JP2007336468A - Re-encoding apparatus, re-encoding method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a re-encoding apparatus capable of generating re-encoded data with which image quality deterioration is appropriately suppressed in a decoding device. <P>SOLUTION: The re-encoding apparatus generates encoded data to be inputted to the decoding device including a filter means and comprises: a decoding means for generating decoded data by decoding inputted encoded data and for detecting information about the encoded data; a filter parameter generating means for generating a filter parameter for controlling the filter means of the decoding device based on the information detected by the decoding means; and an encoding means for generating re-encoded data by encoding the decoded data generated by the decoding means, and outputting the re-encoded data and the filter parameter generated by the filter parameter generating means. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a re-encoding device, a re-encoding method, and a program, and more particularly, to a re-encoding device, a re-encoding method, and a program for re-encoding after decoding encoded digital video data.

  In recent years, the development of digital video coding technology has been remarkable, such as television broadcasting represented by BS digital broadcasting and terrestrial digital broadcasting, and recording devices headed by HDD video recorders. Applications are thriving. In particular, the HDD video recorder not only has an advantage in image quality as compared with a conventional analog video recording apparatus such as VHS, but also has a long time spanning tens of hours to hundreds of hours in the built-in HDD. Since moving images can be saved, the convenience of the user is greatly improved. In addition, some HDD recorders can recode the already recorded video at a higher compression rate to expand the remaining capacity of the HDD and realize longer recording. In the re-encoding, a technique for suppressing deterioration in image quality as much as possible while increasing the compression rate is important.

Hereinafter, a conventional technique for suppressing image quality deterioration in re-encoding will be described. Hereinafter, “digital moving image” is simply referred to as “moving image”.
Conventionally, for example, in the re-encoding device of Patent Document 1, image quality deterioration due to re-encoding is suppressed. FIG. 14 is a schematic block diagram showing a codec system including the re-encoding device of Patent Document 1. This codec system includes a re-encoding device 1300 and a decoding device 1310. Further, the re-encoding device 1300 includes a decoding unit 1301, a filter unit 1302, and an encoding unit 1303, and the decoding device 1310 includes a decoding unit 1311 and a filter unit 1312.

  A decoding unit 1301 decodes the encoded data and outputs image data. The output image data is filtered by the filter unit 1302 and then input to the encoding unit 1303. The encoding unit 1303 encodes the input image data. The filtering effect of the filter unit 1302 is one of the effects of suppressing the deterioration of the image quality in the decoded image of the input encoded data, and the other is the effect described below.

  As illustrated in FIG. 15, the decoding unit 1301 includes an inverse quantization unit 1401 that inversely quantizes encoded data to obtain a coefficient Fd, and an IDCT unit 1402 that performs inverse discrete cosine transform on the coefficient Fd. The encoding unit 1303 includes a DCT unit 1404 that obtains a coefficient Fe by performing discrete cosine transform on an image, and a quantization unit 1405 that quantizes the coefficient Fe.

  Here, if the filter unit 1312 does not exist, an image obtained by the inverse discrete cosine transform of the coefficient Fd by the IDCT unit 1402 of the decoding unit 1301 is directly input to the DCT unit 1404 of the encoding unit 1303 and is input to the discrete cosine. Conversion is performed to obtain a coefficient Fe. Since the inverse discrete cosine transform is an inverse transform of the discrete cosine transform if the calculation error is ignored, the coefficient Fd before the inverse discrete cosine transform and the coefficient Fe after the DCT are almost the same value (Fe≈Fd). Since the coefficient Fd input to the IDCT unit 1402 is inversely quantized, it is a non-continuous value that can take only an integer determined by a quantization step in inverse quantization.

  Since Fe≈Fd, the coefficient Fe can take only an integer determined by the quantization step. When the coefficient Fe which is such a discontinuous value is converted into a discontinuous value determined by another quantization step by the quantization unit 1405 of the encoding unit 1303, the magnitude of the DCT coefficient before conversion and Depending on the size of the quantization step, the error becomes larger than a normal quantization error considered from the quantization step at the time of quantization (a quantization error peculiar to re-encoding occurs).

  In order to suppress image quality degradation due to this error, the conventional re-encoding device 1300 includes a filter unit 1302 between the IDCT unit 1402 and the DCT unit 1404, and the image Ie before being input to the DCT unit 1404 is IDCT. It is different from the subsequent image Id. This process prevents a situation in which the coefficient Fd before being input to the IDCT unit 1402 and the coefficient Fe of the output of the DCT unit 1404 are almost the same. As a result, even if the coefficient Fd before being input to the IDCT unit 1402 has a discontinuous value, the coefficient Fe output from the DCT unit 1404 is not necessarily a discontinuous value that is almost the same as the coefficient Fd. In addition, it is possible to prevent a quantization error peculiar to re-encoding due to re-quantization of discontinuous values. Therefore, when the technique described in Patent Document 1 is used, it is possible to suppress degradation of image quality due to quantization error peculiar to re-encoding to some extent.

  By the way, in order for the user to finally view the moving image data encoded by the re-encoding device 1300, it is necessary to decode and reproduce it with a certain decoding device. In the codec system shown in FIG. 14, the decoding device 1310 corresponds to this decoding device. In addition to the decoding unit 1311 that decodes the re-encoded data, the decoding device 1310 includes a filter unit 1312 for suppressing deterioration in image quality that has occurred during encoding.

  The re-encoding device 1300 in Patent Document 2 suppresses deterioration in image quality by determining the strength of a filter applied at the time of decoding (filter applied by the filter unit 1312) at the time of encoding. The re-encoding device 1300 will be described with reference to FIG. 14 again.

