JP4879151B2 - Encoder device, decoder device, wavelet transform method, and wavelet inverse transform method - Google Patents

Description

  The present invention relates to an encoder device, a decoder device, a wavelet transform method, and a wavelet inverse transform method, and more particularly, to an encoder device, a decoder device, a wavelet transform method, and a wavelet inverse transform method that perform line-based 5 × 3 reversible wavelet transform.

  JPEG2000 standardized in January 2001 by JPEG (Joint Photographic Experts Group) is known. JPEG2000 replaces discrete cosine transform (DCT) used in conventional JPEG with wavelet transform (DWT) to make block distortion inconspicuous when the compression ratio is increased.

Regarding this wavelet transform technique, various techniques are known in order to increase the efficiency and speed of data processing (see, for example, Patent Documents 1 to 4).
FIG. 15 is a diagram illustrating a conventional wavelet transform.

  The wavelet transform unit 75 performs 5 × 3 reversible filter processing in the vertical direction (column direction), and then performs 5 × 3 reversible filter processing in the horizontal direction (row direction), so that the low-frequency component LL of the transform result is obtained. A horizontal high-frequency component LH, a vertical high-frequency component HL, and a diagonal high-frequency component HH are output. That is, two-dimensional wavelet transformation is performed by performing wavelet transformation processing on one direction and transforming the result from another direction again. Note that balloons in the figure indicate the image conversion direction.

FIG. 16 is a diagram illustrating a conventional wavelet transform.
A wavelet transform unit 70 shown in FIG. 16 performs block-based wavelet transform.

The wavelet transform unit 70 includes a buffer 71, a 5 × 3 vertical reversible filter 72, a buffer 73, and a 5 × 3 horizontal reversible filter 74 from the left side in FIG.
The buffer 71 stores the input original image data by dividing it into rectangular block areas.

The 5 × 3 vertical reversible filter 72 performs vertical filtering on the data stored in the buffer 71 to calculate a one-dimensional conversion coefficient.
Here, as shown in the following equations (1) and (2), the one-dimensional conversion coefficient Y (2n + 1) of the pixel data X (2n + 1) existing in the odd-numbered columns is the pixel data X for three pixels adjacent in the vertical direction. (2n), X (2n + 1), and X (2n + 2) are calculated. Further, the one-dimensional conversion coefficient Y (2n) of the pixel data X (2n) existing in the even columns is the pixel data X (2n-2), X (2n-1), X ( 2n), X (2n + 1), and X (2n + 2).

  here,

(Floor operation part) means the maximum integer value not exceeding A.
The buffer 73 stores the one-dimensional conversion coefficient calculated by the 5 × 3 vertical direction reversible filter 72.

The 5 × 3 horizontal reversible filter 74 performs a horizontal filtering process on the one-dimensional transform coefficient stored in the buffer 73 to calculate a two-dimensional transform coefficient.
By the way, a semiconductor integrated circuit having such a wavelet transform function is often used in a small electronic device such as a portable terminal device, and there is a demand for reducing the circuit scale. For this reason, a technique for reducing the size of the wavelet transform unit by reducing the capacity of the buffer is known.

FIG. 17 is a diagram illustrating a wavelet transform unit using a line buffer. In addition, the same code | symbol is attached | subjected to the part provided with the function similar to FIG. 16, and the description is abbreviate | omitted.
The wavelet transform unit 80 performs wavelet transform using a lifting structure.

  The wavelet transform unit 80 includes line buffers 71a to 71d that hold pixel data for one line (for one horizontal column) instead of the buffer 71 shown in FIG. Further, instead of the buffer 73, buffers 73a to 73d each holding pixel data for one pixel are provided.

FIG. 18 is a diagram showing filter processing of the wavelet transform unit shown in FIG.
When the pixel data X (2n + 1) of the odd-numbered columns (first column, third column,...) Is input to the wavelet transform unit 80, the 5 × 3 vertical reversible filter 72 performs the following processing (odd-numbered columns). Line processing).

  As shown in FIG. 18A, the 5 × 3 vertical reversible filter 72 includes four pieces of pixel data X (2n−3) and X (2n) respectively stored in the line buffers 71a to 71d in the previous processing. -2), X (2n-1), X (2n), 6 pixels arranged in the vertical direction composed of the pixel data X (2n + 1) and the pixel data X (2n + 2) inputted this time to the wavelet transform unit 80 Among them, the one-dimensional conversion coefficient Y (2n-1) is calculated according to the equation (1) using three pixels of pixel data X (2n-2), X (2n-1), and X (2n), and the buffers 73a to 73a- Store in any of 73d. The storage position will be described later.

  Further, the pixel data X (2n−2), X (2n−1), X (2n), and X (2n + 1) are converted into the pixel data X (2n−3) of the line buffers 71a to 71d for the subsequent processing. , X (2n-2), X (2n-1), and X (2n) are stored (overwritten).

  On the other hand, when the pixel data X (2n + 2) of the even-numbered columns (second column, fourth column,...) Is input to the wavelet transform unit 80, the 5 × 3 vertical reversible filter 72 performs the following processing ( Even line processing).

  As shown in FIG. 18B, the 5 × 3 vertical reversible filter 72 has pixel data X (2n−2), X (2n−1), and X (2n) stored in the line buffers 71a to 71d. , X (2n + 1) and pixel data X (2n + 2) are used to calculate a one-dimensional conversion coefficient Y (2n) according to equation (2) using 5 pixels arranged in the vertical direction, and one of the buffers 73a to 73d To store.

  Further, pixel data X (2n−1), X (2n), X (2n + 1), and X (2n + 2) are processed for odd-numbered line processing (when pixel data X (2n + 3) is input). The pixel data X (2n-2), X (2n-1), X (2n), and X (2n + 1) of the line buffers 71a to 71d are stored (overwritten).

By adopting a configuration using such line buffers 71a to 71d, the capacity of the buffer can be reduced.
Next, as an example, the processing flow of even column line processing is shown.

19 and 20 are diagrams showing a processing flow of even column line processing.
In the even-numbered line processing, the wavelet transform unit 80 calculates a one-dimensional transform coefficient Y (2n) from five pixels arranged in the vertical direction according to the equation (2), as shown in FIGS. The data is sequentially stored in the buffers 73a to 73d (vertical even processing). In FIGS. 19A to 19D, the hatched portions in the upper stage indicate pixel data read from the line buffers 71a to 71d, and the portions that are filled in with black indicate input pixel data. Further, each black portion in the lower row indicates a one-dimensional conversion coefficient, and a hatched portion indicates that the one-dimensional conversion coefficient is stored in the buffers 73a to 73d.

  Then, as shown in FIG. 20A, when the one-dimensional conversion coefficient Y (2n) is output in a state where the four one-dimensional conversion coefficients are stored in the buffers 73a to 73d, the horizontal alignment 5 A two-dimensional conversion coefficient is calculated from the pixel (horizontal even processing). Next, as shown in FIGS. 20A to 20C, a two-dimensional conversion coefficient is calculated from three pixels arranged in the horizontal direction (horizontal odd number processing). Thereafter, horizontal even processing and horizontal odd processing are alternately performed.

FIG. 21 is a diagram illustrating another conventional wavelet transform.
The wavelet transform unit 90 shown in FIG. 21 is configured so that the number of line buffers and buffers is one less than that of the wavelet transform unit 80 shown in FIG.

FIG. 22 is a diagram showing filter processing of the wavelet transform unit shown in FIG.
In the case of performing odd column line processing, the 5 × 3 vertical reversible filter 72 converts the one-dimensional conversion coefficient Y (2n−1) obtained at the time of even column line processing of the pixel data X (2n−2) to the line buffers 71a˜71. The data is read from any one of 71c and stored in any of buffers 73a to 73c.

