JP5017290B2 - Decoded signal reconstruction device, decoding device, and program - Google Patents

Description

  The present invention relates to a decoded signal reconstruction device, a decoding device, and a program in an encoding scheme involving orthogonal transformation and quantization.

  In the irreversible encoding method with conventional degradation, as shown in FIG. 5, mainly an orthogonal transform processing device 11 that performs orthogonal transform on an input data signal and the orthogonal transform coefficient are quantized. Encoding is performed by the encoding device 10 including the quantization processing device 12.

  On the other hand, as shown in FIG. 6, a decoding device 21 that decodes the encoded signal encoded by the encoding device 10 and an inverse quantization processing device 22 that performs inverse quantization, and inverse orthogonality for the obtained orthogonal transform coefficients. An inverse orthogonal transform processing device 23 that performs transformation and a saturation processing device 24 that performs saturation processing when the decoded signal exceeds the dynamic range are provided. Usually, signals can be reconstructed by inverse orthogonal transform processing. Originally, the orthogonal transformation is a transformation process that does not involve deterioration, and the encoding apparatus 10 can faithfully orthogonally transform the signal within the dynamic range of the input signal.

  However, a signal that loses some information may be generated by quantization after orthogonal transformation. In this case, the decoding device 21 may decode the signal as if a signal outside the dynamic range was observed. When such an originally nonexistent signal is reproduced, the decoding device 21 performs saturation processing for rounding the signal into the dynamic range by the saturation processing device 24.

  In many systems, as this saturation processing, a signal outside the dynamic range is limited to its maximum value or minimum value and reproduced. Such saturation processing is referred to as clip processing or saturation processing. For example, the clip processing is MPEG-2 or H.264. It is also adopted in the H.264 encoding method.

  The saturation processing is processing that faithfully reproduces the transmitted signal and limits the signal within an allowable range of the decoding device 21 because lost information cannot be restored by the decoding device. However, such a signal is originally a signal that is always limited within the dynamic range as an input signal of the encoding device 10, and that a signal outside the dynamic range is decoded by the decoding device 21 is necessarily encoded. This is a phenomenon that confirms that the unit signal contains deterioration (information loss).

  That is, not only the signal reproduced outside the dynamic range but also the signal reproduced within the dynamic range includes deterioration. Therefore, in consideration of the fact that the decoding device 21 has to restore the signal within the dynamic range of the original signal, the signal reproduced within the dynamic range of the encoding unit is also brought closer to the true signal value, and saturation is achieved. There is a restoration method that realizes an effect of reducing deterioration as well as processing.

  A Gerchberg-Papoulis iterative algorithm is known as such a limited signal restoration method (see Non-Patent Document 1, for example). In addition, a technique for increasing the resolution of an image using the principle of the GP iteration method is known (for example, see Non-Patent Document 2).

New Edition Image Analysis Handbook Function Part I Projection 2.8 Super Resolution Eiji Niibori and Mikio Takagi “High-quality image expansion using Gerbert-Papoulis iteration method using DCT”, IEICE Transactions, D-II, Information and Systems, II-Information Processing, Vol. J76-D-2, no. 9, 1993, September 25, pp. 1932-1940

  As described above, in a generally irreversible encoding technique, an efficient compression effect is realized by a combination of the orthogonal transform processing device 11 and the orthogonal transform coefficient quantization processing device 12. In the decoding device 21, the orthogonal transformation coefficient restored by the inverse quantization processing device 22 is restored by the inverse orthogonal transformation processing device 23. Information is deteriorated by this irreversible quantization process, and a signal value outside the original signal existence range (dynamic range) may be reproduced when the information is restored.

  For this reason, the decoding device 21 includes a saturation processing device 24 that limits a signal restored outside the dynamic range within the dynamic range. In general, the saturation processing device 24 performs processing to replace a signal outside the dynamic range with the minimum value or the maximum value, and retains the signal restored within the dynamic range as it is.

  Furthermore, in the iterative algorithm of Gerchberg-Papoulis, when the inverse quantization process is performed, the signal range-orthogonal transformation is performed on the signal outside the measurable frequency component of the observation system and the signal value without degradation. It is repeated in the region and expressed by replacing each other's values to be within known limits, producing a signal that is substantially outside the measurable range.

  In this case, a condition is necessary that the true (or very true) values of some signals are known. Such processing is processing called super-resolution because a signal outside the measurable range is restored in principle.