  In the technique described in Patent Document 2, a filter parameter is determined as a parameter for determining the strength of the filter in the re-encoding device. In the decoding device, the filter applied at the time of decoding has a strength corresponding to the given filter parameter. Apply a filter. In the description in this specification, it is assumed that a stronger filter is applied as the filter parameter is larger.

  Also, the filter parameter is determined by the encoding unit 1303. The encoding unit 1303 encodes the image data, encodes the determined filter parameter, and outputs it to the outside. The output image and filter parameter re-encoded data are transmitted to the decoding unit 1311. The decoding unit 1311 decodes the re-encoded data and obtains image data and filter parameters. The obtained image data and filter parameters are transmitted to the filter unit 1312. The filter unit 1312 performs filtering according to the filter parameter transmitted from the decoding unit 1311. As filtering, for example, if the filter parameter has two values of 1 (meaning weak filter) and 2 (meaning strong filter), the decoded filter parameter is 1 (weak filter). , Filtering by the (1 2 1) // 4 FIR filter, and filtering by the FIR filter of (1 2 3 4 3 2 1) // 16 when the decoded filter parameter is 2 (strong filter) I do.

  Next, how the filter parameter is determined in the encoding unit 1303 will be described. In the technique described in Patent Document 2, the filter parameter is determined by using SNR (Signal to Noise Ratio).

  The SNR is one index indicating the quality of the image after the filter is applied, and the larger the value under this index, the better the filter. There are two methods for determining the filter parameters of Patent Document 2. FIG. 16 is a diagram for explaining a method of determining the first filter parameter. FIG. 16 shows the relationship between the size of the filter parameter and the SNR (the SNR of the post-filtered image and the image before encoding). The horizontal axis of the graph of FIG. 16 is the filter parameter, and the vertical axis is the SNR. In the method of Patent Document 2, first, an image obtained by filtering a filter parameter with various values using the filter parameter is generated. Next, the SNR between the filtered image and the image before encoding is obtained. The filter parameter that maximizes the finally obtained SNR is selected. It can be seen from FIG. 16 that the SNR can be maximized at a specific filter parameter.

  FIG. 17 is a diagram for explaining a method of determining the second filter parameter. FIG. 17 shows the relationship between the SNR and the optimum filter parameter at that SNR. In the second method, a table indicating the relationship between the SNR and the optimum filter parameter in the SNR is obtained in advance by various images and coding parameters. After the table is obtained, an image before encoding and a local decoded image generated at the time of encoding (an image generated by a virtual decoding device in the encoding device. The filter parameter is obtained from the SNR by referring to the table shown in FIG. As shown in FIG. 17, the relationship between the SNR and the optimum filter parameter is that the strength of the optimum filter increases as the SNR is smaller (the image quality is lower).

  The reason for this is as follows. First, when an image is encoded, a quantization error occurs. This error is random. In this case, smoothing with a filter approaches the original image even if random errors are suppressed on average. Since a coarsely quantized image with a low SNR has a larger error, it is necessary to increase the filter strength in order to reduce the error. Further, not only an index for obtaining a high SNR but also an index for achieving a visually good image quality, it is preferable to apply a stronger filter as the SNR of the image before filtering is lower. This is because noise such as block noise and mosquito noise increases as the SNR decreases, so that the filter that suppresses visually noticeable noise must be strengthened.

  Although the method for determining the filter parameter based on the SNR has been described above, as a conventional technique, a technique for determining the filter parameter using the quantization parameter is known in addition to the method using the SNR. One advantage of using the quantization parameter is that the amount of computation can be reduced because it is not necessary to perform processing for obtaining an SNR having a large amount of computation. Another advantage is that, since the quantization parameter is information obtained when decoding, when using the quantization parameter as the filter parameter, it is not necessary to transmit independent information as a filter parameter to the decoding device. Is a point. Hereinafter, a method of using the quantization parameter for determining the filter parameter will be described. The larger the quantization parameter, the coarser the quantization. Therefore, the larger the quantization parameter, the lower the image quality (SNR). As a result, FIG. 18 showing the same relationship as FIG. 17 is obtained. FIG. 18 is a graph showing a table showing the relationship between the quantization parameter and the optimum filter parameter in the quantization parameter. The decoding unit 1311 can determine optimum filter parameters by drawing a table as shown in FIG.

  FIG. 19 is a block diagram illustrating a configuration of a conventional codec system in a case where the re-encoding device includes a loop filter. This codec system includes a re-encoding device 1500 and a decoding device 1510. Further, the re-encoding device 1500 includes a decoding unit 1501, a filter unit 1502, and an encoding unit 1503, and the decoding device 1510 includes a decoding unit 1511. In the configuration of the codec system, the encoding unit 1503 and the decoding unit 1511 include a loop filter unit 1504 and a loop filter unit 1512 therein. The loop filter is a filter used to generate a reference image when performing inter-screen prediction.

  By the way, in the encoding method using inter-screen prediction, the re-encoding device 1500 and the decoding device 1510 must use the same predicted image. The predicted image is an image generated by motion compensation of the reference image. However, in a codec system including a loop filter, the reference image is also generated by the loop filter. In order to use the same prediction image in 1510, it is necessary to apply the same filter in the loop filter unit 1504 of the re-encoding device 1500 and the loop filter unit 1512 of the decoding device 1510 after all. That is, it is necessary to use the same filter parameter in the re-encoding device 1500 and the decoding device 1510. In the technique of Non-Patent Document 1, the filter parameter determined by the re-encoding device 1500 is superimposed on the re-encoded data and transmitted from the re-encoding device 1500 to the decoding device 1510. The decoding device 1510 can use the same filter parameters as the re-encoding device 1500 by decoding and using the encoded filter parameters. In a re-encoding system including such a loop filter, a quantization error caused by re-encoding is suppressed by the loop filter.