  Then, when the pixel data X (2n), the pixel data X (2n + 1) and the one-dimensional conversion coefficient Y (2n-1) input to the wavelet transform unit 90 are input as the next (pixel data X (2n + 2)) (B) Store in the line buffers 71a to 71c for even line processing.

When performing even column processing, the 5 × 3 vertical reversible filter 72 uses the pixel data X (2n) and X (2n + 1) stored in the line buffers 71a to 71c during the odd column processing and the wavelet transform unit 90. The one-dimensional conversion coefficient Y (2n + 1) is calculated using the pixel data X (2n + 2) input to the. Then, the one-dimensional conversion coefficient Y (2n−1) stored in the line buffers 71a to 71c is read, and the one-dimensional conversion coefficient Y (2n) is calculated using Y (2n + 1) calculated this time and the pixel data X (2n). Is calculated and stored in any of the buffers 73a to 73c. Then, pixel data X (2n + 1), X (2n + 2) and a one-dimensional conversion coefficient Y (2n + 1) are converted into pixels for the next odd-numbered line processing (when pixel data X (2n + 3) is input). The data X (2n-1), X (2n) and the one-dimensional conversion coefficient Y (2n-1) are written back to the stored part. According to such a wavelet transform unit 90, wavelet transform can be performed using the three line buffers 71a to 71c.
JP 2002-91943 A JP 2003-142989 A JP 2003-196261 A Japanese Patent Laid-Open No. 2006-29595

As described above, the amount of memory used can be reduced by reducing the number of line buffers.
By the way, as shown in the equation (1), since three pixel data are used to obtain a one-dimensional transformation coefficient, it is difficult to perform a reversible wavelet transformation if the line buffer is further reduced. . For example, if the following equation (4) is approximated as in the following equation (5) and the following equation (6) is approximated as in the following equation (7), an irreversible conversion results in an error.

  For example, if X (2n + 1) = 10, X (2n) =-3, and X (2n + 2) =-1,

on the other hand,

That is, there is a problem that if the floor calculation part is simply expanded and calculated, an error occurs in the calculation result.
The present invention has been made in view of these points, and an object thereof is to provide an encoder device, a decoder device, a wavelet transform method, and a wavelet inverse transform method capable of performing highly accurate wavelet transform even with a small storage capacity. And

  In the present invention, in order to solve the above-described problem, in the encoder device 1 that performs 5 × 3 wavelet transform, the input pixel data X (2n + 1) (n is an integer) and the pixel data X (2n + 1) are input before. The first conversion coefficient defining value for defining the H component conversion coefficient Y (2n + 1) using the pixel data X (2n) adjacent to the pixel data X (2n + 1) stored in the first storage unit 2 The first conversion correction coefficient for correcting the calculation result of the H component conversion coefficient Y (2n + 1) is calculated, and the pixel data X (2n) and the process are already performed and stored in the second storage unit 3. A second correction coefficient defining value for defining the L component conversion coefficient Y (2n) using the H component conversion coefficient Y (2n-1) and a calculation result of the L component conversion coefficient Y (2n) are corrected. 2 conversion correction coefficients are calculated, and the first Filter processing for storing the conversion coefficient specified value and the first conversion correction coefficient in the first storage unit 2 and storing the second conversion coefficient specified value and the second conversion correction coefficient in the second storage unit 3 An encoder device 1 including the unit 4 is provided.

  According to such an encoder device 1, the filter processing unit 4 creates the first conversion coefficient specified value and the first conversion correction coefficient using the pixel data X (2n + 1) and the pixel data X (2n). Is done. Also, a second conversion coefficient specified value and a second conversion correction coefficient are created using the pixel data X (2n) and the H component conversion coefficient Y (2n-1). Then, the first conversion coefficient specified value and the first conversion correction coefficient are stored in the first storage unit 2, and the second conversion coefficient specified value and the second conversion correction coefficient are stored in the second storage unit 3. Stored in

  Further, in the present invention, in order to solve the above-described problem, in a decoder device that performs 5 × 3 wavelet inverse transform, an input L component transform coefficient Y (2n + 2) (n is an integer) and the L component transform coefficient Y ( In order to define the pixel data X (2n + 2) using the H component conversion coefficient Y (2n + 1) input before 2n + 2) and stored in the first storage unit and adjacent to the L component conversion coefficient Y (2n + 2). The first inverse transform coefficient specified value and the first inverse transform correction coefficient for correcting the calculation result of the pixel data X (2n + 2) are calculated, and the H component transform coefficient Y (2n + 1) is already processed. The second inverse transform coefficient defining value for defining the pixel data X (2n + 1) using the pixel data X (2n) stored in the second storage unit and the pixel data X (2n + 1) Compensate calculation results A second inverse transformation correction coefficient is calculated, the first inverse transformation coefficient prescribed value and the first inverse transformation correction coefficient are stored in the first storage unit, and the second inverse transformation coefficient is stored. There is provided a decoder device comprising a filter processing unit for storing a prescribed value and the second inverse transform correction coefficient in the second storage unit.

  According to such a decoder device, the filter processing unit uses the L component conversion coefficient Y (2n + 2) and the H component conversion coefficient Y (2n + 1) to specify the first inverse conversion coefficient specified value and the first inverse conversion correction. And coefficients are created. Further, the second inverse transformation coefficient prescribed value and the second inverse transformation correction coefficient are created using the H component transformation coefficient Y (2n + 1) and the pixel data X (2n). Then, the first inverse transform coefficient prescribed value and the first inverse transform correction coefficient are stored in the first storage unit, and the second inverse transform coefficient prescribed value and the second inverse transform correction coefficient are the second Stored in the storage unit.

  According to the present invention, the first conversion coefficient specified value and the first conversion correction coefficient are stored in the first storage unit, and the second conversion coefficient specified value and the second conversion correction coefficient are stored in the second storage unit. Since it is stored in the storage unit, 5 × 3 reversible wavelet transform can be performed by using values and coefficients stored in the subsequent line processing. As a result, highly accurate image conversion can be performed, and the circuit scale can be reduced.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
First, an outline of the present invention will be described, and then an embodiment will be described.
FIG. 1 is a diagram showing an outline of the present invention.

The encoder device 1 is a device that performs line-based 5 × 3 wavelet transform, and includes a first storage unit 2, a second storage unit 3, and a filter processing unit 4.
The first storage unit 2 and the second storage unit 3 store the pixel data of the original image input from the outside and the values and coefficients output from the filter processing unit 4, respectively.

The filter processing unit 4 uses the input pixel data X (2n + 1) (n is an integer) and the pixel data X (2n) adjacent to the pixel data X (2n + 1) to obtain the H component conversion coefficient Y (2n + 1). The first conversion coefficient defining value (corresponding to an operation value T _odd described later) for defining (deriving the H component conversion coefficient Y (2n + 1)) and the calculation result of the H component conversion coefficient Y (2n + 1) are corrected. 1 conversion correction coefficient (corresponding to a correction value F _odd described later) is calculated. Here, the pixel data X (2n) is input before the pixel data X (2n + 1) and stored in the first storage unit 2.

Further, the filter processing unit 4 includes the pixel data X (2n), the H component conversion coefficient Y (2n−1) already processed by the filter processing unit 4 and stored in the second storage unit 3. Is used to correct a calculation result of a second conversion coefficient specified value (corresponding to a calculation value T _even described later) and an L component conversion coefficient Y (2n) for specifying the L component conversion coefficient Y (2n) using Conversion correction coefficient (corresponding to a correction value F _even described later).

  Then, the filter processing unit 4 stores the first conversion coefficient specified value and the first conversion correction coefficient in the first storage unit 2, and the second conversion coefficient specified value and the second conversion correction coefficient. Store in the second storage unit 3.