  However, when the inverse quantization process is performed, the signal is simply limited in the signal domain-orthogonal transform domain in the same way as the GP iteration method, and the degraded signal is restored to a signal outside the measurable range. Rather than the orthogonal transformation used in many coding schemes for the signal after the orthogonal transformation, for example, a coding scheme using both waveform analysis and quantization such as DCT, DFT, and DST is an irreversible coding scheme. Therefore, since all signals (orthogonal transform coefficients) include an error, it is desired to improve the decoded signal by reducing this error as well.

  Accordingly, an object of the present invention is to provide a decoded signal reconstructing device, a decoding device, and a program for reducing errors in orthogonal transform coefficients of all signals in an irreversible encoding method.

  The present invention improves this saturation processing and not only limits the signal restored outside the dynamic range within the dynamic range, but also processes the signal restored within the dynamic range based on the signal limit. Thus, there are provided a decoded signal reconstruction device, a decoding device, and a program that also have an effect of restoring lost information of an encoded signal.

The decoded signal reconstruction apparatus of the present invention inputs a coded signal in a coding scheme involving orthogonal transformation and quantization, and decodes a decoded signal that reconstructs a decoded signal obtained by performing inverse quantization and inverse orthogonal transformation. When the signal after the inverse orthogonal transform includes a signal outside the predetermined range, the component device is a saturation processing unit that performs saturation processing, and performs orthogonal transform processing on the signal subjected to saturation processing, and re-orthogonal transform A re-orthogonal transform means for transmitting coefficients, a pre-orthogonal transform coefficient obtained when the encoded signal is input and inversely quantized, and the re-orthogonal transform coefficient, and the absolute value of these difference values Is less than the value obtained by multiplying the quantization range of the orthogonal transform coefficient by a weighting coefficient that changes according to the energy distribution of the orthogonal transform coefficient, select the re-orthogonal transform coefficient value ; otherwise, select the previous orthogonal transform coefficient Orthogonal transform to select and send The number comparing means, by performing inverse orthogonal transformation on the orthogonal transformation coefficients sent from the orthogonal transform coefficient comparison means, characterized in that it comprises an inverse orthogonal transform means for generating a decoded signal.

  In the decoded signal reconstruction apparatus of the present invention, the output of the inverse orthogonal transform unit is re-input to the saturation processing unit and is repeated a predetermined number of times.

  The present invention is also characterized as a decoding device including the decoded signal reconstruction device of the present invention.

In addition, the present invention is configured as a decoded signal reconstructing device that inputs an encoded signal in an encoding scheme involving orthogonal transformation and quantization and reconstructs a decoded signal obtained by performing inverse quantization and inverse orthogonal transformation. When the computer to which the signal after inverse orthogonal transformation includes a signal outside the predetermined range, the step of performing the saturation processing, and performing orthogonal transformation processing on the signal subjected to the saturation processing and sending the re-orthogonal transformation coefficient Comparing the pre-orthogonal transform coefficient obtained when the coded signal is input and inversely quantized with the re-orthogonal transform coefficient, and the absolute value of these difference values is quantized to the orthogonal transform coefficient Selecting a re-orthogonal transform coefficient value if the range is less than a value multiplied by a weighting factor that changes corresponding to the energy distribution of the orthogonal transform coefficient ; otherwise, selecting and sending a pre-orthogonal transform coefficient; , By performing inverse orthogonal transformation on the orthogonal transformation coefficient sent from the serial orthogonal transformation coefficient comparison means, and generating a decoded signal, also characterized as a decoded signal reconstruction program for execution.

  According to the present invention, it is possible to reproduce an encoded signal that is more faithful to the original signal in an encoding method that involves degradation due to quantization.

  In particular, according to the present invention, not only the signal is limited within the dynamic range, but also the encoding distortion included in the signal within the dynamic range is simultaneously reduced. In particular, the phenomenon in which a decoded signal outside the dynamic range is reconstructed occurs when a sudden level change is included in the input signal. As an example of this, in the case of a video signal, there are many in the vicinity of the subtitle character font or in the periphery of the outline of the object included in the video, and in the irreversible encoding, the signal degradation of such a portion appears remarkably. Such deterioration can be reduced by using the present invention.