As described above, in the technique described in Patent Document 2, depending on the SNR between the post-filtered image and the image before encoding, or the SNR between the local decoded image and the image before encoding, The filter parameter of the filter to be applied at the time of decoding is determined. The technique described in Patent Document 2 can also suppress deterioration in image quality obtained by decoding by performing filtering at the time of decoding according to the determined filter parameter.
JP 11-41593 A Japanese Patent Laid-Open No. 9-215009 "Information technology -Coding of audio-visual objects- Part10: Advanced Video Coding", ISO Standards, ISO / IEC14496-10: 2005, 2005/12/12

The technique described in Patent Literature 1 includes a decoding device and an encoding device, and in a re-encoding device that re-encodes an image obtained by performing decoding again, the decoded image data Filtering is performed to suppress degradation in image quality caused by encoding at the stage of an image serving as an input for encoding, and further suppress degradation in image quality due to quantization errors peculiar to re-encoding. Is.
However, in the technique described in Patent Document 1, when decoding the re-encoded data output from the re-encoding device, the strength of the post filter required for decoding is not considered. However, the strength of the post filter may become inappropriate. Even when the encoding unit of the re-encoding device includes a loop filter, the strength of the loop filter is not considered, and the strength of the loop filter may be inappropriate. There was a problem that there was.

The technique described in Patent Document 2 is a technique for determining a filter parameter of a post filter applied when decoding moving image data encoded by the encoding apparatus in the encoding apparatus. The filter parameter is determined according to how much the input image to the encoding device is degraded by the conversion.
However, in the technique described in Patent Document 2, the image data input to the encoding unit of the re-encoding device has already deteriorated in image quality due to the original encoding. However, this was not taken into account when determining the filter parameters. Therefore, the strength of the filter is not appropriate in the post filter applied when decoding the re-encoded image data, and the image quality deterioration of the re-encoded image may not be sufficiently suppressed. There was a problem. In addition, since the SNR is used to determine the filter parameter, there is a problem that the amount of calculation is very large.
In addition, in the conventional method in which re-encoding is performed using an encoding method in which the strength of the filter required for decoding is determined by the quantization parameter, the quantization parameter at the time of encoding is reduced. For this reason, there has been a problem that the strength of the filter applied when decoding is inappropriately weakened and image quality degradation caused by re-encoding may not be appropriately suppressed.

  The present invention has been made in view of such circumstances, and an object thereof is to provide a re-encoding device that can generate re-encoded data in which image quality deterioration is appropriately suppressed in the decoding device with a small amount of computation. There is.

  The present invention has been made to solve the above-described problems, and the re-encoding device of the present invention is input in the re-encoding device that generates re-encoded data to be input to the decoding device having the filter means. A decoding unit that decodes the encoded data to generate decoded data, and that detects information related to the encoded data; and a filter that generates a filter parameter for controlling the filter unit based on the information detected by the decoding unit A parameter generating unit and a code that encodes the decoded data generated by the decoding unit to generate re-encoded data, and outputs the re-encoded data with the filter parameter generated by the filter parameter generating unit superimposed thereon And a means for generating.

  As a result, the re-encoding device of the present invention allows the filter parameter generation unit to select a filter parameter that enhances the noise removal effect of the filter unit as the information detected by the decoding unit indicates that the image deterioration is greater. By generating, when decoding is performed by a decoding device including a filter that operates according to the filter parameter, it is possible to generate re-encoded data in which a filter having an appropriate strength is applied and image quality deterioration is appropriately suppressed.

  The re-encoding device of the present invention is the above-described re-encoding device, wherein the information detected by the decoding means is a quantization step.

  As a result, the re-encoding device of the present invention generates a filter parameter such that the noise removal effect of the filter unit becomes stronger as the quantization step detected by the decoding unit increases. When decoding is performed by a decoding device including a filter that operates in accordance with the above, re-encoded data that is filtered with an appropriate strength and appropriately suppresses deterioration in image quality can be generated with a small amount of computation.

  The re-encoding device of the present invention is the above-described re-encoding device, wherein the information detected by the decoding means is a bit rate.

  Thus, the re-encoding device of the present invention generates a filter parameter that causes the noise removal effect of the filter unit to increase as the bit rate detected by the decoding unit decreases, so that the filter parameter generation unit generates the filter parameter according to the filter parameter. When decoding is performed by a decoding apparatus including an operating filter, re-encoded data that is filtered with an appropriate strength and appropriately suppresses deterioration in image quality can be generated with a small amount of calculation.

  The re-encoding device of the present invention is the above-described re-encoding device, wherein the information detected by the decoding means is a frame rate.

  As a result, the re-encoding device of the present invention generates a filter parameter that causes the noise removal effect of the filter unit to increase as the frame rate detected by the decoding unit increases. When decoding is performed by a decoding apparatus including an operating filter, re-encoded data that is filtered with an appropriate strength and appropriately suppresses deterioration in image quality can be generated with a small amount of calculation.

  The re-encoding device of the present invention is the above-described re-encoding device, wherein the information detected by the decoding means is a bit per pixel.

  As a result, the re-encoding device of the present invention generates a filter parameter such that the noise removal effect of the filter unit becomes stronger as the bit per pixel detected by the decoding unit becomes smaller. When decoding is performed by a decoding device including a filter that operates in accordance with the above, re-encoded data that is filtered with an appropriate strength and appropriately suppresses deterioration in image quality can be generated with a small amount of computation.

  The re-encoding device of the present invention is the above-described re-encoding device, wherein the information detected by the decoding means is a picture type.