  According to such an encoder device 1, the filter processing unit 4 creates the first conversion coefficient specified value and the first conversion correction coefficient using the pixel data X (2n + 1) and the pixel data X (2n). Is done. Also, a second conversion coefficient specified value and a second conversion correction coefficient are created using the pixel data X (2n) and the H component conversion coefficient Y (2n-1). Then, the first conversion coefficient specified value and the first conversion correction coefficient are stored in the first storage unit 2, and the second conversion coefficient specified value and the second conversion correction coefficient are stored in the second storage unit 3. Stored in

Embodiments of the present invention will be described below.
FIG. 2 is a block diagram illustrating the encoder device according to the embodiment.
A JPEG 2000 encoder 100 shown in FIG. 2 includes a wavelet transform unit 10, an EBCOT unit 20, and a stream generation unit 30.

The wavelet transform unit 10 performs 5 × 3 reversible wavelet transform on the image data of the original image input by the raster scan method.
The EBCOT unit 20 compresses the output of the wavelet transform.

The stream generation unit 30 generates and outputs a stream.
FIG. 3 is a block diagram illustrating a configuration of the wavelet transform unit according to the embodiment.
The wavelet transform unit 10 includes a line buffer 11a, a line buffer 11b, a 5 × 3 vertical reversible filter 12, a buffer 13a, a buffer 13b, and a 5 × 3 horizontal reversible filter 14.

Each of the line buffers 11a and 11b has a capacity for storing pixel data for one line of the original image.
The 5 × 3 vertical reversible filter 12 performs a vertical wavelet transform using the image data stored in the line buffers 11a and 11b and the input image data to calculate a one-dimensional transform coefficient.

  Each of the buffers 13a and 13b has a capacity for storing pixel data for one pixel. The buffers 13a and 13b store data used in the 5 × 3 horizontal reversible filter 14.

The 5 × 3 horizontal reversible filter 14 has the same function as the 5 × 3 vertical reversible filter 12.
The 5 × 3 horizontal reversible filter 14 performs horizontal wavelet transform using the data stored in the buffers 13 a and 13 b and the one-dimensional transform coefficient output from the 5 × 3 vertical reversible filter 12. To calculate a two-dimensional conversion coefficient. That is, the wavelet transform unit 10 performs wavelet transform processing on a plurality of pixel data arranged in the vertical direction of the input original pixels, arranges the obtained one-dimensional transform coefficients in the horizontal direction, and transforms them again. The two-dimensional wavelet transform is performed.

  Hereinafter, the input of each filter is indicated by X (pixel position) and the output is indicated by Y (pixel position). That is, the 5 × 3 vertical reversible filter 12 receives the pixel data X and outputs a one-dimensional conversion coefficient Y. When the output one-dimensional transformation coefficient Y is processed by the 5 × 3 horizontal reversible filter 14, the one-dimensional transformation coefficient X is replaced based on the horizontal coordinate as the one-dimensional transformation coefficient X, and the same processing as the 5 × 3 vertical reversible filter 12 is performed. And a two-dimensional conversion coefficient Y is output.

  Here, the line buffers 11a and 11b and the 5 × 3 vertical reversible filter 12 constitute one processing unit corresponding to the first storage unit, the second storage unit, and the filter processing unit, respectively, and the buffers 13a and 13b. And the 5 × 3 horizontal reversible filter 14 constitute one processing unit corresponding to the first storage unit, the second storage unit, and the filter processing unit, respectively.

Next, the wavelet transform process of the wavelet transform unit 10 will be described.
FIG. 4 is a diagram conceptually showing the processing of the wavelet transform unit. Here, FIG. 4A is a diagram illustrating odd column line processing, and FIG. 4B is a diagram illustrating even column line processing. In addition, among the arrows shown in FIG. 4, a dotted arrow indicates a part processed before or last time or a part processed after the next time, and a solid line arrow indicates a part processed this time. ing.

When performing odd column line processing, the 5 × 3 vertical reversible filter 12 receives the odd column pixel data X (2n + 1) input to the wavelet transform unit 10 this time and the even column stored in the line buffer 11a. Using the pixel data X (2n) (using two pieces of pixel data arranged in the vertical direction), a calculation value (first conversion) for defining a one-dimensional conversion coefficient (H component conversion coefficient) Y (2n + 1) A coefficient specified value) T _odd and a correction value (first conversion correction coefficient) F _odd are calculated.

Here, the calculated value T _odd and the correction value F _odd are expressed by the following equations (10) and (11), respectively.

Here, the correction value F _odd indicates the lower 1 bit of the pixel data X (2n) (the remainder after division by 2).
The calculated value T _odd and the correction value F _odd are both stored (overwritten) in the line buffer 11a where the pixel data X (2n) was stored.

Here, the number of bits of the operation value T _odd is equal to the number of bits of the pixel data X (2n), and the number of bits of the correction value F _odd is 1 bit. Therefore, the value stored in the line buffer 11a is large. Is the number of bits of pixel data + 1 bit.

For this reason, the size of one storage area of the line buffer 11a is one bit larger than the conventional one.
Simultaneously with the calculation of the calculated value T _odd and the correction value F _odd , the one-dimensional conversion coefficient Y (2n−1) and the pixel data X (2n) stored in the line buffer 11b by the previous process are used. The calculated value (second conversion coefficient specified value) T _even and the correction value (second conversion correction coefficient) F _even are calculated.

Here, the calculated value T _even and the correction value F _even are expressed by the following equations (12) and (13), respectively.

Here, the correction value F _even indicates the lower 2 bits of the one-dimensional conversion coefficient Y (2n−1) (the remainder after division by 4).
The one-dimensional conversion coefficient Y (2n−1) is processed by the 5 × 3 horizontal filter 14 as the one-dimensional conversion coefficient X (2n) or the one-dimensional conversion coefficient X (2n + 1).

Here, since the number of bits of the operation value T _even is equal to the number of bits of the one-dimensional conversion coefficient Y (2n-1) and the number of bits of the correction value F _even is 2 bits, the value written to the line buffer 11b. Is the number of bits of the one-dimensional conversion coefficient + 2 bits.

For this reason, the size of one storage area of the line buffer 11b is increased by 2 bits compared to the conventional case.
On the other hand, when the even-column line processing is performed, the 5 × 3 vertical reversible filter 12 calculates the even-numbered pixel data X (2n + 2) input to the wavelet transform unit 10 this time and the calculation stored in the line buffer 11a. A one-dimensional conversion coefficient Y (2n + 1) is calculated using the value T _odd and the correction value F _odd .

  Here, the one-dimensional conversion coefficient Y (2n + 1) is expressed by the following equation (14).

Further, the one-dimensional conversion coefficient Y (2n) is calculated using the one-dimensional conversion coefficient Y (2n + 1), the calculated value T _even and the correction value F _even acquired from the line buffer 11b.
Here, the one-dimensional conversion coefficient Y (2n) is expressed by the following equation (15).

The calculated one-dimensional conversion coefficient Y (2n) is processed by the 5 × 3 horizontal filter 14 as the one-dimensional conversion coefficient X (2n) or the one-dimensional conversion coefficient X (2n + 1).
Thereafter, the pixel data X (2n + 2) is stored (overwritten) in the line buffer 11a where the calculated value T _odd and the correction value F _odd are stored, and the one-dimensional conversion coefficient Y (2n + 1) is stored in the line buffer 11b. The calculated value T _even and the correction value F _even are stored (overwritten) at the location where they were stored.

