JP2526675B2 - Orthogonal transform coding method of moving picture and its decoding method - Google Patents

Description

The present invention relates to a moving image orthogonal transform coding and decoding system.

(Prior Art) Conventionally, an orthogonal transform coding method is known as a high-efficiency image coding method.

In this method, an image frame is divided into blocks of a predetermined size, each block is orthogonally transformed, and the transform coefficient is quantized and encoded.

 An example of the circuit configuration is shown in FIG.

The image frame signal input to the two-dimensional block orthogonal transformation circuit 33 is first divided into blocks of a predetermined size. The signals of each block are replaced by mutually uncorrelated transform coefficients indicating spatial frequency components by orthogonal transform in the horizontal and vertical directions of the image.

The transform coefficient values have a distribution centered on 0. Further, if they are quantized by the quantizing circuit 34, they can be efficiently coded by variable length coding in the coding circuit 35.

The two-dimensional block orthogonal transformation circuit 33 is realized by cascade-connecting orthogonal transformations in the horizontal and vertical directions of an image as shown in FIG. 5 for an image in which the sampling positions on the horizontal line are the same in each line. It is possible. That is, it is widely used as a separable two-dimensional block orthogonal transform.

(Problems to be Solved by the Invention) When the conventional method is applied to the coding of an interlaced moving image, only the intra-field correlation of the moving image signal is used and the correlation in the time direction remains unchanged. It is considered that the improvement of is not sufficient. Therefore, a method of extending the conventional two-dimensional block orthogonal transformation to three-dimensional block orthogonal transformation can be considered.

However, 3 for interlaced video signals
It is not separable in all dimensions. That is,
It is not possible to realize separate orthogonal transforms in the time direction, vertical direction, and horizontal direction in a cascade connection. This is because in the interlaced video signal, the sampling structure is different between the odd field and the even field as shown in FIG. Further, although it is possible to realize the three-dimensional block orthogonal transform in which the odd field and the even field can be separately separated, in this case, there is a problem that the inter-field correlation is not used.

An object of the present invention is to provide a means for realizing separable three-dimensional block orthogonal transform coding by utilizing inter-field correlation for an interlaced moving image signal.

(Means for Solving the Problems) In order to solve the above-mentioned problems, a first orthogonal transform coding method for moving images according to the present invention is an orthogonal transform coding for moving images, which performs orthogonal transform coding on an interlaced moving image signal. An interpolating means for receiving the interlaced moving image signal, and interpolating 0 as a fixed value to a pixel on a scanning line skipped by interlacing for each field of the moving image signal to generate an interpolated image signal. , A three-dimensional transformation coefficient indicating the spatio-temporal frequency component of the image by receiving the interpolated image signal and performing three-dimensional block orthogonal transformation in the horizontal direction, the vertical direction, and the time direction of the image on a plurality of continuous field signals Block orthogonal transformation means, truncation means for truncating the coefficient of the folding component generated due to the interpolation of the fixed value among the transformation coefficients, and truncation by the truncation means. An orthogonal transform coding method for moving images, comprising: means for coding the remaining transform coefficients without being cut.

In a moving picture orthogonal transform coding method for orthogonal transform coding of an interlaced moving picture signal, a transform base which receives the interlaced moving picture signal and is alternately different for each line of the image with respect to the time direction. Block orthogonal transform means for performing an orthogonal transform using the above, and further performing a two-dimensional block orthogonal transform in the vertical direction and the horizontal direction, or performing orthogonal transform using different transform bases alternately in the vertical direction for each field; With respect to the output of the block orthogonal transform means, the two-dimensional block orthogonal transform means for obtaining the transform coefficient indicating the spatio-temporal frequency component of the image by performing the two-dimensional block orthogonal transform in the time direction and the horizontal direction, and the Truncation means for truncating a coefficient whose value can be estimated to be 0 from the sampling structure of the race video signal,
An orthogonal transform coding method for a moving image, comprising: means for coding the transform coefficients remaining uncensored by the censoring means.

