JPH07249993A - Information converter - Google Patents

Abstract

PURPOSE:To provide the information converter efficiently reducing the amount of processing of the format conversion of the information in performing the communication between dissimilar types to a level of practical use. CONSTITUTION:A semi-decoding means 110 reads code strings read out from a reading means 107 and decodes them to the coefficient matrix in the conversion area. The concrete decoding procedure is the recovery of the allocated code, the two-dimensional rearrangement, and the inverse quantization. A conversion means 112 converts the (m) coefficient matrix into (n) coefficient matrix (in this case, m,n=1,2,3...) in the conversion area. A re-encoding means 114 rearranges coefficient matrix to the one-dimensional coefficient string after the prescribed quantization of the coefficient matrix. Further, the code string performing the prescribed code allocation is generated. A decoding means 116 rearranges the code strings to the two-dimensional coefficient matrix after the prescribed decoding and the coefficient matrix obtained by performing the next prescribed inverse quantization is converted into the inverse eighth order two-dimensional discrete cosine. Thus, the digital picture data in the space area is generated in succession.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information conversion device,
In particular, the present invention relates to an information conversion device used for communication between different types of multimedia communication systems such as video and audio.

[0002]

2. Description of the Related Art In a multimedia communication system,
When communication is performed between models having different resolutions, an information conversion device that converts the resolution is required. Conventionally, as an information conversion apparatus of this type, a conversion block size at the time of inverse conversion of a still image or a moving image subjected to two-dimensional discrete cosine conversion is set to a size other than the conversion block size at the time of two-dimensional discrete cosine conversion. Is known to realize resolution conversion of still images and moving images.
Volume 1, Issue 5 (1992) PP553-560 ".

Further, a component forming a predetermined number of macroblocks of a two-dimensional discrete cosine-transformed still image or moving image matches the natural frequency of this component among the components forming the target number of macroblocks. An information conversion device that realizes resolution conversion of a still image or a moving image by copying the component values has been considered.

[0004]

However, in the conventional information conversion apparatus using the former inverse conversion, since the resolution conversion is realized by changing the substantial block size at the time of the inverse conversion, the processing amount is increased. However, the processing system in which the block size is variable at the time of the inverse conversion may not be practical in terms of processing time. Further, in the latter information conversion apparatus that copies the macroblock constituent component to a component value that matches the natural frequency, if the degree of matching of the natural frequency is low, the image quality of the converted image may be significantly degraded. .

An object of the present invention is to provide an information conversion device that efficiently reduces the processing amount of format conversion of information when performing communication between different models to a practical level.

Another object of the present invention is to provide an information conversion apparatus capable of high-speed conversion processing by designating a conversion formula that gives an optimum amount of calculation for conversion accuracy.

[0007]

In order to achieve the above object, the present invention provides a code obtained by coding an information sequence obtained by sampling source information at a predetermined rate with a coding algorithm including a transform coding process. A read means for obtaining a sequence from a storage medium or a communication medium; a semi-decoding means for reading a target code sequence extracted from the read means and decoding up to a coefficient matrix composed of transform coefficients in a transform region; and a half-decoding means. Each coefficient of the generated coefficient matrix is calculated by using a conversion formula according to a desired conversion rate, and a conversion means for generating a coefficient matrix composed of conversion coefficients having a desired conversion rate, and the conversion means are extracted. Re-encoding means for reconstructing and re-encoding the transform coefficient of the coefficient matrix and re-encoding means for decoding the output code sequence of the re-encoding means and then performing inverse quantization to transform at the desired conversion rate. The de It is obtained by a configuration and a decoding means for outputting digital data.

Further, the information converting apparatus of the present invention is a storage medium or a communication medium, which stores a code string obtained by coding an information string obtained by sampling source information at a predetermined rate by a coding algorithm including a transform coding process. From the read means, a half-decoding means for reading the target code string extracted from the read means, and decoding up to a coefficient matrix made up of transform coefficients in the transform domain, and data indicating a desired transform processing precision in advance. The registering means and the coefficients of the coefficient matrix generated by the half-decoding means are calculated using a conversion formula according to the desired conversion rate to generate a coefficient matrix composed of conversion coefficients with the desired conversion rate. Based on the conversion means, the data registered by the registration means, the adjustment means for switching the conversion equation used by the conversion means, and the reconstruction coefficient of the coefficient matrix extracted from the conversion means. Re-encoding means for generating a re-encoded code string, and decoding for decoding the output code string of the re-encoding means and then performing inverse quantization to output digital data converted at the desired conversion rate. It can be configured to include a means.

[0009]

According to the present invention, the half-decoding means decodes the code string up to the transform coefficient, and then the transform means performs an operation using the transform formula to match the natural frequencies of the coefficients before and after the transform. The desired frequency component value is obtained by copying based on a plurality of conversion coefficients and a desired frequency component value is not obtained because the natural frequencies of the coefficients do not match and the desired frequency component value is written at a predetermined position. A coefficient matrix composed of conversion coefficients of the conversion rate is generated. After that, the transform coefficient of the output coefficient matrix of this transform means is re-encoded by the re-encoding means.