Then, the pixel data X (2n + 2) and the one-dimensional conversion coefficient Y (2n + 1) are read and processed during odd column line processing when the pixel data X (2n + 3) is input.
In this way, the 5 × 3 vertical reversible filter 12 outputs one-dimensional conversion coefficients Y (2n−1), Y (2n), Y (2n + 1),. This processing flow is the same as the processing flow described with reference to FIGS. That is, for example, when a line at a vertical odd-numbered position is input to the wavelet transform unit 10, the 5 × 3 vertical reversible filter 12 always performs odd-numbered line processing. The value output from the 5 × 3 vertical reversible filter 12 is input to the 5 × 3 horizontal reversible filter 14, which alternately repeats even line processing and odd line processing within the 5 × 3 horizontal reversible filter 14. .

  Next, the internal configuration of the 5 × 3 vertical reversible filter 12 and the 5 × 3 horizontal reversible filter 14 will be described. The 5 × 3 vertical direction reversible filter 12 and the 5 × 3 horizontal direction reversible filter 14 have the same configuration except for the data to be processed. The internal structure of will be described.

FIG. 5 is a block diagram showing a configuration of a 5 × 3 vertical reversible filter.
The 5 × 3 vertical reversible filter 12 includes an odd column line processing unit 12a that performs odd column line processing and an even column line processing unit 12b that performs even column line processing.

FIG. 6 is a block diagram illustrating a configuration of an odd-numbered line processing unit of the encoder.
The odd-numbered line processing unit 12a includes a calculation unit 121a, a calculation unit 122a, a subtraction unit 123a, a calculation unit 124a, a calculation unit 125a, and an addition unit 126a. Here, as an example, processing when pixel data X (2n + 1) is input is shown.

  The calculation unit 121a divides the pixel data X (2n) read from the line buffer 11a by 2, performs a floor calculation on the division result (X (2n) / 2), and outputs the result to the subtraction unit 123a.

The arithmetic unit 122a divides the pixel data X (2n) read from the line buffer 11a by 2, and outputs the remainder X (2n) mod2 to the line buffer 11a.
The subtraction unit 123a subtracts the floor calculation result of the calculation unit 121a from the input pixel data X (2n + 1), and outputs the subtraction result to the line buffer 11a.

  The calculation unit 124a reads the one-dimensional conversion coefficient Y (2n-1) from the line buffer 11b, divides it by 4, performs a floor calculation on the division result Y (2n-1) / 4, and outputs the result to the addition unit 126a.

The odd-numbered line processing unit 12a outputs the one-dimensional conversion coefficient Y (2n−1) read from the line buffer 11b.
The arithmetic unit 125a divides the one-dimensional conversion coefficient Y (2n-1) read from the line buffer 11b by 4, and outputs the remainder Y (2n-1) mod 4 to the line buffer 11b.

  The addition unit 126a adds the floor calculation result output from the calculation unit 124a and the pixel data X (2n) read from the line buffer 11a, and outputs the addition result to the line buffer 11b.

Here, the value output from the subtractor 123a is the calculated value T _odd , the value output from the calculator 122a is the correction value F _odd , the value output from the adder 126a is the calculated value T _even , and the calculation unit 125a outputs The value to be corrected is the correction value F _even .

Next, the configuration of the even-numbered line processing unit 12b will be described.
FIG. 7 is a block diagram illustrating a configuration of an even-numbered line processing unit of the encoder.
The even-numbered line processing unit 12b includes a calculation unit 121b, an addition unit 122b, a calculation unit 123b, a calculation unit 124b, an addition unit 125b, a subtraction unit 126b, a calculation unit 127b, an addition unit 128b, and an addition unit. 129b, a calculation unit 130b, a calculation unit 131b, an addition unit 132b, and an addition unit 133b. Here, as an example, processing when pixel data X (2n + 2) is input is shown.

The calculation unit 121b divides the input pixel data X (2n + 2) by 2, and outputs the remainder X (2n + 2) mod 2 to the addition unit 122b.
The adder 122b adds the remainder X (2n + 2) mod2 output from the calculator 121b and the correction value F _odd read from the line buffer 11a, and adds the result (F _odd + X (2n + 2) mod 2) to the calculator 123b. Output to.

The calculation unit 123b divides the addition result (F _odd + X (2n + 2) mod 2) output from the addition unit 122b by 2, and performs a floor operation on the division result (F _odd + X (2n + 2) mod 2) / 2 to add the addition unit 125b. Output to.

The calculation unit 124b divides the input pixel data X (2n + 2) by 2, performs a floor calculation on the division result (X (2n + 2) / 2), and outputs the result to the addition unit 125b.
The addition unit 125b adds the floor calculation result output from the calculation unit 123b and the floor calculation result output from the calculation unit 124b, and outputs the addition result to the subtraction unit 126b.

The subtractor 126b subtracts the addition result output by the adder 125b from the calculated value _Todd read from the line buffer 11a, and outputs the subtraction result to the calculator 127b and the calculator 131b. Further, after the calculation of the adder 133b is completed, the subtraction result of the subtractor 126b is stored (overwritten) in the line buffer 11b.

The calculation unit 127b divides the subtraction result output from the subtraction unit 126b by 4, and outputs the remainder to the addition unit 128b.
The adder 128b adds the correction value F _even read from the line buffer 11b and the remainder output from the calculator 127b, and outputs the addition result to the adder 129b.

The adding unit 129b adds 2 to the addition result output from the adding unit 128b, and outputs the addition result to the arithmetic unit 130b.
The calculation unit 130b divides the addition result output from the addition unit 129b by 4, performs a floor calculation on the division result, and outputs the result to the addition unit 132b.

The calculation unit 131b divides the subtraction result output from the subtraction unit 126b by 4, performs a floor operation on the division result, and outputs the result to the addition unit 132b.
The addition unit 132b adds the floor calculation result output from the calculation unit 130b and the floor calculation result output from the calculation unit 131b, and outputs the addition result to the addition unit 133b.

The adder 133b adds the operation value T _even read from the line buffer 11b and the addition result output from the adder 132b.
Here, the subtraction result output from the subtraction unit 126b is a one-dimensional conversion coefficient Y (2n + 1), and the addition result output from the addition unit 133b is a one-dimensional conversion coefficient Y (2n).

The even column line processing unit 12b stores (overwrites) the input pixel data X (2n + 2) in the line buffer 11a.
According to such a JPEG2000 encoder 100, 5 × 3 wavelet transform can be performed with no error _even by using only the two line buffers 11a and 11b by performing calculations using the correction values F _odd and F _even . The data storage unit of one line buffer has some exceptions, but one line buffer has the number of bits of pixel data + 1 bit, and the other line buffer has the number of bits of pixel data + 2 bits. This is the case of the 5 × 3 reversible filter 12 in the vertical direction. In the 5 × 3 reversible filter 14 in the horizontal direction, the number of bits of the primary transform coefficient + 1 or the number of bits of the primary transform coefficient + 2, that is, the input of the filter. The number of bits is +1, +2. The same applies to the decoder, and the number of input bits of the filter is the reference. Thereby, the capacity of the line buffer can be reduced as compared with the conventional case.

  As already described, for the horizontal wavelet transform, the wavelet transform may be applied to the one-dimensional transform coefficient group generated by the vertical wavelet transform in the same manner as described above. Note that when data outside the screen is required, the data inside the adjacent screen is folded and used as in the case of the vertical wavelet transform.

  In the present embodiment, in order to make the explanation easy to understand, the 5 × 3 vertical reversible filter 12 is provided with each functional block included in the odd column line processing unit 12a and the even column line processing unit 12b separately. Although described, the present invention is not limited to this, and some of the functional blocks provided in the odd-numbered line processing unit 12a and the even-numbered line processing unit 12b may be shared.

<Decoding process>
Next, the decoding process will be described.
FIG. 8 is a block diagram illustrating a decoder device according to the embodiment.

The JPEG 2000 decoder 200 restores the bit stream output from the JPEG 2000 encoder 100 to an original image (restored image).
The JPEG2000 decoder 200 is symmetrical to the JPEG2000 encoder 100, and includes a stream analysis unit 40, an inverse EBCOT unit 50, and a wavelet inverse transformation unit 60.