  As a result, the re-encoding device of the present invention has a noise removal effect of the filter means when the picture type detected by the decoding means is an intra-picture coded picture compared to when the picture type is an inter-picture coded picture. Filter parameter generation means generates a filter parameter that enhances the image quality, so that when decoding is performed by a decoding device including a filter that operates according to the filter parameter, a filter with an appropriate strength is applied and image quality deterioration is appropriately suppressed. Re-encoded data can be generated with a small amount of computation.

  The re-encoding apparatus of the present invention is the above-described re-encoding apparatus, wherein the information detected by the decoding unit is a block type.

  As a result, the re-encoding device of the present invention has a stronger noise removal effect of the filter means when the block type detected by the decoding means is an intra-picture coded block than when it is an inter-picture coded block. The filter parameter generation means generates such a filter parameter, so that when decoding is performed by a decoding device including a filter that operates according to the filter parameter, re-encoding is performed in which a filter with an appropriate strength is applied and image quality deterioration is appropriately suppressed. Data can be generated with a small amount of computation.

  The re-encoding device of the present invention is the above-described re-encoding device, wherein the information detected by the decoding unit is a quantization step, and the filter parameter generation unit is configured to detect the quantum detected by the decoding unit. A filter parameter for controlling the filter unit is generated based on the encoding step and a quantization step when the encoding unit encodes the decoded data.

  Thereby, in the re-encoding device of the present invention, the smaller the difference between the quantization step detected by the decoding means and the quantization step when the encoding means performs encoding, the stronger the noise removal effect of the filter means. Such filter parameters are generated by the filter parameter generation means. For this reason, when the re-encoding device of the present invention performs decoding by a decoding device including a filter that operates according to the filter parameter, image quality degradation that occurs because the difference between the two quantization steps is small, has an appropriate strength. Re-encoded data that is appropriately suppressed can be generated by applying the filter.

  The re-encoding device of the present invention is the above-described re-encoding device, wherein the filter parameter generating unit encodes the quantization step detected by the decoding unit, and the encoding unit encodes the decoded data. A filter parameter for controlling the filter means is generated in accordance with a difference from the quantization step at the time of performing.

  Further, the re-encoding device of the present invention decodes input encoded data to generate decoded data, and a decoding unit that detects a quantization step of the encoded data, and the decoding unit detects Quantization step setting means for detecting the minimum value of the quantization step, and encoding for encoding the decoded data generated by the decoding means in the quantization step having a value larger than the minimum value detected by the quantization step setting means Means.

  As a result, the re-encoding device of the present invention performs re-encoding in which image quality deterioration is appropriately suppressed by applying a filter with an appropriate strength when decoding is performed by a decoding device that determines the strength of the filter based on the quantization step. Data can be generated with a small amount of calculation.

  The re-encoding method of the present invention is a re-encoding method in a re-encoding device that generates re-encoded data to be input to a decoding device having a filter means. A first step of detecting decoded information and detecting information relating to the encoded data, and a filter for controlling the filter means based on the information detected by the decoding means by the re-encoding device A second step of generating parameters, and the re-encoding device encodes the decoded data generated in the first step to generate re-encoded data, and the re-encoded data includes the second And a third step of superimposing and outputting the filter parameters generated in the step.

  The re-encoding method of the present invention is the re-encoding method in the re-encoding device, wherein the re-encoding device decodes the input encoded data to generate decoded data, and the encoded data A first process for detecting the quantization step of the second, a second process for the re-encoding device to detect a minimum value of the quantization step detected in the first process, and a re-encoding device for And a third step of encoding the decoded data generated in the first step in a quantization step having a value larger than the minimum value detected in the second step.

  The program of the present invention is a program for causing a computer to function as the re-encoding device described above.

  According to the present invention, the filter parameter generation unit generates the filter parameter such that the noise detection effect of the filter unit becomes stronger as the information detected by the decoding unit indicates that the image deterioration becomes larger. It is possible to provide a re-encoding device that generates re-encoded data in which a filter having an appropriate strength is applied and image quality deterioration is appropriately suppressed when decoding is performed by a decoding device including a filter that operates according to a parameter.

[First Embodiment]
The first embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a schematic block diagram showing a configuration of a moving image codec system using the re-encoding device 100 according to the first embodiment of the present invention. This codec system includes a re-encoding device 100 and a decoding device 110. The re-encoding device 100 includes a decoding unit 101, a filter unit 102, an encoding unit 103, and a filter parameter generation unit 104. The decoding device 110 includes a decoding unit 111 and a filter unit 112.

  The decoding unit 101 decodes the input encoded data. Image data obtained by decoding is transmitted to the filter unit 102.

  The filter unit 102 filters image data and suppresses deterioration in image quality caused by encoding. The filtered image data is transmitted to the encoding unit 103.

  The decoding unit 101 also obtains information obtained when decoding the encoded data, for example, a quantization step, a quantization parameter, a bit rate, a frame rate, bpp (bit per pixel), a picture type (an intra-picture encoded picture, a screen Information such as an inter-coded picture is transmitted to the filter parameter generation unit 104. Note that bpp (bit per pixel) is the number of bits of encoded data per unit pixel, and is obtained by the calculation of equation (1).

  The filter parameter generation unit 104 determines the filter parameter FP based on the information transmitted from the decoding unit 101 (a method for determining the filter parameter FP will be described later). The determined filter parameter FP is transmitted to the encoding unit 103.

  The encoding unit 103 outputs re-encoded data obtained by encoding the image transmitted from the decoding unit 101. Further, the encoding unit 103 encodes the filter parameter FP transmitted from the filter parameter generation unit 104, and superimposes the filter parameter FP on the re-encoded data and outputs the result.

  The re-encoded data output from the re-encoding device 100 is transmitted to the decoding device 110. In the decoding device 110, first, the re-encoded data is decoded by the decoding unit 111 to obtain image data and a filter parameter FP. The obtained image data and filter parameter FP are transmitted to the filter unit 112.