0 as a fixed value for the transform coefficient truncated by the encoding
And a means for decoding a non-interlaced moving image signal by performing a three-dimensional block inverse orthogonal transform on the conversion coefficient, or a conversion coefficient obtained by interpolating 0 in the vertical and temporal directions of the image. Dimensional block inverse orthogonal transform is performed, and then at least one of the means for decoding the interlaced video signal by performing the inverse orthogonal transform in the horizontal direction only on the necessary line is provided. The signal encoded by the method according to claim 1 or 2 is decoded.

(Operation) In the interlaced video image encoding method using the orthogonal transformation of the first invention, first, a fixed value is set to a pixel indicated by an X in FIG. 6, that is, a pixel on a scanning line skipped by interlacing. Interpolate 0.

Next, the continuous multiple field signals are divided into blocks of a predetermined size. As an example, if the continuous field number is L, the block size is N × M ×
It can be expressed as L (N pixel, M line, L field). Here, the horizontal direction of the block is x, the vertical direction is y,
Also, the time direction is t, and the block signal is I (x, y, t)
It is represented by. The image signal is subjected to three-dimensional block orthogonal transformation of x-axis, y-axis, and t-axis for each block, and N × M × L transform coefficients C representing the intra-block space-time frequency component
Get (u, v, f).

In this conversion formula, u, v, f are the horizontal, vertical, and temporal frequencies of the intra-block image signal, and a1, a2, a3 are horizontal,
The orthogonal transformation bases in the vertical and time directions are shown. The three-dimensional block orthogonal transformation represented by the above transformation equation can be separated for each direction as shown in FIG. 4, and may be performed in any order.

By the way, since the conversion coefficient is the result of orthogonal conversion of the interpolated fixed value 0 on the interlaced scanning line, it is necessary to correct the gain.

Here, if N, M, and L are all multiples of 2, the values of the conversion coefficients may be all doubled. Further, the interlaced moving image signal having the sampling structure as shown in FIG. 6 does not have the frequency components in the shaded portion in FIG. In FIG. 7, V _L is the interlaced line frequency and f _F is the field frequency.

However, since the conversion coefficient C (u, v, f) is a coefficient obtained by orthogonally converting after interpolating the fixed value 0 in the interlaced signal, the coefficient corresponding to the shaded portion in FIG. have. Since these coefficients increase the number of coefficients and lower the coding efficiency, they are cut off and the remaining transform coefficients are quantized and coded.

By the above method, a separable three-dimensional block orthogonal transform coding method using inter-field correlation can be realized without coding unnecessary transform coefficients.

In the interlaced moving picture coding method using the orthogonal transformation of the second invention, first, a plurality of continuous field signals are divided into blocks of a predetermined size.

That is, when the number of consecutive fields is L, the size of the block can be expressed as N × M × L (N pixels, M lines, L fields). However, the number of lines M is the number of scanning lines including the lines skipped by interlacing. Here, the horizontal direction of the block is x, the vertical direction is y, and the time direction is t, and the intra-block signal is represented by I (x, y, t). Next, orthogonal transformation is performed on each block in the horizontal direction x, the vertical direction y, and the time direction t of the image. Here, L is an even number, and a case where orthogonal transformation is performed in the time direction t will be described first.

As shown in FIG. 6, orthogonal transformation in the time direction indicated by t is performed to obtain L transform coefficients H (x, y, f) for the frequency f in the time direction for each pixel position (x, y). .

By the way, as can be seen from FIG. 6, since the sampling positions are different between the odd line and the even line, even if the same orthogonal transform base is used, it seems that different orthogonal transforms have been performed in actual calculation.

That is, the orthogonal transformation is performed by using the odd-numbered coefficient of t in the odd-numbered line and the even-numbered coefficient of t in the even-numbered line of the orthogonal transformation basis coefficient sequence a3 (f, t) of length L. If it shows with a formula (Odd line) (Even line) Becomes

Next, the transform coefficient H (x, y, f) is subjected to a two-dimensional block orthogonal transform in the horizontal direction x and the vertical direction y of the image, and N × M × indicating the intra-block space-time frequency component of the image.
Obtain L transform coefficients C (u, v, f).

a1 (u, x) and a2 (v, y) are orthogonal transformation bases for the horizontal and vertical directions of the image, respectively.