Therefore, in the present invention, the decoding required for the resolution conversion of the coded information may be up to the conversion coefficient. Similarly, the decoding and re-encoding procedures required for resolution conversion can be partially omitted. Further, even in the conversion mode in which the natural frequencies before and after the conversion are often inconsistent with each other, it is possible to provide a process of performing a predetermined calculation from a plurality of other coefficients.

Further, according to the present invention, by the adjusting means,
Since the conversion formula used by the conversion means is switched according to the desired conversion accuracy, it is possible to specify the conversion formula that can obtain the optimum calculation amount for the conversion accuracy without increasing the calculation amount more than necessary. it can.

[0012]

EXAMPLES Next, examples of the present invention will be described. 1 is a block diagram of a first embodiment of an information conversion apparatus of the present invention,
FIG. 2 is a block diagram of the essential parts of FIG. 1, and FIG. 3 is a flowchart for explaining the operation of the first embodiment of the present invention.

First, the configuration and functional outline of the information conversion apparatus of this embodiment will be described with reference to FIG. The information conversion apparatus of this embodiment includes a medium unit 101 and a reading unit 10.
7, the semi-decoding means 110, the conversion means 112, the re-encoding means 114 and the decoding means 116 are the main constituent elements.

The medium means 101 is an auxiliary storage device capable of storing, as a file, a code string obtained by coding an information string obtained by sampling source information at a predetermined rate by a coding algorithm including a transform coding process, The target code string can be selectively read by the reading means 107.
Here, in FIG. 1, for example, the 0th one is 102, the first one is 103, the xth one is 104, and the nth one is 105.

The reading means 107 is one of the above code strings.
This is an information processing system capable of selectively reading out a target code string (here, the code string 104). The half-decoding means 110 is an information processing system that reads the code string read by the reading means 107 and decodes the coefficient matrix in the frequency domain described later. Specific decoding procedures are restoration of assigned codes, two-dimensional rearrangement, and inverse quantization.

Further, the conversion means 112 has a frequency domain m.
N coefficient matrices (where m, n = 1,
2, 3, ...) is an information processing system. Here, a system for holding the values of m and n, and a system for setting values to this holding system from the outside are necessary, but these are secondary elements to the essence of the present invention, and the figures are unnecessary. To avoid complication, the illustration and description are omitted.

The re-encoding means 114 rearranges the coefficient matrix into a predetermined one-dimensional coefficient sequence after predetermined quantization,
Further, it is an information processing system that generates a code string to which a predetermined code is assigned.

The decoding means 116 rearranges the code sequence into a predetermined two-dimensional coefficient matrix after predetermined decoding, and then performs predetermined dequantization to obtain the coefficient matrix obtained as follows.
Further, it is an information processing system for sequentially generating digital image data in the spatial domain by performing inverse 8th order two-dimensional discrete cosine transform.

Next, the outline of the operation of this embodiment will be described with reference to FIG. First, the half-decoding means 110, the re-encoding means 114, and the decoding means 116 are quantized and dequantized, and a table for bidirectional rearrangement, and a code allocation table and a code allocation decoding table. Equivalent table values are set respectively. Next, the degree of resolution conversion of the digital image data is set by giving appropriate values to the m and n of the conversion means 112.

Next, the reading means 107 searches the medium means 101 for a compressed file (here, the code string 104) of the image data according to the request of the request source, reads it out through the code transfer path 106, and performs resolution conversion. If the image format conversion of No. 1 is required, it is output to the code transfer path 109, and if the image format conversion of resolution conversion is not necessary, it is output to the code transfer path 108.

The half-decoding means 110 reads the code string from the code transfer path 109, decodes it according to the set code allocation table value, and arranges it in a two-dimensional coefficient matrix according to the set rearrangement table value. Then, the coefficient matrix obtained by performing inverse quantization according to the set quantization table value is sequentially generated, and the coefficient transfer path 111
Output to.

The conversion means 112 reads m coefficient matrices in the frequency domain from the coefficient transfer path 111, and outputs n coefficients.
It is converted to a coefficient matrix and output to the re-encoding means 114 via the coefficient transfer path 113. The re-encoding means 114 is
The n coefficient matrices are read from the coefficient transfer path 113, quantized according to the set quantization table value, sorted into a one-dimensional coefficient string according to the set sorting table value, and further set. A code string to which code is assigned according to the code assignment table value is generated and output to the code transfer path 115.