  The filter unit 112 filters the image data transmitted from the decoding unit 111 using the filter parameter FP transmitted from the decoding unit 111. The output of the decoding device 110 is image data in which deterioration of image quality is suppressed by filtering.

  In this embodiment, as a moving image encoding technique, each image constituting a moving image is converted into a block of a predetermined size (a block of M pixels × N pixels, M and N are integers of 1 or more, for example, 4 8), a discrete cosine transform (DCT) is performed on each block unit, and compression is performed by quantizing the obtained DCT coefficients. Such techniques include, for example, ISO / IEC 14496-2 and ISO / IEC 14496-10.

FIG. 2 is a schematic block diagram illustrating a configuration of a function for encoding an input moving image in the encoding unit 103. The encoding unit 103 includes a DCT unit 1031 and a quantization unit 1032.
The DCT unit 1031 divides each image constituting the input moving image into blocks of a predetermined size, performs discrete cosine transform for each block unit, and obtains DCT coefficients.
The quantization performed by the quantization unit 1032 is expressed by Expression (2).

Here, F is a value before quantization (here, the DCT coefficient generated by the DCT unit 1031), and is hereinafter referred to as an original signal. QF is a value after quantization and is called a quantization coefficient. QSTEP is a quantization step. λ is a value that determines quantization characteristics such as rounding down, rounding up, and rounding off.
The quantization unit 1032 removes the decimal point and converts it to an integer by a floor operation. An error occurs when converting to an integer.

FIG. 3 is a schematic block diagram showing a configuration of a function of decoding encoded data encoded using the above encoding technique in the decoding unit 101. The decoding unit 101 includes an inverse quantization unit 1011 and an IDCT (Inverse Discrete Cosine Transform) unit 1012.
Inverse quantization for the quantization coefficient QF performed by the inverse quantization unit 1011 is expressed as in Expression (3).
F ′ = (QF + μ) · QSTEP (3)
Here, F ′ is a value obtained by inverse quantization. Since F ′ is a signal for reproducing the original signal (F≈F ′), it is hereinafter referred to as a reproduction signal. μ is a value that determines the characteristics of inverse quantization. QSTEP is a quantization step used in quantization. The quantization step QSTEP is transmitted to the inverse quantization unit 1011 by embedding a quantization parameter QP that is information on the quantization step QSTEP in the encoded data when encoding a moving image. . The inverse quantization unit 1011 extracts a quantization parameter QP from the encoded data, and obtains a quantization step QSTEP from the quantization parameter QP.

  Since the inverse quantization unit 1011 performs inverse quantization by performing transformation of Equation (3) on the integerized quantization coefficient, the range of values that can be taken by the inversely quantized reproduction signal F ′ is It is determined by the conversion step QSTEP. The range of values that the reproduction signal F 'can take is not continuous such that continuous integers can be taken as values, but is non-continuous where only predetermined integers can be taken as values. For example, when μ is 0, the reproduction signal F ′ can take only a multiple of the quantization step QSTEP.

  The error between the reproduced signal F ′ and the original signal F is called a quantization error. The quantization error increases as the quantization step QSTEP increases. On the other hand, when the quantization step QSTEP is increased, the range that the quantization coefficient can take is reduced, so that the code amount of the moving image data obtained by encoding the quantization coefficient is reduced.

  In the technique of performing compression by such quantization, when the compression rate is increased (the code amount is decreased), the quantization step QSTEP is increased. When the quantization step QSTEP is increased, the reproduction signal F ′ of each DCT coefficient is greatly separated from the original signal F. Therefore, when IDCT is performed, a large error occurs in the reproduction pixel value of the entire block. Since the way in which this error occurs differs between blocks, as a result, an edge (step) that does not exist in the original image occurs between adjacent blocks. Such edge distortion is visually noticeable and is called block noise. Block noise becomes more noticeable visually as the inside of the block is flatter. In addition to block noise, noise generated by encoding includes noise called ringing noise. Ringing noise is caused by the fact that strong edges such as the boundary of an object cannot be sufficiently expressed as a result of coarse quantization of high-frequency components, and is visually perceived as a wavy pattern near the edges. Ringing noise is also called mosquito noise because it looks like a pattern of mosquitoes flying around.

  The filter unit 102 filters the image after the decoding unit 101 decodes the encoded data in order to suppress such image quality deterioration. Such a filter is called a post filter because it is a filter that is performed as post processing after main processing (decoding). In particular, in the case of a filter for suppressing block noise, it is also called a deblocking filter.

  Next, the operation of the filter unit 102 will be described. In general, there is a property that it is better to increase the strength of the filter as the deterioration of the image quality is larger, and the deterioration of the image quality itself depends on the quantization step. Therefore, in this embodiment, the strength of the filter is changed according to the magnitude of the quantization step. Here, a strong filter means that the filter has a strong noise removal effect. That is, it indicates that the amount of change in the image data by the filter is larger, or that the condition for determining noise in the filter is looser.

The filter unit 102 first obtains a filter parameter TH1 from the quantization step QSTEP1 transmitted from the decoding unit 101 by the following equation.
TH1 = QSTEP1 (4)
The filter parameter TH1 means the strength of the filter and is one of the filter parameters. This is used as a threshold value TH for determining whether or not to apply a filter in the filter method described below. Subsequently, when the filter unit 102 determines to apply a filter, the filter unit 102 filters the block boundary.