FIG. 9 is a block diagram showing the configuration of the wavelet inverse transform unit.
The wavelet inverse transform unit 60 includes line buffers 61a and 61b for storing data for one column, a 5 × 3 vertical reversible filter 62, buffers 63a and 63b, and a 5 × 3 horizontal reversible filter 64. ing. The functions of these line buffers 61a and 61b correspond to the line buffers 11a and 11b, the functions of the 5 × 3 vertical reversible filter 62 correspond to the 5 × 3 vertical reversible filter 12, and the functions of the buffers 63a and 63b Corresponding to the buffers 13a and 13b, the function of the 5 × 3 horizontal reversible filter 64 corresponds to the 5 × 3 horizontal reversible filter 14.

  Here, the line buffers 61a and 61b and the 5 × 3 vertical reversible filter 62 constitute one processing unit corresponding to the first storage unit, the second storage unit, and the filter processing unit, respectively, and the buffers 63a and 63b. And the 5 × 3 horizontal reversible filter 64 constitute one processing unit corresponding to the first storage unit, the second storage unit, and the filter processing unit, respectively.

  The wavelet inverse transform unit 60 performs an operation reverse to that of the wavelet transform unit 10. That is, first, the 5 × 3 horizontal reversible filter 64 performs horizontal wavelet inverse transformation, and then the 5 × 3 vertical reversible filter 62 performs vertical wavelet inverse transformation.

  Hereinafter, the input of each filter is indicated by Y (pixel position) and the output is indicated by X (pixel position). That is, the 5 × 3 horizontal reversible filter 64 receives the two-dimensional conversion coefficient Y and outputs the one-dimensional conversion coefficient X. When the output one-dimensional conversion coefficient X is processed by the 5 × 3 vertical reversible filter 62, the one-dimensional conversion coefficient X is replaced with a one-dimensional conversion coefficient Y based on the vertical coordinate, and pixel data X is output.

FIG. 10 is a diagram conceptually illustrating the wavelet inverse transform process.
As shown in the following equations (16) and (17), the pixel data X (2n) is obtained from three one-dimensional conversion coefficients Y (2n−1), Y (2n), and Y (2n + 1) adjacent in the vertical direction. calculate. The pixel data X (2n + 1) is calculated from five one-dimensional conversion coefficients Y (2n−1), Y (2n), Y (2n + 1), Y (2n + 2), and Y (2n + 3) adjacent in the vertical direction. .

Next, the wavelet inverse transformation process of the wavelet inverse transformation unit 60 will be described.
FIG. 11 is a diagram conceptually showing processing of the wavelet inverse transform unit. Here, FIG. 11A is a diagram illustrating even-numbered column line processing, and FIG. 11B is a diagram illustrating odd-numbered column line processing. In addition, among the arrows shown in FIG. 11, a dotted arrow indicates a portion processed before or last time or a portion processed after the next time, and a solid line arrow indicates a portion processed this time. ing.

When performing even-column line processing, the 5 × 3 vertical reversible filter 62 replaces the one-dimensional transform coefficient output from the current 5 × 3 horizontal reversible filter 64 with the even-numbered one-dimensional transform coefficient Y ( 2n + 2) and the one-dimensional transformation coefficient Y (2n + 1) (two one-dimensional transformation coefficients arranged in the vertical direction) stored in the line buffer 61a, and the pixel data (L component inverse transformation coefficient) X (2n + 2). A calculation value (first inverse transformation coefficient regulation value) T _EVEN and a correction value (first inverse transformation correction coefficient) F _EVEN for defining are calculated.

Here, the calculated value T _EVEN and the correction value F _EVEN are expressed by the following equations (18) and (19), respectively.

Here, the correction value F _EVEN indicates the lower 2 bits of the one-dimensional conversion coefficient Y (2n + 1) (the remainder after division by 4).
The calculated value T _EVEN and the correction value F _EVEN are both stored (overwritten) in the line buffer 61a where the one-dimensional conversion coefficient Y (2n + 1) was stored.

Here, since the number of bits of the operation value T _EVEN is equal to the number of bits of the one-dimensional conversion coefficient and the number of bits of the correction value F _EVEN is 2 bits, the size of the value written in the line buffer 61a is the primary value. The number of bits of the original transform coefficient + 2 bits.

For this reason, the size of one storage area of the line buffer 61a is increased by 2 bits compared to the conventional case.
Simultaneously with the calculation of the calculation value T _EVEN and the correction value F _EVEN , the calculation is performed using the pixel data X (2n) stored in the line buffer 61b by the previous process and the one-dimensional conversion coefficient Y (2n + 1). A value (second inverse transform coefficient prescribed value) T _ODD and a correction value (second inverse transform correction coefficient) F _ODD are calculated.

Here, the calculated value T _ODD and the correction value F _ODD are expressed by the following equations (20) and (21), respectively.

Here, the correction value _FODD indicates the lower 1 bit (the remainder after division by 2) of the pixel data X (2n).
Then, the pixel data X (2n) is output, and the calculated value T _ODD and the correction value F _ODD are both stored (overwritten) in the same line buffer 61b where the pixel data X (2n) was stored. .

Here, since the number of bits of the operation value T _ODD is equal to the number of bits of the one-dimensional conversion coefficient and the number of bits of the correction value F _ODD is 1 bit, the size of the value written to the line buffer 61b is the primary value. The number of bits of the original transform coefficient + 1 bit.

For this reason, the size of one storage area of the line buffer 11b is one bit larger than the conventional one.
On the other hand, when performing odd-column line processing, the 5 × 3 vertical reversible filter 62 outputs an odd-numbered one-dimensional conversion coefficient Y (2n + 3) in the vertical direction output from the 5 × 3 horizontal reversible filter 64 and a line buffer. Pixel data X (2n + 2) is calculated using the calculated value T _EVEN and the correction value F _EVEN stored in 61a.

  Here, the pixel data X (2n + 2) is expressed by the following equation (22).

Also, the pixel data X (2n + 1) is calculated using the pixel data X (2n + 2), the calculated value T _ODD and the correction value F _ODD acquired from the line buffer 61b.
Here, the pixel data X (2n + 1) is expressed by the following equation (23).

Then, the calculated pixel data X (2n + 1) is output.
Further, the one-dimensional conversion coefficient Y (2n + 3) is stored (overwritten) at the place where the calculated value T _EVEN and the correction value F _EVEN of the line buffer 61a are stored, and the pixel data X (2n + 2) is stored in the line buffer 61b. The calculated value T _ODD and the correction value F _ODD are stored (overwritten) at the location where the calculated value T _ODD and the correction value F _ODD were stored.

  Then, the one-dimensional conversion coefficient Y (2n + 3) and the pixel data X (2n + 2) are read and processed when the even-numbered line processing is executed when the one-dimensional conversion coefficient Y (2n + 4) is input.

  In this manner, the 5 × 3 vertical direction reversible filter 62 receives the pixel data X (2n−1), X (2n), X (2n + 1),. Output.

  Next, the internal configuration of the 5 × 3 vertical reversible filter 62 and the 5 × 3 horizontal reversible filter 64 will be described. The 5 × 3 vertical direction reversible filter 62 and the 5 × 3 horizontal direction reversible filter 64 have the same configuration except for the data to be processed. The internal structure of will be described.

FIG. 12 is a block diagram showing the configuration of the 5 × 3 vertical reversible filter of the wavelet inverse transform unit.
The 5 × 3 vertical reversible filter 62 includes an even-numbered line processing unit 62a that performs even-numbered line processing and an odd-numbered line processing unit 62b that performs odd-numbered line processing.