The decoding means 116 reads the above-mentioned code string from the code transfer path 115 when the image format conversion for resolution conversion is necessary, and the code string from the code transfer path 108 when the image format conversion for resolution conversion is not necessary, and is set. After the decoding is performed according to the code allocation table value of, the two-dimensional coefficient matrix is rearranged according to the set rearrangement table value, and the inverse quantization is performed according to the set quantization table value. Next, the high-definition digital image data in the spatial domain subjected to the two-dimensional discrete cosine transformation is output to the digital image data transfer path 117.

Next, the configuration of the half-decoding means 110, the conversion means 112 and the re-encoding means 114, which are the main parts of this embodiment, will be described in more detail with reference to FIG.

As shown in FIG. 2, the half decoding means 110.
Is a code table 201, a code comparator 202, a code transfer gate 203, a difference information restorer 204, and a rearranger 20.
5, an inverse quantizer 206, and a quantization table 207. In addition, the re-encoding means 114 uses the quantizer 2
23, the rearranger 224, the difference extractor / substitute 225, and the coefficient encoder 226 are main components.

The code table 201 of the half decoding means 110
Is a table showing the correspondence between the code allocation, the decoding coefficient of the code allocation, and the code. Half-decoding means 11
The code comparator 202 of 0 is an information processing system capable of comparing the coefficient for decoding of the code allocation with the code and outputting the comparison result.

The difference information restorer 20 of the half-decoding means 110
4 is 1 for the coefficient described by the difference among the coefficients.
It is restored by substituting with the sum of the previous restored coefficient,
It is an information processing system capable of outputting the restoration result. here,
It is necessary to store the previous restored coefficient, but this is a side element to the essence of the present invention, and these are used to restore the difference information in order to prevent the figure from becoming unnecessarily complicated. It is provided in the container 204.

The rearranger 205 of the half-decoding means 110
Is an information processing system that rearranges the one-dimensional coefficient sequence 211 into a two-dimensional coefficient matrix 212 and can output the result of this rearrangement. Although it is necessary to store the rearrangement procedure, it is a secondary element to the essence of the present invention, and in order to prevent the figure from becoming unnecessarily complicated, these are the rearrangers. It shall be provided in 205.

Also, the inverse quantizer 2 of the half decoding means 110
Reference numeral 06 denotes an information processing system capable of multiplying each coefficient of the two-dimensional coefficient matrix by a predetermined value for inverse quantization and outputting the result of the multiplication. Further, the quantization table 207 of the half-decoding means 110 is a table showing the correspondence between the coefficients in the quantization and the dequantization and the values for the multiplication.

The conversion means 112 is an information processing system that obtains and outputs the matrix x from the four matrices 00, 01, 10, and 11.
Here, the number described in the frame of each matrix above indicates an address in the matrix and corresponds to each address.
Any one of 1023 (in the case of 11-bit precision) is set. This value is data in the frequency domain of the image data being processed.

Here, the value in the frame with a halftone dot of the matrix x shown by 222 in FIG. 2 is the energy level of the second harmonic of the matrix 00, 01, 10, 11 for half wavelength if necessary. It is the average value after phase adjustment. Further, the value in the frame without the halftone dot in the same row as the halftone dot frame of the matrix x is the same row of the matrix x that is the frequency group of the Fourier series expansion of the horizontal natural frequency of the frame. The values in the frame with halftone dots are summed by multiplying the values by a predetermined coefficient.

The values in the frame without halftone dot in the same column as the halftone dot frame of the matrix x are the same as those of the matrix x which is the frequency group of the Fourier series expansion of the vertical natural frequency of the frame. The sum of the values in the shaded frame of the column multiplied by a predetermined coefficient. Also, the values “9”, “27”, in the frame of the matrix x,
“45” and “63” are the values “0”, “18”, “36” and “54” in the frame of the matrix x, which is the frequency group of the Fourier series expansion of the natural frequency of the frame, The sum of the values multiplied by a predetermined coefficient is used. The values of the matrix x that cannot be obtained by performing the above calculation are the known values of the matrix x, which is the frequency group of the Fourier series expansion of the horizontal or vertical natural frequency of the frame, respectively, and the predetermined values It is the sum of the multiplied values.

This will be described more specifically. The halftone dot addresses “0”, “2” of the matrix x,
"4", "6", "16", "18" ,. ． ． , "5
4 ”is the address“ 0 ”to“ 3 ”,“ 8 ”to“ 1 ”of the four matrices 00 to 11 shown by adding 217 to 220 in FIG.
1 ”,“ 16 ”to“ 19 ”,“ 24 ”to“ 27 ”. The calculation formula is as follows.