A filter method (referred to as filter A) common to the filter unit 102 and the filter unit 112 of this embodiment will be described with reference to FIG. FIG. 4 shows two blocks each having 8 pixels × 8 pixels. The center line indicates the boundary (vertical boundary) between the two blocks. In the figure, p1, p2, p3, p4, p5, p6, p7, and p8 indicate pixel values (integer values) of pixels in a direction perpendicular to the block boundary. The pixel values of the boundary pixels are p4 and p5. In the filter, a (1 2 1) // 4 FIR filter is applied in a direction perpendicular to the block boundary.
When | p4-p5 | <TH, the following filter is applied.
p4 ′ = (p3 + 2 · p4 + p5 + 2) / 4 (5)
p5 ′ = (p4 + 2 · p5 + p6 + 2) / 4 (6)
Here, A / B represents an operation of truncating the fractional part of the quotient obtained by dividing A by B.
If | p4-p5 | ≧ TH, no filter is applied.
Here, TH is a predetermined threshold value, and | x | is an operation for obtaining an absolute value of x.

Further, instead of this filtering method, another filtering method, for example, a stronger (1 2 3 4 3 2 1) // 16 7-tap FIR filter may be used. In this case, the filtered pixel value is
p4 '= (p1 + 2, p2 + 3, p3 + 4, p4 + 3, p5 + 2, p6 + p7 + 8) / 16 (7)
p5 ′ = (p2 + 2 · p3 + 3 · p4 + 4 · p5 + 3 · p6 + 2 · p7 + p8 + 8) / 16 (8)
It becomes. This filter has a stronger filter strength and a higher effect of reducing block noise. However, when a strong filter is applied, there is a problem that the image after filtering is blurred. Therefore, when reducing block noise by filtering, it is important to set the filter strength appropriately.

  Since an image is divided into rectangular blocks and encoded, there are two image boundaries, a horizontal boundary and a vertical boundary. In order to filter to the horizontal boundary and the vertical boundary, in the filtering, first the horizontal boundary is filtered, and then the result filtered to the horizontal boundary is filtered to the vertical boundary. Shall be applied.

  In the above filter A, the difference (| p4-p5 |) of the pixel value at the boundary is calculated and compared with the threshold value TH. This difference means the size of the edge existing at the block boundary after decoding. ing. There are two types of edges that exist at the block boundary after decoding, such as edges that existed before encoding, such as object boundaries and patterns, and edges that are caused by quantization errors. An edge caused by a quantization error is an unnecessary edge, so that it is preferable to blur it by filtering. However, in the case of an edge such as an object boundary, it is not preferable to blur by filtering because it is an important edge.

  The determination using the threshold value TH in the above-described filter considers this property. If the size of the edge is less than the threshold value TH, the edge is regarded as an edge caused by a quantization error, and the filter is determined. When the edge size is equal to or greater than the threshold value TH, it is regarded as an edge existing in the original image and is not filtered. With the above configuration, it is possible to prevent the image from being unnecessarily blurred by the filter. The filter described above is a deblocking filter that suppresses block noise, but may be a deringing filter that suppresses ringing noise.

  In the present embodiment, the configuration including the filter unit 102 is shown as the best mode. However, the realization of the filter unit 102 requires an enormous amount of computation when implemented in software, and is implemented in hardware. Even when doing so, a large circuit scale is required. In order to reduce the amount of calculation of software and the circuit scale of hardware, a configuration without the filter unit 102 is also possible. Since the characteristic part of the present invention is in the filter parameter generation unit 104, the characteristic of the invention is not lost even if the filter unit 102 does not exist.

  When the encoding method of the encoded data decoded by the decoding unit 101 uses a loop filter, the decoding unit 101 includes a loop filter unit therein. In this case, since the output image of the decoding unit 101 has already been filtered in the loop filter unit, the filter unit 102 is unnecessary.

Next, the operation of the filter unit 112 in the decoding device 110 will be described.
In the filter unit 112, first, the filter parameter TH2 is obtained from the following equation (9) from the filter parameter FP transmitted from the decoding unit 111 in the same manner as the quantization step QSTEP2 transmitted from the decoding unit 111.
TH2 = QSTEP2 + FP (9)
The filter parameter TH2 means the strength of the filter and is one of the filter parameters.

In the present embodiment, the sum of the quantization step QSTEP2 and the filter parameter FP is used as the final filter parameter (here, the filter parameter TH2), but TH2 = FP is used instead of Expression (9), and the filter The parameter FP may be directly used as the final filter parameter used in the filter unit 112.
The filter unit 112 applies the filter in the order of the horizontal boundary and the vertical boundary in accordance with the filter A in the same manner as the filter unit 102. Here, the TH of the filter A is TH = TH2. In this embodiment, the filter unit 102 and the filter unit 112 use the same algorithm according to the filter A, but separate algorithms may be used.

FIG. 5 is a block diagram showing the configuration of the re-encoding system. In the re-encoding system of FIG. 5, the deterioration of the image quality between the input original image and the final output image will be described, and then the concept of the present invention regarding a suitable filter strength, that is, a filter parameter generation unit The generation of the filter parameter FP in 104 and the generation of the filter parameter FP in the filter unit 112 will be described.
In the re-encoding system of FIG. 5, the original image I0 input to the system is converted according to the flow of the encoding unit 301, the decoding unit 302, the re-encoding unit 303, and the decoding unit 304. In this re-encoding system, if the image quality deterioration due to encoding in the encoding unit 301 is larger, that is, the image quality of the image I1 input to the re-encoding unit 303 is deteriorated with respect to the original image I0. There is a feature that the image quality of the final image I2, which is the output of re-encoding, is deteriorated as it is more.

  FIG. 6 is a graph showing this feature. The horizontal axis in FIG. 6 is the quantization step in the encoding unit 301, and the vertical axis is the SNR between the original image I0 and the final output image (= output image of the decoding unit 304) I2 of this re-encoding system. The quantization step in the encoding by the re-encoding unit 303 is assumed to be constant. In this case, as shown in the graph of FIG. 6, even when the quantization step in the encoding in the re-encoding unit 303 is constant, the quantization step in the encoding in the encoding unit 301 becomes large. Accordingly, as the image I1 input to the re-encoding unit 301 becomes deteriorated, the SNR of the final image I2 also decreases. When the image I2 is filtered, the lower the SNR, the greater the block noise and mosquito noise, and the pattern (texture) existing in the original image is also distorted. It is desirable to apply a strong filter.