FIG. 13 is a block diagram showing the configuration of the even column line processing unit of the decoder. Here, as an example, processing when a one-dimensional conversion coefficient Y (2n + 2) is input is shown.
The even-numbered line processing unit 62a includes a calculation unit 621a, a calculation unit 622a, a subtraction unit 623a, a calculation unit 624a, a calculation unit 625a, and an addition unit 626a.

  The calculation unit 621a divides the one-dimensional conversion coefficient Y (2n + 1) read from the line buffer 61a by 4, performs a floor calculation on the division result (Y (2n + 1) / 4), and outputs the result to the subtraction unit 623a.

The calculation unit 622a divides the one-dimensional conversion coefficient Y (2n + 1) read from the line buffer 61a by 4, and outputs the remainder Y (2n + 1) mod4 to the line buffer 61a.
The subtraction unit 623a subtracts the floor calculation result output from the calculation unit 621a from the one-dimensional conversion coefficient Y (2n + 2), and outputs the subtraction result to the line buffer 61a.

  The calculation unit 624a divides the pixel data X (2n) read from the line buffer 61b by 2, performs a floor calculation on the division result (X (2n) / 2), and outputs the result to the addition unit 626a.

The even column line processing unit 62a outputs the pixel data X (2n) read from the line buffer 61b.
The arithmetic unit 625a divides the pixel data X (2n) read from the line buffer 61b by 2, and outputs the remainder X (2n) mod2 to the line buffer 61b.

  The addition unit 626a adds the floor calculation result output from the calculation unit 624a and the one-dimensional conversion coefficient Y (2n + 1) read from the line buffer 61a, and outputs the addition result to the line buffer 61b.

Here, the value output from the subtractor 623a is the calculated value T _EVEN , the value output from the calculator 622a is the correction value F _EVEN , the value output from the adder 626a is the calculated value T _ODD , and the calculator 625a outputs The value to be corrected becomes the correction value _FODD .

Next, the configuration of the odd column line processing unit 62b will be described.
FIG. 14 is a block diagram showing the configuration of the odd-numbered line processing unit of the decoder.
The odd-numbered line processing unit 62b includes a calculation unit 621b, an addition unit 622b, an addition unit 623b, a calculation unit 624b, a calculation unit 625b, an addition unit 626b, a subtraction unit 627b, a calculation unit 628b, and an addition unit. 629b, a calculation unit 630b, a calculation unit 631b, an addition unit 632b, and an addition unit 633b. Here, as an example, processing when a one-dimensional conversion coefficient Y (2n + 3) is input is shown.

The calculation unit 621b divides the input one-dimensional conversion coefficient Y (2n + 3) by 4, and outputs the remainder Y (2n + 3) mod 4 to the addition unit 622b.
The adder 622b adds the remainder Y (2n + 3) mod 4 output from the calculator 621b and the correction value F _EVEN read from the line buffer 61a, and adds the result (F _EVEN + Y (2n + 3) mod 4) to the adder 623b. Output to.

The adder 623b adds 2 to the addition result output from the adder 622b, and outputs the addition result to the calculator 624b.
The calculation unit 624b divides the addition result output from the addition unit 623b by 4, performs a floor calculation on the division result, and outputs the result to the addition unit 626b.

The calculation unit 625b divides the input one-dimensional conversion coefficient Y (2n + 3) by 4, performs a floor calculation on the division result (Y (2n + 3) / 4), and outputs the result to the addition unit 626b.
The addition unit 626b adds the floor calculation result output from the calculation unit 625b and the floor calculation result output from the calculation unit 624b, and outputs the addition result to the subtraction unit 627b.

The subtraction unit 627b subtracts the addition result output from the addition unit 626b from the calculation value T _EVEN read from the line buffer 61a, and outputs the subtraction result to the calculation unit 628b and the calculation unit 631b. Further, after the calculation of the adder 633b is completed, the subtraction result of the subtractor 627b is stored (overwritten) in the line buffer 61b.

The calculation unit 628b divides the subtraction result output from the subtraction unit 627b by 2, and outputs the remainder to the addition unit 629b.
Addition section 629b includes a calculation value _{F ODD} read from the line buffer 61b, and a remainder calculating unit 628b is output the sum, and outputs the addition result to the arithmetic unit 630b.

The calculation unit 630b divides the addition result output from the addition unit 629b by 2, performs a floor operation on the division result, and outputs the result to the addition unit 632b.
The calculation unit 631b divides the subtraction result output from the subtraction unit 627b by 2, performs a floor operation on the division result, and outputs the result to the addition unit 632b.

The adder 632b adds the floor calculation result output from the calculator 630b and the floor calculation result output from the calculator 631b, and outputs the addition result to the adder 633b.
The adder 633b adds the operation value _TODD read from the line buffer 61b and the addition result output from the adder 632b, and outputs the addition result.

Here, the subtraction result output from the subtraction unit 627b is pixel data X (2n + 2), and the addition result output from the operation addition unit 633b is pixel data X (2n + 1).
The odd-numbered line processing unit 62b stores (overwrites) the input one-dimensional conversion coefficient Y (2n + 3) in the line buffer 61a.

According to such a JPEG2000 decoder 200, the same effect as the JPEG2000 encoder 100 can be obtained. That is, by performing calculations using the correction values F _ODD and F _EVEN , it is possible to perform reversible 5 × 3 wavelet inverse transformation that does not cause an error even with only the two line buffers 61a and 61b.

  As described above, the encoder device, the decoder device, the wavelet transform method, and the wavelet inverse transform method of the present invention have been described based on the illustrated embodiment, but the present invention is not limited to this, and the configuration of each unit is as follows. Any structure having a similar function can be substituted. Moreover, other arbitrary structures and processes may be added to the present invention.

In addition, the present invention may be a combination of any two or more configurations (features) of the above-described embodiments.
(Supplementary Note 1) In an encoder device that performs 5 × 3 wavelet transform,
Input pixel data X (2n + 1) (n is an integer) and pixel data X input before the pixel data X (2n + 1) and stored in the first storage unit and adjacent to the pixel data X (2n + 1) A first conversion coefficient defining value for defining the H component conversion coefficient Y (2n + 1) using (2n) and a first conversion correction coefficient for correcting the calculation result of the H component conversion coefficient Y (2n + 1); To calculate
The L component conversion coefficient Y (2n) is defined using the pixel data X (2n) and the H component conversion coefficient Y (2n-1) already processed and stored in the second storage unit. A second conversion coefficient specified value for calculating the second conversion correction coefficient for correcting the calculation result of the L component conversion coefficient Y (2n),
The first conversion coefficient specified value and the first conversion correction coefficient are stored in the first storage unit, and the second conversion coefficient specified value and the second conversion correction coefficient are stored in the second storage unit. A filter processing unit to be stored in the storage unit;
An encoder device comprising:

  (Supplementary Note 2) The first conversion correction coefficient is a coefficient for correcting an error caused by a floor calculation of the first conversion coefficient specified value, and the second conversion correction coefficient is the second conversion coefficient specifying The encoder apparatus according to appendix 1, wherein the encoder apparatus corrects an error caused by a value floor calculation.

  (Supplementary Note 3) The pixel data X (2n + 1) is pixel data in an odd-numbered column of an original image, and the pixel data X (2n) is pixel data in an even-numbered column of the original image. The encoder device according to appendix 1.