[0034]

(Value of address 0 of matrix x) = {(value of address 0 of matrix 00) + ... + (value of address 0 of matrix 11)} / 4 (1) (address 2 of matrix x Value) = (value of address 0 of matrix 00) − (value of address 0 of matrix 01) − (value of address 0 of matrix 10) − (value of address 0 of matrix 11) (2) (address of matrix x 4 value) = (value of address 2 of matrix 00) +. ． ． ++ (value of address 2 of matrix 11) (3) Further, addresses “6”, “16”, “18” of matrix x,
"22", "38", "48", "50", "52",
As for the value of “54”, the addresses of the matrices 00 to 11 of the equation (2) are respectively set to the address of the matrix x / 2. Addresses of matrix x "4", "20", "32", "34", "3"
For the value of “6”, the addresses of the matrices 00 to 11 of the equation (3) are respectively set to the address of the matrix x / 2.

The same row as the halftone dot of the matrix x, that is, the addresses "1", "3", "5", "7", "17", "1".
9 ”,“ 21 ”,“ 23 ”,“ 33 ”,“ 35 ”,“ 3 ”
7 ”,“ 39 ”,“ 49 ”,“ 51 ”,“ 53 ”,“ 5 ”
The value of "5" is the horizontal natural frequency (Hz) of the address,
θ / 2 (addresses 1, 17, 33, 49), 3θ / 2
(Address 3, 19, 35, 51), 5θ /
2 ,. ． ． , 7θ / 2, a cosine of 7θ / 2 is obtained and an approximate expression obtained by Fourier series expansion is obtained and substituted.

[0036]

Cos θ / 2 = a ₀ cosb ₀ θ + a ₁ cosb ₁ θ + a ₂ cosb ₂ θ + a ₃ cosb ₃ θ (4) Similarly, cos (3θ / 2), cos (5θ / 2), c
An approximate expression is also obtained and substituted for os (7θ / 2).
Substitution means the value set corresponding to the address whose natural frequency is the frequency given by b ₀ θ to b ₃ θ in the equation (4), and cosb ₀ θ to cosb ₃ θ in the equation (4). Substitute as the value of. After substituting, the right side is calculated and set to the address shown on the left side.

Values that cannot be obtained by the above calculations are similarly obtained from the values obtained by the above calculations. Here, when a plurality of equations such as the equation (4) are obtained, each equation is substituted and operated, and the operation results are averaged to obtain a value.

Quantizer 223 of the re-encoding means 114
Is an information processing system that divides each coefficient of the two-dimensional coefficient matrix by a predetermined value for quantization and outputs the division result. Here, it is assumed that the quantizer 223 is provided with preprocessing such as level shift necessary for setting the division result to a value obtained by rounding off the effective minimum digit.

The rearranger 224 of the re-encoding means 114.
Is an information processing system that rearranges the two-dimensional coefficient matrix 227 into a one-dimensional coefficient column 228 and outputs the result of the rearrangement, and has means for storing the rearrangement procedure.

Difference extraction / substitution device 2 of re-encoding means 114
25 is one of the coefficients to be described as a difference among the coefficients.
An information processing system that substitutes with the difference between the previous restored coefficient and outputs the substitution result includes means for storing the previous restored coefficient. Further, the coefficient encoder 226 of the re-encoding means 114 refers to the code allocation table,
This is an information processing system that converts a coefficient into a code corresponding to the table and outputs the code.

Next, the operation of this embodiment will be described with reference to the flowchart of FIG. First, prior to the operation of this embodiment, the code table 2 of the half-decoding means 110.
01 and a table for sorting in the sorter 205,
The quantization table 207 and the re-encoding means 11
4 sets the table value in the rearrangement table in the rearranger 224.

Here, the above-mentioned rearrangement table, which arranges the one-dimensional (serial) signal sequence generally called zigzag scan in a two-dimensional matrix (usually 8 × 8), or defines the opposite operation. Then, it is generally used in the field of image coding.

Further, the conversion means 112 converts the conversion rate (n /
m) and the precision / loss ratio, the coefficient matrix 0
Corresponding relational expressions of 0, 01, 10, 11 and the coefficient of the coefficient matrix x are obtained, and the conversion procedure is established. In particular, the coefficient sequence necessary to obtain the value in the frame without dots in the same row as the frame with dots in the matrix x is the Fourier series expansion of the horizontal natural frequency of the frame without dots. It is obtained by doing, and is held while keeping the predetermined correspondence.

Further, the coefficient sequence necessary for obtaining the value in the frame without the halftone dot in the same column as the halftone dot frame of the matrix x is
The vertical natural frequency of the frame without halftone dots is obtained by performing Fourier series expansion, and is retained while maintaining a predetermined correspondence. Further, the coefficient string necessary for obtaining the value in the frame of 9, 27, 45, 63 of the matrix x is obtained by performing Fourier series expansion of the spatial frequency of the frame without the halftone dots, and the predetermined correspondence is obtained. Hold it.

Next, after the above setting, the medium means 10 of FIG.
The target 1 file is received from 1 and buffering is performed (step 301 in FIG. 3). The medium means 101 is an auxiliary storage device such as a hard disk device or a communication line (public network, LAN, etc.), and usually a plurality of files can be accessed. Therefore, a search is performed and one target file is loaded into a buffer provided in the system. Of.

