a block diagram of the fourth image synthesizing device .

Referring to FIG. 5, a horizontal weight creating section 501 is provided.
Scans the area information in the horizontal direction to obtain a portion where the area information is 1, and obtains a weighting function corresponding thereto. Specifically, the coordinate x0 of the leftmost point of the area and the horizontal length N of the area
And the horizontal weighting function as shown in FIG. 12A is calculated.

The weighting function may be created by combining straight lines, or may be created by combining straight lines and trigonometric functions. As an example of the latter, when the width of the trigonometric function part is W,
If N ＞ 2W, then when 0 ≦ x <W sin [(x + 1/2) π / (2W)] * sin [(x + 1/2) π / (2W)] W ≦ x <NW When 1 NW ≦ x <N sin [(x-N + 2W + 1/2) π / (2W)] * sin [(x-N + 2W + 1/2) π / (2W)] N ≦ For 2W, sin2 [(x + 1/2) π / N] * sin [(x + 1/2) π / N] can be used. However, the point x0 at the left end of the area is 0
I am trying.

The vertical weight creating unit 502 scans the area information in the vertical direction to obtain a portion where the area information is 1, and obtains a vertical weight function corresponding to the portion. Specifically, the coordinate y0 of the point at the upper end of the area and the vertical length M of the area are obtained, and the vertical weighting function as shown in FIG. 12B is calculated.

The multiplier 503 is a horizontal weighting unit 50.
1 and the output of the vertical weight generation unit 502 are multiplied for each pixel position to generate weight information. If the weight information is created in this way, the weight information that matches the shape of the area information can be obtained with a small amount of calculation.

Next , an example of the fifth image synthesizing device will be described. The fifth image synthesizing apparatus relates to a mode switching method for adaptively switching between intraframe coding and interframe predictive coding in predictive coding of a lower layer or an upper layer. FIG. 6 is a block diagram of the fifth image synthesizing device .

In FIG. 6, the average value calculation unit 601 receives the original image and the area information and calculates the average of the pixel values of the pixel values inside the area. The average value is input to the differentiator 603 and the storage unit 602. The differentiator 603 includes a previous average value stored in the storage unit 602 and an average value calculation unit 601.
Calculate the difference from the current average value output from.

Judging section 604 compares the absolute value of the difference value calculated by differentiator 603 with a predetermined threshold value and outputs the mode switching information. When the absolute value of the difference value is larger than the threshold value, it is determined that there is a scene change in the selected area, and mode switching information is generated so that intraframe coding is always performed.

As described above, by performing the mode switching while judging the scene change of the selected area, a good encoded image can be obtained even when a person appears from a shadow or the front and back of an object are reversed. Obtainable.

The fifth image synthesizing apparatus can be applied to a method of separately coding a selected area and other areas in the coding of the lower layer. In that case, the area information is input to the lower layer. Furthermore, the fifth image synthesizing apparatus can also be applied to the encoding of only the selected region of the upper layer.

Next , an example of the sixth image synthesizing device will be described. The sixth image synthesizing device relates to data amount control when a selected region and a region other than the selected region are separately encoded in encoding of a lower layer. FIG. 7 shows the sixth
It is a block diagram of an image synthesizing device .

In FIG. 7, the coding section 703 separates the selected area and the other area and codes them. Area information is input to the area determination unit 701, and it is determined whether the encoded area is inside or outside the selected area. The generated code amount calculation unit 705 calculates the generated code amount in each area based on this determination result.

The target code amount distribution ratio calculation unit 704 determines the distribution ratio of the target code amount to be assigned to each area in frame units. The method of determining the distribution ratio will be described later. The quantization width calculating unit 702 determines the quantization width according to the target code amount, and this determination method is the same as the conventional method.

Here, a method of determining the distribution ratio in the target code amount distribution ratio calculation unit 704 will be described. First, the target code amount Bi of the corresponding frame is calculated using the following equation. Bi
= (Usable code amount-use code amount up to the previous frame) / number of remaining frames This target code amount Bi is assigned to the inside and outside of the selected area at a certain ratio, but here, an appropriate fixed ratio R0 and The ratio is determined using the previous frame complexity ratio Rp.

The previous frame complexity ratio Rp is determined by the following equation. Rp = (gen # bitF * avg # qF) / (gen # bitF * avg # qF + gen # bitB * avg
#qB) where gen # bitF: Code amount generated in the previous frame selected area gen # bitB: Code amount generated outside the previous frame selected area avg # qF: Average quantization width in the selected area of the previous frame avg # qB: Previous frame This is the average quantization width outside the selected area.

In order to improve the quality of the selected area, the quantization width is controlled so that the average quantization width within the selected area is kept to some extent smaller than the average quantization width outside the selected area, and within the moving image sequence. It is desirable to follow changes in the image at.

Generally, the allocation by the fixed ratio R0 is suitable for keeping the relationship of the average quantization widths inside and outside the selected area substantially constant, and the allocation by the previous frame complexity ratio Rp is within the moving image sequence. Suitable for following changes in images.

[0083] Therefore, the goal bit allocation ratio Ra is a fixed ratio R0 and the average of the previous frame complexity ratio Rp, performs control that combines the advantages of both. That is, Ra = (R0 + Rp) / 2.

For example, it is assumed that the fixed ratio R0 and the previous frame complexity ratio Rp in the selected area are as shown by the dotted lines in FIG. 13 in the entire moving image sequence. At this time, the target code amount distribution ratio Ra becomes as shown by the solid line in FIG. 13, and it is understood that the target code amount distribution ratio Ra does not deviate from the fixed ratio R0 so much and the image change is reflected to some extent.