  (Supplementary Note 4) When the pixel data X (2n + 2) continuous in one direction with respect to the pixel data X (2n) and the pixel data X (2n + 1) is input to the filter processing unit, the pixel data X (2n + 2) A first calculation unit that calculates the H component conversion coefficient Y (2n + 1) using the first conversion coefficient specified value and the first conversion correction coefficient; and the H component conversion coefficient Y (2n + 1); A second calculation unit that calculates the L component conversion coefficient Y (2n) using the second conversion coefficient specified value read from the second storage unit and the second conversion correction coefficient; The encoder device according to Supplementary Note 1, wherein

  (Additional remark 5) The said filter process part stores the said pixel data X (2n + 2) in the said 1st storage part after the calculation of a said 1st calculating part, and the said 1st conversion factor prescription | regulation value and said 1st When generating the conversion correction coefficient, the pixel data X (2n + 2) is read as the pixel data X (2n), and the H component conversion coefficient Y (2n + 1) is calculated as the second data after the calculation by the second calculation unit. The supplementary note 4 according to claim 4, wherein the second conversion coefficient prescribed value and the second conversion correction coefficient are stored in a storage unit and read out as the H component conversion coefficient Y (2n-1) when the second conversion correction coefficient is created. Encoder device.

  (Supplementary Note 6) Each of the first storage unit and the second storage unit constitutes two regions arranged in the vertical direction of two line buffers that store pixel data for one row of the original image. The encoder device as set forth in appendix 1, wherein:

  (Supplementary Note 7) The capacities of the first storage unit and the second storage unit are at most the total number of bits of the first conversion coefficient specified value and the first conversion correction coefficient or the second conversion coefficient specification. The encoder apparatus according to appendix 1, wherein the encoder device is equal to the larger of the value and the total number of bits of the second conversion correction coefficient.

(Supplementary Note 8) In a decoder device that performs 5 × 3 wavelet inverse transformation,
The input L component conversion coefficient Y (2n + 2) (n is an integer) and input before the L component conversion coefficient Y (2n + 2), stored in the first storage unit, and the L component conversion coefficient Y (2n + 2) The first inverse transform coefficient defining value for defining the pixel data X (2n + 2) using the H component transform coefficient Y (2n + 1) adjacent to the first and the calculation result of the pixel data X (2n + 2) are corrected. And the inverse transformation correction coefficient of
A second for defining pixel data X (2n + 1) using the H component conversion coefficient Y (2n + 1) and the pixel data X (2n) already processed and stored in the second storage unit. And a second inverse conversion correction coefficient for correcting the calculation result of the pixel data X (2n + 1),
The first inverse transform coefficient prescribed value and the first inverse transform correction coefficient are stored in the first storage unit, and the second inverse transform coefficient prescribed value and the second inverse transform correction coefficient are obtained. A filter processing unit to be stored in the second storage unit;
A decoder device comprising:

  (Supplementary Note 9) The first inverse transformation correction coefficient is a coefficient for correcting an error caused by floor calculation of the first inverse transformation coefficient specified value, and the second inverse transformation correction coefficient is the second inverse transformation correction coefficient. The decoder device according to appendix 8, wherein the decoder device is a coefficient for correcting an error caused by a floor calculation of the inverse transform coefficient specified value.

  (Supplementary Note 10) The pixel data X (2n + 1) is pixel data in an odd-numbered column of an original image, and the pixel data X (2n) is pixel data in an even-numbered column of the original image. The decoder device according to appendix 8.

(Supplementary Note 11) When the H component conversion coefficient Y (2n + 3) continuous in one direction with respect to the H component conversion coefficient Y (2n + 1) and the L component conversion coefficient Y (2n + 2) is input to the filter processing unit, A first calculation unit that calculates the pixel data X (2n + 2) using an H component conversion coefficient Y (2n + 3), the first inverse conversion coefficient specified value, and the first inverse conversion correction coefficient;
The pixel data X (2n + 1) is calculated using the pixel data X (2n + 2), the second inverse transform coefficient specified value read from the second storage unit, and the second inverse transform correction coefficient. 9. The decoder device according to appendix 8, wherein the decoder device comprises two arithmetic units.

  (Additional remark 12) The said filter process part stores the said H component conversion coefficient Y (2n + 3) in the said 1st storage part after the calculation of a said 1st calculating part, and the said 1st inverse conversion coefficient prescription | regulation value and the said When creating the first inverse conversion correction coefficient, the H component conversion coefficient Y (2n + 3) is read as the H component conversion coefficient Y (2n + 1), and the pixel data X (2n + 2) is calculated after the calculation by the second calculation unit. ) Is stored in the second storage unit, and the pixel data X (2n + 2) is converted into the pixel data X (2n) when the second inverse transform coefficient prescribed value and the second inverse transform correction coefficient are created. The decoder device according to appendix 11, wherein the decoder device is read as

  (Supplementary Note 13) Each of the first storage unit and the second storage unit constitutes two areas arranged in the vertical direction of two line buffers each storing data for one column of the original image. Item 9. The decoder device according to appendix 8.

(Supplementary Note 14) In the wavelet transform method for performing 5 × 3 wavelet transform,
Input pixel data X (2n + 1) (n is an integer) and pixel data X input before the pixel data X (2n + 1) and stored in the first storage unit and adjacent to the pixel data X (2n + 1) A first conversion coefficient defining value for defining the H component conversion coefficient Y (2n + 1) using (2n) and a first conversion correction coefficient for correcting the calculation result of the H component conversion coefficient Y (2n + 1); To calculate
The L component conversion coefficient Y (2n) is defined using the pixel data X (2n) and the H component conversion coefficient Y (2n-1) already processed and stored in the second storage unit. A second conversion coefficient specified value for calculating the second conversion correction coefficient for correcting the calculation result of the L component conversion coefficient Y (2n),
The first conversion coefficient specified value and the first conversion correction coefficient are stored in the first storage unit, and the second conversion coefficient specified value and the second conversion correction coefficient are stored in the second storage unit. Store in the storage,
A wavelet transform method characterized by that.

(Supplementary Note 15) In the wavelet inverse transform method for performing 5 × 3 wavelet inverse transform,
The input L component conversion coefficient Y (2n + 2) (n is an integer) and input before the L component conversion coefficient Y (2n + 2), stored in the first storage unit, and the L component conversion coefficient Y (2n + 2) The first inverse transform coefficient defining value for defining the pixel data X (2n + 2) using the H component transform coefficient Y (2n + 1) adjacent to the first and the calculation result of the pixel data X (2n + 2) are corrected. And the inverse transformation correction coefficient of
A second for defining pixel data X (2n + 1) using the H component conversion coefficient Y (2n + 1) and the pixel data X (2n) already processed and stored in the second storage unit. And a second inverse conversion correction coefficient for correcting the calculation result of the pixel data X (2n + 1),
The first inverse transform coefficient prescribed value and the first inverse transform correction coefficient are stored in the first storage unit, and the second inverse transform coefficient prescribed value and the second inverse transform correction coefficient are obtained. Storing in the second storage unit;
A wavelet inverse transformation method characterized by that.

It is a figure which shows the outline | summary of this invention. It is a block diagram which shows the encoder apparatus of embodiment. It is a block diagram which shows the structure of the wavelet transformation part of embodiment. It is a figure which shows notionally the process of a wavelet transformation part. It is a block diagram which shows the structure of a 5x3 vertical direction reversible filter. It is a block diagram which shows the structure of the odd-numbered line processing part of an encoder. It is a block diagram which shows the structure of the even number column line process part of an encoder. It is a block diagram which shows the decoder apparatus of embodiment. It is a block diagram which shows the structure of a wavelet inverse transformation part. It is a figure which shows notionally a wavelet inverse transformation process. It is a figure which shows notionally the process of a wavelet inverse transformation part. It is a block diagram which shows the structure of the 5x3 vertical direction reversible filter of a wavelet inverse transformation part. It is a block diagram which shows the structure of the even column line process part of a decoder. It is a block diagram which shows the structure of the odd column line process part of a decoder. It is a figure which shows the conventional wavelet transform It is a figure which shows the conventional wavelet transformation. It is a figure which shows the wavelet transformation part using a line buffer. It is a figure which shows the filter process of the wavelet transformation part shown in FIG. It is a figure which shows the processing flow of even number line processing. It is a figure which shows the processing flow of even number line processing. It is a figure which shows the other conventional wavelet transformation. It is a figure which shows the filter process of the wavelet transformation part shown in FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Encoder apparatus 2 1st storage part 3 2nd storage part 4 Filter processing part 10 Wavelet transformation part 11a, 11b, 61a, 61b Line buffer 12, 62 5 * 3 vertical direction reversible filter 12a, 62b Odd column line processing part 12b, 62a Even line processing unit 13a, 13b, 63a, 63b Buffer 14, 64 5 × 3 horizontal reversible filter 20 EBCOT unit 30 Stream generation unit 40 Stream analysis unit 50 Inverse EBCOT unit 60 Wavelet inverse conversion unit 100 JPEG2000 encoder 200 JPEG2000 decoder