Next, the four macroblocks are decoded (step 302 in FIG. 3). This is the table 201 of FIG.
And the sign comparator 202. That is, the code comparator 202 of the half-decoding means 110 uses the code table 201.
The code assignment and the code for decoding of the code assignment supplied from the code sequence and the code string and the set code supplied from the code transfer path 109 are read, and sequentially compared while changing the code until they match. 4 macroblocks, that is, 256 data (1 macroblock is 6
Variable length decoding is performed on the code string corresponding to (composed of four data), and it is output.

The code table 201 is used for the code comparator 2
The coefficient corresponding to the code supplied to 02 is simultaneously output to the code transfer gate 203 of the half decoding means 110. The code transfer gate 203 compares this input coefficient with the code comparator 202.
According to the comparison result of the difference information restorer 20
4 or is cut off to prevent output to the difference information restorer 204. Specifically, the code comparator 202
If the comparison result of indicates a sign match, it connects, and if it indicates a mismatch, disconnects.

Next, the difference information restorer 204 restores the difference of the lowest frequency component into the coefficient (step 30 in FIG. 3).
3). That is, the difference information restorer 204 reads the coefficient that has passed through the code transfer gate 203, and restores the coefficient, which is described by the difference among the coefficients, by the sum of the previous restored coefficient. On the other hand, among the coefficients, the coefficients not described by the difference are output as they are. At this time, the coefficient which is always present in one macroblock 64 data (the above coefficient) and which is described by the difference, that is, the difference-encoded data (the lowest frequency component, that is, the DC component) is immediately before. The data is restored by adding it to the similar data processed in.

Next, the rearranger 205 is placed in the virtual two-dimensional space.
To rearrange (step 304 in FIG. 3). That is, the rearranger 205 reads the above-mentioned data (coefficients) and stores them in a predetermined number, here, in each macroblock 64.
Each piece of data is arranged in a one-dimensional coefficient string 211, and the one-dimensional coefficient string 211 is rearranged in a predetermined manner in a two-dimensional coefficient matrix 212 (zigzag scanning is performed and converted into four 8 × 8 two-dimensional matrices). The rearranged two-dimensional coefficient matrix (two-dimensional matrix) 212 is output to the inverse quantizer 206 in a predetermined order for each coefficient.

Next, the inverse quantizer 206 performs inverse quantization (step 305 in FIG. 3). That is, the inverse quantizer 206 reads the two-dimensional coefficient matrix (two-dimensional matrix) 212 in each coefficient in a predetermined order, and from the quantization table 207 in which each macroblock is arranged in a matrix of the same size (8 × 8). , Corresponding values, that is, values at the same position on the matrix are sequentially read out and multiplied, and the multiplication results are assembled again in a matrix to maintain positional correspondence, and the coefficient transfer path 1 is formed as a two-dimensional coefficient matrix.
Output to 11. Here, the assembled two-dimensional coefficient matrices are the matrices 00 to 11 in FIG.

Next, the four macroblocks are converted into one macroblock (step 306 in FIG. 3). That is, the conversion means 112 converts the coefficient matrix into the coefficient transfer path 11
The matrix 0 read from 1 and indicated by 217 to 220 in FIG.
The matrix 0 is obtained by arranging the coefficients 0 to 11 in order in correspondence with each other, and determining the resolutions m and n of the resolution of the set digital image data in the horizontal and vertical directions respectively.
0 to 11 and the coefficient of the matrix x and the coefficient sequence for calculation, the matrix x is calculated from the values of the predetermined coefficients of the matrices 00 to 11.
Each coefficient is calculated by the above-mentioned calculation as needed, and is output to the coefficient transfer path 113.

Next, the quantizer 223 of the re-encoding means 114 performs quantization (step 307 in FIG. 3).
That is, the quantizer 223 uses the coefficient transfer path 113.
The coefficient matrix x of one macroblock read from is divided by the corresponding quantization coefficient from the quantization table 207 of the same size (8 × 8) (that is, between the values at the same position on the matrix).

Next, the rearranger 224 of the re-encoding means 114 performs one-dimensional rearrangement (step 308 in FIG. 3). That is, the rearranger 224 rearranges the data of 64 macroblocks, which is the two-dimensional coefficient matrix 227, into a one-dimensional coefficient string 228 by zigzag scanning.

Next, the difference extracting / substituting unit 225 of the re-encoding means 114 obtains and substitutes the difference of the coefficient of the lowest frequency component (step 309 in FIG. 3). That is,
The differential extraction / substitution unit 225 generates differential encoded data in which one always exists among 64 macroblock data which is a one-dimensional coefficient sequence input from the rearrangement unit 224. This is performed by obtaining the difference from the similar data of the macroblock processed immediately before, which is the immediately preceding coefficient sequence, and substituting it. However, if there is no previous data, nothing is done.