  The SNR is measured by the following formula (10).

ここで、ＭＳＥは平均二乗誤差を表しており、以下の式（１１）から計算される。

Here, MSE represents a mean square error and is calculated from the following equation (11).

Here, x _{i, j} represents the pixel value in the image sample at the position (i, j) of the image before encoding, and y _{i, j} is in the image sample at the position (i, j) of the filtered image. It represents a pixel value. M and N are the width and height of the image, respectively.

  Based on the above properties, the filter parameter generation unit 104 of the present invention sets the filter strength to be increased by the filter unit 112 at the time of decoding as the deterioration in image quality at the time of input to the encoding unit 103 increases. The deterioration in image quality at the time of input to the encoding unit 103 becomes larger as the quantization parameter obtained by the decoding unit 101 becomes larger. Therefore, in the filter parameter generation unit 104 of this embodiment, the decoding unit 101 performs encoding. A filter parameter FP is set so that the stronger the quantization parameter obtained when decoding data, the stronger the filter.

The filter parameter generation unit 104 uses the quantization step QSTEP1 transmitted from the decoding unit 101 to calculate the filter parameter FP from the following equation.
_{FP = CLIP3 (FP min, FP} max, a · QSTEP1 + b) (12)
Here, a and b are predetermined constants (where a is 0 or more and 1 or less), and CLIP3 (min, max, x) means an operation of clipping x to a range of min or more and max or less. FP _min and FP _max are predetermined values, and mean the minimum and maximum values of the filter parameter FP (where FP _min <FP _max ).

  The filter parameter FP is not limited to a linear function (value obtained by clipping) as described above, but a table indicating a relationship between the quantization step and the filter parameter FP in the quantization step is defined. The filter parameter FP may be determined by drawing a predetermined table.

  When the filter parameter FP is determined as shown in Expression (12), if the quantization step QSTEP1 in the decoding unit 101 is large, the filter parameter FP becomes a large value accordingly.

As described above, the filter parameter FP is converted into the filter parameter TH2 in the filter unit 112 according to the equation (9). The filter parameter TH2 is obtained from the equations (9) and (12).
TH2 = QSTEP2 + FP
_{= QSTEP2 + CLIP3 (FP min,} FP max, a · QSTEP1 + b)
(13)
It becomes.

  Equation (13) indicates that not only the filter parameter TH2 increases as the quantization step QSTEP2 in decoding by the decoding unit 111 increases, but also when the quantization step QSTEP1 in the decoding unit 101 increases. Means to grow.

  By the way, when re-encoding is performed, a quantization error due to re-encoding occurs. Therefore, even when the quantization step QSTEP2 in the re-encoding is made small, the image quality of the output image of the decoding unit 111 is deteriorated compared to the input image of the encoding unit 103 (= the output image of the filter unit 102). Become. The output image of the filter unit 102 is improved in visual image quality with respect to the input image of the filter unit 102 by the filter unit 102, but the effect of improving the SNR with respect to the original image is small. Therefore, except when the quantization error in re-encoding is extremely small (when the quantization step QSTEP2 is very small), the input image of the decoding unit 111 has a higher quality than the output image of the decoding unit 101 for the original image. It becomes deteriorated. Therefore, it is preferable that the filter strength at the filter 112 is stronger than the filter strength at the filter 102.

  The filter parameter generation unit 104 may calculate the filter parameter FP as follows in consideration of the above property. In this case, since the filter parameter generation unit 104 requires the quantization step QSTEP2, the encoding unit 102 transmits the quantization step QSTEP2 to the filter parameter generation unit 104. The filter parameter generation unit 104 sets the filter parameter FP as shown in the following equation (14).

  When the filter unit 112 receives the filter parameter FP set by the equation (14), the filter unit 112 calculates the filter parameter TH2 by the equation (15) obtained from the equations (9) and (14).

In this case, regardless of which one of the quantization step QSTEP2 and the quantization step QSTEP1 is large, the value obtained by adding a times the value of the other quantization step to the value of the larger quantization step, and further adding b is used. It will be.

  The filter parameter TH1 of the filter unit 102 is QSTEP1, but the filter parameter TH2 of the filter unit 112 is greater than or equal to the value of the larger quantization step, so TH2 ≧ TH1 is always satisfied. Thus, even when the quantization step QSTEP2 is smaller than the quantization step QSTEP1, the filter strength of the filter unit 112 is higher than that of the filter unit 102.

  By the way, depending on the quantization / inverse quantization method, especially when the quantization step QSTEP1 used in the first encoding is close to the quantization step QSTEP2 used in the second encoding, the quantization specific to the requantization is performed. There are cases where the error in the conversion tends to increase.

  The filter parameter generation unit 104 may set the filter parameter FP as follows in consideration of this property.

In this case, the closer the quantization step QSTEP1 and the quantization step QSTEP2 are, the larger the filter parameter has a value. In addition, when QSTEP2 satisfies the relationship of QSTEP2> QSTEP1, the larger the quantization step QSTEP1, the closer to the quantization step QSTEP2. Therefore, the larger the quantization step QSTEP1, the stronger the filter is applied. When QSTEP2 = QSTEP1, there is no significant deterioration in image quality, so that no particularly large value is set for the filter parameter FP.