It is a figure which shows the decoding apparatus of the Example by this invention. It is a figure which shows the decoded signal reconstruction apparatus of the Example by this invention. It is an operation | movement flowchart of an example of the decoding apparatus of the Example by this invention. It is a figure which shows roughly the effect of the decoding apparatus of the Example by this invention. It is the schematic of the conventional encoding apparatus. It is the schematic of the conventional decoding apparatus.

  Hereinafter, a decoded signal reconstruction device, a decoding device, and a program according to embodiments of the present invention will be described.

  FIG. 1 shows a decoding apparatus according to an embodiment of the present invention. The decoding device 1 of the present embodiment is a device that decodes a signal encoded by an encoding device based on an existing irreversible encoding method as shown in FIG. A dequantization processing device 2 that performs quantization, an inverse orthogonal transformation processing device 3 that performs inverse orthogonal transformation on the obtained orthogonal transformation coefficient, and a decoded signal reconstruction device 4.

  FIG. 2 shows a decoded signal reconstruction apparatus according to an embodiment of the present invention. The decoded signal reconstruction device 4 of this embodiment includes a saturation processing unit 40, an orthogonal transformation processing unit 41, an orthogonal transformation coefficient comparison unit 42, and an inverse orthogonal transformation processing unit 43.

In the case where the saturation processing unit 40 includes the signal after the inverse orthogonal transformation supplied from the inverse orthogonal transformation processing device 3, the saturation processing similar to the conventional case, that is, the signal after the inverse orthogonal transformation includes a signal outside the determined range. in, subjected to saturation processing, and sends the signal resulting to the orthogonal transform processing section 41.

  The orthogonal transform processing unit 41 generates a re-orthogonal transform coefficient obtained by performing orthogonal transform processing on the signal subjected to saturation processing.

The orthogonal transform coefficient comparison unit 42 compares an orthogonal transform coefficient sequence (hereinafter referred to as a pre-orthogonal transform coefficient sequence) obtained when an encoded signal is input and dequantized, and the re-orthogonal transform coefficient. When the absolute value of these difference values is within a predetermined limit value, the re-orthogonal transform coefficient value is selected, and otherwise, the pre-orthogonal transform coefficient is selected and sent to the inverse orthogonal transform processing unit 43 . Predetermined limit value for this is suitably a limiting value weighted that changes in response to the energy distribution of the orthogonal transform coefficients.

  The inverse orthogonal transform processing unit 43 performs inverse orthogonal transform on the input re-orthogonal transform coefficient sequence or the pre-orthogonal transform coefficient sequence (that is, the orthogonal transform coefficient transmitted from the orthogonal transform coefficient comparison unit), and the decoded data Send a signal. Note that the decoding device 1 may be configured to re-input the output of the inverse orthogonal transform unit 43 to the saturation processing unit 40 and repeat the output a predetermined number of times.

  Further, as one aspect of the present invention, the decoding device 1 can be configured as a computer, and a program for realizing each function is stored in a storage unit (not shown) provided inside or outside each computer. The The data signal input by the decoding device 1 can be stored in this storage unit. Such a storage unit can be realized by an external storage device such as an external hard disk or an internal storage device such as ROM or RAM. The control unit that executes the program can be realized by a central processing unit (CPU) or the like. That is, the CPU can appropriately read from the storage unit a program in which the processing content for realizing the function of each component is described, and implement each device on the computer. Here, the function of any means may be realized by all or part of the hardware.

  In the embodiment described above, the program describing the processing content for realizing the function of the decoding device 1 can be recorded on a computer-readable recording medium. As the computer-readable recording medium, any recording medium such as a magnetic recording device, an optical disk, a magneto-optical recording device, and a semiconductor memory may be used.

  The input data signal may be input via a portable recording medium such as a DVD or a CD-ROM, or may be input via an interface dedicated to video input.

  Next, an example of the operation of the decoding device 1 will be described with reference to FIG.

  In step S1, the decoding apparatus 1 inputs an encoded data signal for a certain pixel block.

  In step S2, the inverse quantization processing device 2 performs inverse quantization to generate a DCT sequence J [k].

  In step S <b> 3, the inverse orthogonal transform processing device 3 performs inverse orthogonal transform on the DCT sequence J [k], and then the saturation processing unit 40 executes saturation processing.

  In step S4, a re-orthogonal transform coefficient sequence (DCT sequence j [k]) obtained by performing orthogonal transform again on the signal subjected to saturation processing by the orthogonal transform processing unit 41, and a pre-orthogonal transform after inverse quantization processing For the coefficient sequence (DCT sequence J [k]), the orthogonal transform coefficient comparison unit 42 calculates the absolute value of these differences.