Claims (10)

In an encoder device that performs 5 × 3 wavelet transform,
Input pixel data X (2n + 1) (n is an integer) and pixel data X input before the pixel data X (2n + 1) and stored in the first storage unit and adjacent to the pixel data X (2n + 1) A first conversion coefficient defining value for defining the H component conversion coefficient Y (2n + 1) using (2n) and a first conversion correction coefficient for correcting the calculation result of the H component conversion coefficient Y (2n + 1); To calculate
The L component conversion coefficient Y (2n) is defined using the pixel data X (2n) and the H component conversion coefficient Y (2n-1) already processed and stored in the second storage unit. A second conversion coefficient specified value for calculating the second conversion correction coefficient for correcting the calculation result of the L component conversion coefficient Y (2n),
The first conversion coefficient specified value and the first conversion correction coefficient are stored in the first storage unit, and the second conversion coefficient specified value and the second conversion correction coefficient are stored in the second storage unit. A filter processing unit to be stored in the storage unit;
An encoder device comprising:   The first conversion correction coefficient is a coefficient for correcting an error caused by the floor calculation of the first conversion coefficient specified value, and the second conversion correction coefficient is a floor calculation of the second conversion coefficient specified value. The encoder apparatus according to claim 1, wherein the coefficient is a coefficient for correcting an error caused by.   When the pixel data X (2n + 2) connected in one direction to the pixel data X (2n) and the pixel data X (2n + 1) is input to the filter processing unit, the pixel data X (2n + 2) and the first data A first calculation unit that calculates the H component conversion coefficient Y (2n + 1) using the conversion coefficient specified value and the first conversion correction coefficient; and the H component conversion coefficient Y (2n + 1) and the second conversion coefficient And a second calculation unit that calculates the L component conversion coefficient Y (2n) using the second conversion coefficient specified value read from the storage unit and the second conversion correction coefficient. The encoder device according to claim 1.   The filter processing unit stores the pixel data X (2n + 2) in the first storage unit after the calculation of the first calculation unit, and stores the first conversion coefficient specified value and the first conversion correction coefficient. At the time of creation, the pixel data X (2n + 2) is read as the pixel data X (2n), and the H component conversion coefficient Y (2n + 1) is stored in the second storage unit after the calculation by the second calculation unit. 4. The encoder apparatus according to claim 3, wherein when the second conversion coefficient specified value and the second conversion correction coefficient are created, the H component conversion coefficient Y (2n-1) is read. In a decoder device that performs 5 × 3 wavelet inverse transformation,
The input L component conversion coefficient Y (2n + 2) (n is an integer) and input before the L component conversion coefficient Y (2n + 2), stored in the first storage unit, and the L component conversion coefficient Y (2n + 2) The first inverse transform coefficient defining value for defining the pixel data X (2n + 2) using the H component transform coefficient Y (2n + 1) adjacent to the first and the calculation result of the pixel data X (2n + 2) are corrected. And the inverse transformation correction coefficient of
A second for defining pixel data X (2n + 1) using the H component conversion coefficient Y (2n + 1) and the pixel data X (2n) already processed and stored in the second storage unit. And a second inverse conversion correction coefficient for correcting the calculation result of the pixel data X (2n + 1),
The first inverse transform coefficient prescribed value and the first inverse transform correction coefficient are stored in the first storage unit, and the second inverse transform coefficient prescribed value and the second inverse transform correction coefficient are obtained. A filter processing unit to be stored in the second storage unit;
A decoder device comprising:   The first inverse transform correction coefficient is a coefficient for correcting an error caused by a floor calculation of the first inverse transform coefficient specified value, and the second inverse transform correction coefficient is the second inverse transform coefficient specified. 6. The decoder apparatus according to claim 5, wherein the decoder apparatus corrects an error caused by a value floor calculation. When the H component conversion coefficient Y (2n + 3) continuous in one direction with respect to the H component conversion coefficient Y (2n + 1) and the L component conversion coefficient Y (2n + 2) is input to the filter processing unit, the H component conversion coefficient A first computing unit that computes the pixel data X (2n + 2) using Y (2n + 3), the first inverse transformation coefficient specified value, and the first inverse transformation correction coefficient;
The pixel data X (2n + 1) is calculated using the pixel data X (2n + 2), the second inverse transform coefficient specified value read from the second storage unit, and the second inverse transform correction coefficient. 6. The decoder device according to claim 5, further comprising: 2 arithmetic units.   The filter processing unit stores the H component conversion coefficient Y (2n + 3) in the first storage unit after the calculation of the first calculation unit, and the first inverse conversion coefficient specified value and the first inverse When generating the conversion correction coefficient, the H component conversion coefficient Y (2n + 3) is read as the H component conversion coefficient Y (2n + 1), and the pixel data X (2n + 2) is calculated after the calculation by the second calculation unit. 2 and reading out the pixel data X (2n + 2) as the pixel data X (2n) when creating the second inverse transform coefficient prescribed value and the second inverse transform correction coefficient. The decoder device according to claim 7. In the wavelet transform method for performing 5 × 3 wavelet transform,
Input pixel data X (2n + 1) (n is an integer) and pixel data X input before the pixel data X (2n + 1) and stored in the first storage unit and adjacent to the pixel data X (2n + 1) A first conversion coefficient defining value for defining the H component conversion coefficient Y (2n + 1) using (2n) and a first conversion correction coefficient for correcting the calculation result of the H component conversion coefficient Y (2n + 1); To calculate
The L component conversion coefficient Y (2n) is defined using the pixel data X (2n) and the H component conversion coefficient Y (2n-1) already processed and stored in the second storage unit. A second conversion coefficient specified value for calculating the second conversion correction coefficient for correcting the calculation result of the L component conversion coefficient Y (2n),
The first conversion coefficient specified value and the first conversion correction coefficient are stored in the first storage unit, and the second conversion coefficient specified value and the second conversion correction coefficient are stored in the second storage unit. Store in the storage,
A wavelet transform method characterized by that. In the wavelet inverse transform method for performing 5 × 3 wavelet inverse transform,
The input L component conversion coefficient Y (2n + 2) (n is an integer) and input before the L component conversion coefficient Y (2n + 2), stored in the first storage unit, and the L component conversion coefficient Y (2n + 2) The first inverse transform coefficient defining value for defining the pixel data X (2n + 2) using the H component transform coefficient Y (2n + 1) adjacent to the first and the calculation result of the pixel data X (2n + 2) are corrected. And the inverse transformation correction coefficient of
A second for defining pixel data X (2n + 1) using the H component conversion coefficient Y (2n + 1) and the pixel data X (2n) already processed and stored in the second storage unit. And a second inverse conversion correction coefficient for correcting the calculation result of the pixel data X (2n + 1),
The first inverse transform coefficient prescribed value and the first inverse transform correction coefficient are stored in the first storage unit, and the second inverse transform coefficient prescribed value and the second inverse transform correction coefficient are obtained. Storing in the second storage unit;
A wavelet inverse transformation method characterized by that.

Images