The two-dimensional block orthogonal transformation expressed by the above transformation equation is the fifth
It can be separated as shown. Further, the orthogonal transformation in each of the time direction, the vertical direction, and the horizontal direction may be performed in any order.

By the way, although the interlaced video signal does not have a frequency component shown by hatching in FIG. 7 due to its sampling structure, the coefficient C corresponding to the frequency component is calculated in the above calculation.
(U, v, f) appears. These coefficients are unnecessary and can be truncated, and only the remaining transform coefficients are quantized and encoded.

By the above method, a separable three-dimensional block orthogonal transform coding method using inter-field correlation can be realized.

In the orthogonal transform of this method, first, orthogonal transform is performed in the vertical direction y of the image using different transform bases for each field, and M transform coefficients G are set at each (x, t) position.
(X, v, t) is obtained, and then the transformation coefficient G (x, v, t) is transformed into the transformation coefficient C (u , v, f) can be obtained. In this case, in the orthogonal transformation in the vertical direction, in the orthogonal transformation basis coefficient sequence a2 (v, y) of length M, the odd-numbered coefficient of y is used in the odd field and the even-numbered coefficient of y is used in the even field. The conversion is done.

Also in this case, the two-dimensional block orthogonal transform can be separated, and the orthogonal transforms in the vertical direction, the time direction, and the horizontal direction may be performed in any order.

In the method of decoding an image from a three-dimensional block orthogonal transform coefficient of an interlaced video signal of the third invention, first,
The encoded transform coefficient is inversely encoded and inversely quantized to return to the transform coefficient.

Next, the value of 0 is interpolated in the coefficient portion which is cut off on the transform side to this transform coefficient.

A non-interlaced moving image can be obtained as a decoded image by subjecting the interpolated transform coefficient to three-dimensional block inverse orthogonal transform of the image in the time direction, the vertical direction, and the horizontal direction.

In addition, if the two-dimensional block inverse orthogonal transform of the image in the time direction and the vertical direction is performed on the transform coefficient obtained by interpolating 0, and then the inverse orthogonal transform is performed for each line in the horizontal direction, the decoded image is interlaced. You can get a moving image of the race. Here, the interlaced line is the same as the interlaced line of the interlaced moving image which is the original image.

By the above method, both the non-interlaced video signal and the interlaced video signal can be decoded from the same transform coefficient, and an unnecessary inverse orthogonal transform operation is unnecessary when decoding the interlaced video signal. It can be omitted.

(Examples) Examples of the present invention will be described below with reference to FIGS.

FIG. 1 is a block diagram of an embodiment for realizing an orthogonal transform coding system for interlaced video signals of the first invention.

The interlaced moving image signal, which is the original signal, is first input to the interpolation circuit 1 and 0 is interpolated as a fixed value to the pixels of the line not scanned by the interlaced scanning for each image field.

Next, the interpolated image signal is replaced by transform coefficients in the three-dimensional block orthogonal transform circuit 2.

In the three-dimensional block orthogonal transformation circuit 2, first, the input image signal is divided into blocks having a certain size in the space-time direction. As an example, in the present embodiment, the continuous L field signals are divided into blocks each having a size of N pixels × M lines × L fields.

The block-shaped image signal is orthogonally transformed in the horizontal direction, the vertical direction and the time direction of the image, replaced with a transform coefficient having the same size as the image block, and output. The orthogonal transformation can be performed separately for each direction as shown in FIG. 4, and may be performed in any order. In FIG. 4, the transform bases 26, 27 and 28 are orthogonal transform bases of lengths N, M and L, respectively.

Next, the truncation circuit 3 truncates the portion of each conversion coefficient corresponding to the high-frequency component indicated by the diagonal lines in FIG. The remaining transform coefficients are gain-corrected and quantized by the quantizing circuit 4, and further variable-length coded by the coding circuit 5.