  In step S5, the orthogonal transform coefficient comparison unit 42 determines whether or not the calculated difference value is within the quantization range.

  In step S6, when the difference value calculated by the orthogonal transform coefficient comparison unit 42 is not within the quantization range, the DCT sequence j [k] obtained by the saturation process is selected, and a new DCT sequence j [k] is selected. Replace with].

  In step S7, when the difference value calculated by the orthogonal transform coefficient comparison unit 42 is within the quantization range, the DCT sequence J [k] after the inverse quantization process is selected, and a new DCT sequence j [ k].

  In step S8, the inverse orthogonal transform processing unit 43 performs inverse DCT transformation on the new DCT sequence j [k] and stores it in the decoded data signal sequence i.

  In step S9, the decoding apparatus 1 determines whether or not the processing has been completed for all the pixel blocks. If the processing has not been completed, the decoding device 1 repeats the processing for another pixel block from step S1 and ends for all the pixel blocks. If so, the decoding process ends. Further, step S9 may be repeated for a predetermined number of times from step S4 to S8 and then processed for other pixel blocks.

  The above example program will be described below with reference to a description example in C language.

  First, the saturation processing procedure will be described. When encoding with 8-bit accuracy (dynamic range: 0 to 255), the signal value is limited to a range of 0 to 255 in the dynamic range of the signal. The signal subjected to the inverse orthogonal transform is subjected to the following processing procedure on the decoded signal in order to limit the signal within this range.

[Description example in C language]
i = idct (J); // Function to inverse DCT transform DCT coefficient sequence J and store the value in decoded signal sequence i
for (k = 0; k <N; k ++)
i [k] = (i [k] <0? 0: (i [k]> 255? 255: i [k])); // i [k]: decoded signal

  The above process is a process of using a decoded signal as it is for a signal decoded within the dynamic range and limiting the signal decoded outside the dynamic range to its minimum value and maximum value. In other words, if i <0, the signal is limited to the minimum value 0, if i> 255, the signal is limited to the maximum value 255, and if 255 ≧ i ≧ 0, i is kept.

  The simplest description example of the decoded signal reconstruction device 4 is as follows.

[Description example in C language]
void reconstruct (int i []) {
int j [64];
int k;
j = fdct (i); // function to DCT transform signal sequence i and store value in DCT transform coefficient sequence j
for (k = 0; k <N; k ++) {
j [k] = fabs (j [k] -J [k]) <q [k]? j [k]: J [k];
// DCT coefficient (re-DCT coefficient value) sequence j [k] of the clipped signal,
// Received stored DCT coefficient sequence J [k],
// compare and if the difference is within the quantization range
// Select re-DCT coefficient value
}
i = idct (j); // function for inverse DCT transforming DCT coefficient sequence j and storing the value in decoded signal sequence i
for (k = 0; k <N; k ++) {
i [k] = (i [k] <0? 0: (i [k]> 255? 255: i [k]));
}
}

  The simplest description example of the decoding device 1 is as follows.

In the above iterative process (within the reconstruct function), the orthogonal transform coefficient is converted into the quantization range as follows:
j [k] = fabs (j [k] -J [k]) <q [k]? j [k]: J [k];
However, it is not limited to this mode.

  For example, when encoding processing is performed on video, generally, high-frequency components are roughly quantized according to human visual characteristics, so q [k] in the equation is larger than the low-frequency component values. It is done.

In such a case, the width of the DCT coefficient value to be corrected becomes large and the error becomes large. So in such a system:
j [k] = fabs (j [k] -J [k]) <w [k] * q [k]? j [k]: J [k];
Multiplying the limit condition by a weighting factor will be more effective. Here, w [k] is a weighting coefficient corresponding to the component.

  This w is a parameter designed in a system corresponding to the input signal on the encoding device side. Hereinafter, an example method for setting this w will be described.

  In the case of video signal encoding, DCT is generally used for encoding as orthogonal transform, and the DCT coefficient energy distribution exists in a Laplace distribution (exponential function distribution) as a general video signal property. is there. Therefore, it is preferable to design this w so as to follow the energy distribution of the signal.

For example, when using an 8 × 8 two-dimensional DCT and performing weighting linearly,
w [8 * l + n] = (1.0-l / 7) * (1.0-n / 7)
And when weighting exponentially,
w [8 * l + n] = exp (-(l ² + n ² ))
And Note that l and n are the indexes of the matrix in the 8 × 8 two-dimensional DCT.