Subsequently, the coefficient encoder 226 sequentially encodes one macroblock (step 31 in FIG. 3).
0). That is, the coefficient encoder 226 reads the difference encoded data input from the difference extractor / substitute 225, and sets the set value according to the variable length dictionary for data conversion defined on the set code assignment table 201. The matching variable length code is output to the code transfer path 115.

Next, it is judged whether or not the encoding for one file has been completed (step 311 in FIG. 3). If not completed, the process returns to the step 302 and the processes of the steps 302 to 310 are executed again. . When the encoding for one file is completed, the one file is formed into a predetermined number of packets and then transmitted to the medium (step 31 in FIG. 3).
2).

That is, since the above processing is processing for every 4 or 1 macroblocks, in the case of conversion from normal high definition to standard definition, by repeating the above processing 20480 times, one screen of image data can be obtained. Is read / written as one file.

As described above, according to the present embodiment, since the decoding process required for the resolution conversion of the coded image is up to the transform coefficient in the frequency domain, the conventional spatial domain can be decoded. The required processing load can be reduced as compared with the case where it is performed. Further, since the decoding and re-encoding procedures required for resolution conversion of the encoded image are partially omitted, it is possible to eliminate errors in the omitted procedures.

Further, even in the conversion mode in which the natural frequencies before and after the conversion are largely different from each other, the image quality can be improved with a small amount of calculation by providing a process of performing a predetermined calculation from a plurality of other coefficients.

Next, a second embodiment of the present invention will be described. FIG. 4 is a block diagram of a second embodiment of the information converting apparatus according to the present invention, FIG. 5 is a configuration diagram of a main part of FIG. 4, and FIG. 6 is a flowchart for explaining the operation of the second embodiment of the present invention.

First, referring to FIG. 4, an outline of the configuration and function of the information conversion apparatus of this embodiment will be described. The information conversion apparatus of this embodiment includes a medium unit 101 and a reading unit 10.
7, the half-decoding means 110, the conversion means 112, the re-encoding means 114, the decoding means 116, the registration means 401, and the adjustment means 403 are the main components. The same components as those in FIG. 1 are designated by the same reference numerals and the description thereof will be omitted.

When the registering means 401 transforms m coefficient matrices in the transform frequency domain of the half decoding means 110 into n coefficient matrices (here, m, n = 1, 2, 3 ...). Is a system for storing the switching information of the coefficient conversion formulas and outputting the switching information of the coefficient conversion formulas as needed, and here is a small-scale memory that can be output to the adjusting means 403.

The adjusting means 403 converts the m coefficient matrices in the frequency domain into n coefficient matrices (where m, n =
1, 2, 3 ...) is a control system that controls the conversion accuracy of the conversion based on the switching information by switching the used conversion formula.

Next, the outline of the operation of the information conversion apparatus according to the present invention will be described with reference to FIG. As described above, first, the semi-decoding means 110, the re-encoding means 114, and the decoding means 11
The table values for the quantization and the dequantization, the table for the bidirectional rearrangement, and the table for the code allocation and the decoding for the code allocation of 6 are set to the same table values, respectively.

Next, the registration means 401 and the adjustment means 40
In 3, the conversion formula selection information and each conversion formula are set. Next, the conversion means 112 is set by giving an appropriate value to the resolution conversion degree of the digital image data for m and n (here, m = 4, n = 1).

Next, the reading means 107 searches for the compressed file of the image data, that is, the code string x according to the request of the request source, reads it out through the code transfer path 106, and when the image format conversion of resolution conversion is necessary. To the code transfer path 109, and to the code transfer path 108 when image format conversion for resolution conversion is unnecessary.

Next, the semi-decoding means 110 reads the code string from the code transfer path 109, decodes it according to the set code allocation table value, and then uses the two-dimensional coefficient according to the set rearrangement table value. The matrix is rearranged, and then the inverse quantization is performed according to the set quantization table value to sequentially generate the coefficient matrix, and the coefficient matrix is output to the coefficient transfer path 111.

The conversion means 112 reads m coefficient matrices in the frequency domain from the coefficient transfer path 111, and uses a conversion equation that maintains the accuracy according to the adjustment content registered by the registration means 401, to obtain n coefficients. It is converted into a coefficient matrix and output to the coefficient transfer path 113.

Next, the re-encoding means 114 reads the coefficient matrix from the coefficient transfer path 113, quantizes it according to the set quantization table value, and one-dimensional coefficient string according to the set rearrangement table value. Then, the code sequence is rearranged and the code sequence is generated according to the set code assignment table value, and the code sequence is output to the code transfer path 115.