FIG. 2 is a block diagram of an embodiment which realizes the orthogonal transform coding method of the interlaced video signal of the second invention.

The interlaced video signal input to the block orthogonal transformation circuit 6 is first divided into blocks of a certain size in the spatiotemporal direction. As an example, it is divided into continuous L field signals N pixels × M lines × L field size blocks. However, the number of lines M is the number of scanning lines including the lines skipped by interlacing.

Next, the blocked image signal is orthogonally transformed in either the vertical direction or the time direction of the image to be transformed into N × M × L transform coefficients.

Here, since the image signal has a sampling structure as shown in FIG. 6, the conversion base is changed between an odd field and an even field in the case of orthogonal transformation in the vertical direction, and an odd number in the case of orthogonal transformation in the time direction. Change the conversion basis between lines and even lines. The two types of conversion bases are given from the respective conversion bases A11 and B12, and are alternately supplied to the conversion circuit 6 for each field or each line.

Next, in the two-dimensional block orthogonal transform circuit 7, the transform coefficient is orthogonally transformed in the direction not selected by the transform circuit 6, that is, either the vertical direction or the time direction of the image and the horizontal direction. When the transform circuit 6 performs the orthogonal transform in the time direction, the two-dimensional block orthogonal transform can be realized in a form separated in the vertical direction and the horizontal direction as shown in FIG. 5, and whichever is performed first. Good. In FIG. 5, transform bases 31 and 32 are orthogonal transform bases having lengths N and M, respectively.

When the transformation circuit 6 performs orthogonal transformation in the vertical direction, the two-dimensional block orthogonal transformation circuit 7 can be realized by replacing the vertical direction with the time direction in FIG. 5, and in this case, the time direction orthogonal transformation circuit. Is supplied with a transform base of length L.

The transformation coefficient which is the result of the series of transformations is truncated by the truncation circuit 8 to the coefficient of the high frequency component corresponding to the shaded area in FIG. The remaining coefficients are quantized by the quantization circuit 9 and further variable-length coded by the coding circuit 10.

FIG. 3 is a block diagram of an embodiment for realizing the decoding of the non-italaced video signal and the interlaced video signal from the orthogonal transform coefficient of the interlaced video signal of the third invention.

Here, the input coding transform coefficient is obtained by the method of the first invention or the second invention.

The encoded transform coefficient is first inverse-encoded by the inverse encoding circuit 13 and further inversely quantized by the inverse quantization circuit 14 to be decoded into a transform coefficient. Next, the interpolating circuit 15 gives 0 as an interpolation value to the transform coefficient portion where the transform coefficient is truncated on the encoding side.

The interpolated transform coefficient is inversely transformed by the time direction inverse orthogonal transform circuit 16 and the vertical direction inverse orthogonal transform circuit 17 in the time direction and the vertical direction of the image, respectively. The transform circuit 16 is supplied with a transform base 20 and the transform circuit 17 is supplied with a transform base 21 of orthogonal transform bases of lengths L and M, respectively.

Further, this conversion order may be exchanged. The transform coefficient obtained by the inverse transform is inversely transformed in the horizontal direction of the image in the horizontal inverse orthogonal transform circuit 18 to be decoded into a non-interlaced moving image. Here, the orthogonal transform basis of length L used for the inverse transform is supplied from the transform basis 22.

Further, the interlaced moving image can be decoded by inversely transforming the transform coefficient every other line using the transform basis in the line-interlaced horizontal inverse orthogonal transform circuit 19.
However, the line interlacing is the same interlace scanning as the interlace of the original image.