  FIG. 4 schematically shows the effects obtained by the above description. For example, consider a case where a data signal having an amplitude of A1 in the signal string direction is encoded and decoded. The decoded data signal obtained by the conventional saturation process clips the Sd portion of the rapidly changing amplitude portion within the dynamic range, or creates a signal outside the dynamic range of the Sd portion. It has no effect on the ripple R1 in the region within the dynamic range. On the other hand, according to the decoding device 1 of the present embodiment, since DCT coefficient replacement based on saturation processing is involved, a reduced ripple R2 (<R1) in a region within another dynamic range can be obtained. This means that errors in orthogonal transform coefficients of all signals in the lossy encoding method are reduced.

  Although the decoding apparatus 1 of the above-described embodiment has been described as a representative example of processing for DCT, it will be apparent to those skilled in the art that many changes and substitutions can be made within the spirit and scope of the present invention. For example, DST or the like can be used instead of DCT. Accordingly, the invention should not be construed as limited by the embodiments described above, but only by the claims.

  According to the present invention, the lost information of the encoded signal is also restored by processing the signal restored within the dynamic range based on the signal limit amount. Useful for applications that require saturation processing.

DESCRIPTION OF SYMBOLS 1 Decoding device 2 Inverse quantization processing device 3 Inverse orthogonal transformation processing device 4 Decoded signal reconstruction device 10 Encoding device 11 Orthogonal transformation processing device 12 Quantization processing device 21 Decoding device 22 Inverse quantization processing device 23 Inverse orthogonal transformation processing device 24 Saturation processing device 40 Saturation processing unit 41 Orthogonal transformation processing unit 42 Orthogonal transformation coefficient comparison unit 43 Inverse orthogonal transformation processing unit

Claims (4)

A decoded signal reconstruction device for inputting a coded signal in a coding scheme involving orthogonal transformation and quantization and reconstructing a decoded signal obtained by performing inverse quantization and inverse orthogonal transformation,
Saturation processing means for performing saturation processing when the signal after the inverse orthogonal transform includes a signal outside a predetermined range;
Re-orthogonal transform means for generating a re-orthogonal transform coefficient obtained by performing orthogonal transform processing on the signal subjected to saturation processing;
The pre-orthogonal transform coefficient obtained when the coded signal is input and dequantized is compared with the re-orthogonal transform coefficient, and the absolute value of these difference values is orthogonal to the quantization range of the orthogonal transform coefficient. Orthogonal transform coefficient comparison means for selecting a re-orthogonal transform coefficient value if it is smaller than a value multiplied by a weighting coefficient that changes corresponding to the energy distribution of the transform coefficient, and for selecting and sending a pre-orthogonal transform coefficient otherwise. When,
An inverse orthogonal transform unit that performs an inverse orthogonal transform on the orthogonal transform coefficient transmitted from the orthogonal transform coefficient comparison unit to generate a decoded signal;
A decoded signal reconstructing device comprising: The output of the inverse orthogonal transform unit, and re-input to the saturation processing section, characterized by being configured to repeat a predetermined number of times, the decoded signal reconstruction apparatus according to claim 1.   A decoding device comprising the decoded signal reconstruction device according to claim 1. In a computer configured as a decoded signal reconstruction device that inputs an encoded signal in an encoding scheme involving orthogonal transformation and quantization, and reconstructs a decoded signal obtained by performing inverse quantization and inverse orthogonal transformation,
A step of performing saturation processing when the signal after inverse orthogonal transformation includes a signal outside a predetermined range;
Performing orthogonal transform processing on the signal subjected to saturation processing and sending a re-orthogonal transform coefficient;
The pre-orthogonal transform coefficient obtained when the coded signal is input and dequantized is compared with the re-orthogonal transform coefficient, and the absolute value of these difference values is orthogonal to the quantization range of the orthogonal transform coefficient. Selecting a re-orthogonal transform coefficient value if smaller than a value multiplied by a weighting coefficient that changes corresponding to the energy distribution of the transform coefficient, and selecting and sending a pre-orthogonal transform coefficient otherwise;
Performing an inverse orthogonal transform on the orthogonal transform coefficient sent from the orthogonal transform coefficient comparing means to generate a decoded signal;
A decoded signal reconstruction program for executing