Next, the decoding means 116 uses the code transfer path 11 when image format conversion for resolution conversion is necessary.
5, when the image format conversion of resolution conversion is not necessary, the code string is read from the code transfer path 108, decoded according to the set code allocation table value, and the two-dimensional code is read according to the set rearrangement table value. The coefficient matrix is rearranged, and then the inverse quantization is performed according to the preset quantization table value, and the high-definition digital image data 117 in the spatial domain subjected to the inverse 8th order two-dimensional discrete cosine transform is transferred to the different definition digital image data. Output to the route 117.

Next, the configuration of the half-decoding means 110, the conversion means 112, the re-encoding means 114, the registration means 401 and the adjusting means 403, which are the main parts of this embodiment, will be described in more detail with reference to FIG. FIG. 5 is a configuration diagram of a main part of FIG. 4, and the same components as those in FIGS. 2 and 4 are designated by the same reference numerals and the description thereof will be omitted.

As shown in FIG. 5, the half decoding means 110.
Is a code table 201, a code comparator 202, a code transfer gate 203, a difference information restorer 204, and a rearranger 20.
5, an inverse quantizer 206, and a quantization table 207. In addition, the re-encoding means 114 includes the quantizer 22.
3, the rearranger 224, the difference extractor / substitute 225, and the coefficient encoder 226 are main components.
These are the same as the configuration of the first embodiment shown in FIG.

In FIG. 5, the conversion means 112 uses the matrix 0.
An information processing system that obtains and outputs a matrix x from four matrices 0, 01, 10, and 11 using a conversion formula, and the conversion operation itself is the same as that of the first embodiment. However, since the above-mentioned conversion formula is a polynomial approximation formula, the amount of calculation increases more than necessary when trying to realize higher precision when the precision is relatively high.
Therefore, in the present embodiment, a conversion formula having a calculation amount corresponding to the required image quality is prepared in advance in the conversion means 508, 509, 510, 511, and is registered in the registration means 401 at the time of use to generate an appropriate conversion expression. specify.

Here, the conversion means 508, 509, 51
0, 511 and switches 5 provided on their input side
04, 505, 506 and 507 respectively constitute the adjusting means 403. This switch 504 ~
Any one of 507 is turned on by the output of the registration means 401, and acts so as to use the conversion formula of the corresponding conversion means.

In FIG. 5, the conversion means 508 to 511 are shown.
16 conversion formulas and a maximum of 64 conversion formulas are required. Of these, the conversion means 508 is shown to have the minimum of 16 conversion expressions, but the other conversion means 5
Although not shown, 09 to 511 have 16 to 64 conversion formulas.

Next, the operation of this embodiment will be described with reference to the flowchart of FIG. 6, those steps which are the same as those corresponding steps in FIG. 3 are designated by the same reference numerals, and a description thereof will be omitted. Also in this embodiment, similarly to the first embodiment, step 301
Of the half-decoding means 110 prior to the above processing.
01 and a table for sorting in the sorter 205,
The quantization table 207 and the re-encoding means 11
4 setting table values for each of the sorting tables in the sorting unit 224, and the conversion means 1
12 conversion rates (n / m) and precision / loss ratios are set.

In particular, the coefficient sequence necessary for obtaining the value in the frame without dots in the same row as the frame with dots in the matrix x shown by 222 in FIG. The horizontal natural frequency that is possessed is obtained by performing Fourier series expansion, and is retained while maintaining a predetermined correspondence. Further, the coefficient sequence necessary for obtaining the value in the frame without a dot in the same column as the frame with a dot in the matrix x is subjected to Fourier series expansion of the vertical natural frequency of the frame without a dot. It is obtained by the above, and is held while maintaining a predetermined correspondence. Also, the matrix x 9,2
The coefficient sequence necessary for obtaining the values within the frames of 7, 45, and 63 is obtained by performing Fourier series expansion on the spatial frequency of the above-mentioned frame without halftone dots, and held while maintaining a predetermined correspondence.

Further, in the present embodiment, in order to set the conversion formula of the optimum calculation amount corresponding to the image quality of the converted image, the conversion unit having the conversion formula among the conversion units 508, 509, 510 and 511. The data specifying the registration means 4
Register in 01 beforehand.

After the above initialization, as shown in FIG. 6, after receiving the packet of the medium means 101 from the reading means 107 and buffering the target 1 file data (step 301), 4 macroblocks. Is decoded, the difference of the lowest frequency component is restored to a coefficient (steps 302 and 303), rearranged in a virtual two-dimensional space, and then dequantized (steps 304 and 305).

Next, according to the conversion processing accuracy registered in the registration means 401 of FIG. 5, that is, the data classified into requirements for the image quality obtained by the conversion, the conversion set for each area or each image. The expression (degree of conversion) is read (step 601).