As described above, according to the present invention, it is possible to realize a separable three-dimensional block orthogonal transform coding method utilizing inter-field correlation without coding unnecessary transform coefficients. Both non-interlaced video signals and interlaced video signals can be combined from the same transform coefficient, and unnecessary inverse orthogonal transform operation can be omitted when decoding interlaced video images.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment for realizing an orthogonal transform coding system for interlaced video signals of the first invention, and FIG. 2 is an orthogonal transform coding system for interlaced video signals of the second invention. FIG. 3 is a block diagram of an embodiment for realizing the invention, and FIG. 3 is a block diagram of an embodiment for realizing the decoding method of the orthogonal transform coefficient of the interlaced video signal of the invention. FIG. 4 is a three-dimensional view of the first invention. FIG. 5 is a block diagram of an embodiment for realizing the block orthogonal transform, FIG. 5 is a block diagram of an embodiment for realizing the two-dimensional block orthogonal transform, and FIG. 6 is a sampling structure in the vertical direction and the time direction of an interlaced video signal. FIG. 7 is a diagram for explaining the frequency distribution in the vertical direction and the time direction of the interlaced video signal, and FIG. 8 is a block diagram of a conventional orthogonal transform coding system. 1,15 …… Interpolation circuit, 2 …… 3D block orthogonal transformation circuit, 3,8 …… Cutoff circuit, 4,9,34 …… Quantization circuit, 5,1
0,35 …… Encoding circuit, 6 …… Block orthogonal transformation circuit, 7,
33 …… 2D block orthogonal transformation circuit, 11,12,20,21,22,2
6,27,28,31,32 …… Conversion basis, 13 …… Inverse encoding circuit, 14
…… Inverse quantization circuit, 16 …… Time direction inverse orthogonal transform circuit, 17
... vertical reverse orthogonal transform circuit, 18 ... horizontal reverse orthogonal transform circuit, 19 ... line interlaced horizontal reverse orthogonal transform circuit, 23,29 ... horizontal orthogonal transform circuit, 24,30 ... vertical orthogonal Conversion circuit, 25 ... Time direction orthogonal conversion circuit.

Claims (3)

(57) [Claims] 1. An orthogonal transform coding method for a moving image, which orthogonally transform-codes an interlaced moving image signal, receives the interlaced moving image signal, and interlaces each field of the moving image signal by interlacing. Interpolating means for interpolating 0 as a fixed value to pixels on the scanning line to generate an interpolated image signal, and receiving the interpolated image signal, the horizontal and vertical directions of the image and the time direction of the image with respect to continuous field signals. A three-dimensional block orthogonal transform means for obtaining a transform coefficient indicating a spatiotemporal frequency component of an image by performing a three-dimensional block orthogonal transform, and a coefficient of a folding component generated due to the interpolation of the fixed value among the transform coefficients. Orthogonal transformation of a moving image, which comprises a truncation means and a means for encoding the transform coefficients remaining uncensored by the truncation means. Goka system. 2. An orthogonal transform coding method for a moving image, which orthogonally transform codes an interlaced moving image signal, receives the interlaced moving image signal, and receives the image signal line by line in the image in the time direction. Orthogonal transformation using alternating transformation bases is performed, and further 2
Dimensional block orthogonal transformation, or block orthogonal transformation means for performing orthogonal transformation using different transformation bases for each field in the vertical direction, and the output of the block orthogonal transformation means, in the time direction and the horizontal direction. A two-dimensional block orthogonal transform means for obtaining a transform coefficient indicating a spatiotemporal frequency component of an image by performing a two-dimensional block orthogonal transform;
The present invention is characterized by further comprising: truncation means for truncating a coefficient of which a value can be estimated to be 0 from the sampling structure of the interlaced video signal among the transform coefficients, and means for coding the transform coefficient remaining uncensored by the truncation means. Orthogonal transform coding method for moving images. 3. An interpolating circuit for interpolating 0 as a fixed value into a transform coefficient cut off by the encoding, and means for decoding a non-interlaced moving image signal by performing a three-dimensional block inverse orthogonal transform on the transform coefficient. , Or by performing the inverse orthogonal transform in the horizontal direction only on the necessary lines after performing the two-dimensional block inverse orthogonal transform in the vertical direction and the time direction of the image by interpolating the transform coefficient of 0. 3. A method for decoding a signal encoded by the method according to claim 1, further comprising at least one decoding means of the means for decoding.