At this time, the fixed ratio outside the selected area is (1-R0),
If the previous frame complexity ratio is (1-Rp), the target code amount distribution ratio (1-Ra), which is the average of both, becomes as shown by the solid line in FIG. The value obtained by adding the target code amount allocation ratio is 1.

In this way, the quantization width can be controlled appropriately, but since the generated code amount may exceed the target code amount Bi depending on the frame, the bit rate of the entire sequence becomes a predetermined bit rate. It may happen that it does not fit. In such a case, the following method can be adopted.

The target code amount distribution ratio Ra in the selected area is the average of the fixed ratio R0 and the previous frame complexity ratio Rp, as described above. On the other hand, the target code amount allocation ratio outside the selected area is the fixed ratio outside the selected area (1-R0) and the previous frame complexity ratio (1-Rp).
And the minimum value of Rm.

In this way, the change in the target code amount distribution ratio Rm outside the selected area becomes as shown by the solid line in FIG. 15, for example. At this time, since Ra + Rm ≦ 1, the target code amount of the corresponding frame can be reduced. That is, by suppressing the target code amount of the background area, the bit rate of the entire moving image sequence can be kept within a predetermined bit rate.

[0089]

According to the image coding apparatus and the image decoding apparatus of the present invention, by adding the information indicating the presence or absence of the change of the area shape based on the inter-frame difference of the area information,
The region shape can be efficiently lossless-encoded / decoded.

[Brief description of drawings]

1 is a block diagram illustrating a first image synthesizing apparatus.

2 is a block diagram illustrating a second image synthesizing apparatus.

3 is a block diagram illustrating an encoding side of the real施例of the present invention.

4 is a block diagram illustrating a decoding side of the actual施例of the present invention.

5 is a block diagram illustrating a fourth image synthesizing apparatus.

6 is a block diagram illustrating the fifth image synthesizing apparatus.

7 is a block diagram illustrating an image composition apparatus according to a sixth.

FIG. 8 is a block diagram illustrating a conventional encoding device and decoding device.

FIG. 9 is a block diagram illustrating a conventional code amount control unit.

FIG. 10 is a diagram illustrating an effect of the first image synthesizing device .

11 is an explanatory diagram showing an example of approximating the region information in the real施例of the present invention.

FIG. 12 is an explanatory diagram showing an example of a method of creating weight information in the fourth image synthesizing device .

FIG. 13 is a diagram illustrating a target code amount ratio of a selected area by a code amount control method in the sixth image synthesizing device .

FIG. 14 is a diagram illustrating a target code amount ratio outside a selected area by the code amount control method in the sixth image synthesizing apparatus .

FIG. 15 is a diagram illustrating a target code amount ratio outside a selected area by another code amount control method in the example of the sixth image synthesizing device .

[Fig. 16] Fig. 16 is a diagram for describing the concept of conventional moving image hierarchical encoding.

FIG. 17 is a diagram illustrating an eight-direction quantized code.

[Fig. 18] Fig. 18 is a diagram for describing a problem of conventional moving image hierarchical encoding.

[Explanation of symbols]

101 First Area Extraction Unit 102 second region extraction unit 103 controller 104 switch 105 Interpolation image creation unit 106 Weighted average part 301 Area information approximation unit 302 area approximation information coding unit 303 area information coding unit 304 Selector 401 Selector 402 area approximation information decoding unit 403 Area information decoding unit

─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Toshio Nomura 22-22 Nagaike-cho, Abeno-ku, Osaka City, Osaka Prefecture Sharp Corporation (56) Reference JP-A-8-116542 (JP, A) Kakuno, Tsuda, Bun , Eito, Video Coding Using Transparency Information for Region Representation, Research Report of Information Processing Society of Japan Audio-Visual Complex Information Processing 95-AVM-9, Nihon, July 14, 1995, Vol. 95, No. 64, p. 1-8 Eito, Bun, Kakuno, Image Coding Using Hierarchical Representation and Multiple Templates, IEICE Technical Report, Japan, March 17, 1995, Vol. 94, No. 549, p. 99-106 (58) Fields surveyed (Int.Cl. ⁷ , DB name) H04N ⁷ /24-7/68 G06T 9/20 JISST file (JOIS)

Claims (2)

(57) [Claims] 1. A specific part area in a moving image sequence is defined as a floor.
Layer- encoded lower layer and upper layer encoded data
An image encoding device comprising moving image encoding means for generating a data and area information encoding means for encoding area information indicating an area shape of the specific component area, wherein the area information encoding means is , Area information approximating means for approximating area information,
Based on the difference data, an inter-frame difference data calculating unit that obtains an inter-frame difference of the approximated area information,
Change information coding means for coding change information indicating whether or not there is a change in the approximated area shape, and a selection for selecting output data.
Have a-option unit, encoded data and the near region information not approximate
An image coding apparatus characterized by selectively outputting coded data of similar area information . 2. A specific part area in a moving image sequence is a floor.
A video decoding means for decoding a layer to the lower layer and the upper layer obtained by encoding the encoded data, and the area information decoding means the area information indicating the area shape of the specific part region decoding coded encoded data In the image decoding device having the above, the area information decoding means is a code of the area information that is not approximated.
Select the encoded data of the area information that is approximated to the encoded data
And a change information decoding unit that decodes change information indicating whether or not there is a change between frames of the approximated region shape, and a region where the change information is approximated. To indicate that there is no change in shape between frames,
An image decoding device characterized by using area information of a previous frame as the approximated area information.