Here, as the image quality, for example, "POOR" for obtaining only the halftone dot portion of the matrix x, "FAIR" for obtaining the halftone dot portion of the matrix x and the same row as the halftone dot, and the halftone dot of the matrix x "GOOD" for obtaining the same row and column as the dot portion and the halftone dot, and "FIN" for obtaining all the halftone dot portions of the matrix x (a single expression for a portion that is not the same row or column as the halftone dot portion)
E ", all the halftone dot parts of the matrix x (the part which is not the same row or column as the halftone dot part is the average of plural expressions)
There is LENT.

Next, the conversion formula information read from the registration means 401 is sent to the switch 50 via the control path 402.
4 to 507 are respectively applied as control information, and one of them is selected to switch the conversion formula (step 60).
2). In FIG. 5, the switch 505 is turned on and the conversion formula of the conversion means 509 is selected. As a result, the conversion is performed by the conversion unit 112 using the conversion formula with the accuracy corresponding to the required image quality.

Next, the conversion means 112 performs conversion from 4 macroblocks to 1 macroblock using the switched conversion formulas (step 603). That is,
A specified conversion formula is used when obtaining each value of one macroblock from four macroblocks. Generally, the shaded portion of the matrix x in FIG. 5 can perform high-precision processing with a relatively small amount of calculation, and therefore, a common equation is given to any conversion equation. On the other hand, the portion of the matrix x other than the shaded area has a relatively small effect on the amount of calculation, so that the user has a choice.

After that, as in the first embodiment, the quantizer 2
Quantization by 23 (step 307), rearranger 22
4 to rearrange the two-dimensional coefficient sequence 227 into the one-dimensional coefficient matrix 228 (step 308), the difference extraction / substitution device 22.
Substitution processing for the coefficient described by the difference of 5 (step 309) and sequential encoding of one macroblock by the coefficient encoder 236 (step 310) are sequentially performed.

The above procedure is performed until the end of one image file, and finally one image file is packetized into a predetermined number and then sent to the medium to end (step 31).
1, 312). As described above, according to the present embodiment, in addition to the effect of the first embodiment, there is an effect that the resolution conversion can be performed by the optimum calculation amount according to the target image quality.

[0086]

As described above, according to the present invention,
Since the decoding required for resolution conversion of the encoded information is limited to the conversion coefficient, the required processing load can be reduced as compared with the case where decoding is performed up to the conventional spatial domain.

Further, according to the present invention, since the decoding and re-encoding procedures required for resolution conversion can be partially omitted, the error in the omitted procedure can be eliminated.

Further, even in the conversion mode in which the natural frequencies before and after the conversion are largely different from each other, the image quality can be improved with a small amount of calculation by providing a process of performing a predetermined calculation from a plurality of other coefficients.

Further, in the present invention, the conversion formula used in the conversion means is switched according to the desired conversion precision, so that the optimum calculation amount for the conversion precision can be obtained without increasing the calculation amount more than necessary. Since an expression is specified, high-speed conversion processing can be performed.

[Brief description of drawings]

FIG. 1 is a block diagram of a first embodiment of an information conversion apparatus of the present invention.

FIG. 2 is a configuration diagram of a main part of FIG.

FIG. 3 is a flowchart for explaining the operation of the first embodiment of the present invention.

FIG. 4 is a block diagram of a second embodiment of the information conversion system of the invention.

5 is a configuration diagram of a main part of FIG.

FIG. 6 is a flowchart for explaining the operation of the second embodiment of the present invention.

[Description of Codes] 101 ... Medium Means, 107 ... Read Means, 110 ... Half Decoding Means, 112 ... Conversion Means, 114 ... Re-Encoding Means, 116 ... Decoding Means, 201 ... Code Table, 20
2 ... Sign comparator, 203 ... Coefficient transfer gate, 204 ... Difference information restorer, 205, 224 ... Rearranger, 206 ...
Inverse quantizer, 207 ... Quantization table, 217, 21
8, 219, 220, 222 ... Matrix 00, 01, 10, 11, X, 2
23 ... Quantizer, 225 ... Difference extractor / alternator, 226 ...
Coefficient encoder, 401 ... Registration means, 403 ... Adjustment means, 504, 505, 506, 507 ... Switch, 50
8,509,510,511 ... Conversion means.

Claims (1)

[Claims] 1. A reading means for obtaining, from a storage medium or a communication medium, a code string obtained by coding an information string obtained by sampling source information at a predetermined rate by a coding algorithm including a transform coding process, and the reading. A half-decoding means for reading the target code string extracted from the means and decoding up to a coefficient matrix composed of transform coefficients in the transform domain; and each coefficient of the coefficient matrix generated by the half-decoding means,
A conversion unit that performs a calculation using a conversion formula according to a desired conversion rate to generate a coefficient matrix composed of conversion coefficients with a desired conversion rate; and a conversion coefficient of the coefficient matrix extracted from the conversion unit is reconstructed and Re-encoding means for generating a re-encoded code string, and decoding for decoding the output code string of the re-encoding means and then performing inverse quantization to output digital data converted at the desired conversion rate And an information conversion device.