  In the above description, the filter parameter generation unit 104 uses the quantization step QSTEP1 for setting the filter parameter FP. However, other information obtained when the decoding unit 101 decodes the encoded data, For example, a bit rate, a frame rate, bpp (bit per pixel), a picture type, or a block type may be used. The quantization step is often large when the bit rate is low, when the frame rate is high, or when bpp (bit per pixel) is small. Using this relationship, the filter parameter generation unit 104 can set the filter parameter FP in consideration of the degree of deterioration in image quality at the time of input to the encoding unit 103.

  First, the filter parameter generation unit 104 may set the filter parameter FP according to the following equation using the bit rate for setting the filter parameter.

Here, bitrate is a bit rate detected by the decoding unit 101 for the encoded data and transmitted to the filter parameter generation unit 104. According to Expression (17), the filter parameter FP increases as the bit rate decreases, and a strong filter is applied. This is because the smaller the bit rate, the greater the degradation of the image decoded by the decoding unit 101. Therefore, the degradation of the image quality of the image decoded by the decoding unit 111, which is an image obtained by re-encoding the image once decoded, also increases. Become. Therefore, the strength of the filter applied by the filter unit 112 is increased.

Next, a case where the filter parameter generation unit 104 uses a frame rate for setting filter parameters will be described. When the bit rate is the same, the higher the frame rate, the higher the compression rate, and the lower the image quality. Therefore, in this case, the filter parameter generation unit 104 sets the filter parameter FP according to the following equation (18).
FP = CLIP3 (FP _min , FP _max , a · framerate + b) (18)
Here, framerate is a frame rate detected by the decoding unit 101 for the encoded data and transmitted to the filter parameter generation unit 104. According to Expression (18), the larger the frame rate, the larger the filter parameter FP is set.

  Next, a case where the filter parameter generation unit 104 uses bpp (bit per pixel) for setting filter parameters will be described. Since bpp means the average bit amount used per pixel as shown in Expression (1), the image quality deteriorates as bpp is smaller. Therefore, in this case, the filter parameter generation unit 104 sets the filter parameter FP according to the following equation.

Here, bpp is a bit per pixel detected by the decoding unit 101 for the encoded data and transmitted to the filter parameter generation unit 104. According to the equation (21), a larger filter parameter FP is set as bpp is smaller.

  Next, a case where the filter parameter generation unit 104 uses a picture type for setting a filter parameter will be described. In the case of an intra-picture coded picture, it is known that block noise is larger than that of an inter-picture coded picture. On the other hand, in the case of an inter-picture coded picture, block noise does not increase compared to an intra-picture coded picture. Therefore, the filter parameter generation unit 104 may determine the filter parameter FP according to the picture type detected by the decoding unit 101 for the encoded data and transmitted to the filter parameter generation unit 104. For example, if the picture type is intra-picture encoding, the filter parameter generation unit 104 sets a predetermined value for the filter parameter FP, and otherwise sets zero for the filter parameter FP.

  In addition, for a frame whose picture type is an inter-picture coded picture, there is an encoding method that can select a prediction method for performing intra-picture prediction or inter-picture prediction for each block. Which prediction method each block uses is called a block type. When the codec system is configured so that the filter parameter FP can be changed for each block or set of blocks, the filter parameter generation unit 104 detects the encoded data by the decoding unit 101 and generates the filter parameter. If the block type transmitted to unit 104 is an intra-coded block (or the ratio of intra-coded blocks for a set of blocks is greater than or equal to a predetermined value), a predetermined value is set in filter parameter FP; Set parameter FP to zero.

  As described above, according to the method of the present embodiment, the filter parameter generation unit 104 determines a filter parameter according to information obtained when decoding encoded data input to the re-encoding device. As a result, it is possible to suppress image quality degradation in the case of re-encoding as compared with the conventional method in which the filter parameter is determined using only information at the time of encoding. Compared to the method using SNR, the method according to the present embodiment does not require the SNR to be calculated, and thus the amount of calculation can be reduced.

[Second Embodiment]
A second embodiment of the present invention will be described. FIG. 7 is a schematic block diagram showing the configuration of the codec system according to the second embodiment of this invention. The codec system of the present invention includes a re-encoding device 500 and a decoding device 510.
The re-encoding device 500 includes a decoding unit 101, a filter unit 102, an encoding unit 503, and a filter parameter generation unit 504. The decoding device 510 includes a decoding unit 511. The codec system of the second embodiment shown in FIG. 7 is different from the codec system of the first embodiment described with reference to FIG. 1, and the encoding unit 503 and the decoding unit 511 are internally provided with a loop filter unit 505 and a loop. A filter unit 512 is provided.
In the figure, portions corresponding to those in FIG. 1 are denoted by the same reference numerals, and description thereof is omitted.

  When a loop filter is used as the encoding method, the same filter must be applied in the loop filter unit 505 of the encoding unit 503 and the loop filter unit 512 of the decoding unit 511. For this reason, in the present embodiment, a filter parameter for controlling the strength of the filter in the loop filter is transmitted from the re-encoding device 500 to the decoding device 510, and the same filter is applied on the encoding side and the decoding side. .

 The operation of re-encoding apparatus 500 will be described. First, when the encoded data is input, the decoding unit 101 decodes the input encoded data to obtain image data. The decoded image data is filtered by the filter unit 102. The filtered image data is transmitted to the encoding unit 503. Further, the decoding unit 101 generates filter parameters based on information obtained when decoding the encoded data, for example, information such as a quantization step, a quantization parameter, a bit rate, a frame rate, bpp (bit per pixel), and a picture type. The data is transmitted to the unit 504. The filter parameter generation unit 504 determines a filter parameter based on the information transmitted from the decoding unit 101 and transmits it to the encoding unit 503. The encoding unit 503 encodes the image data transmitted from the filter unit 102 and outputs the encoded image data. The encoding unit 503 performs filtering similar to the filter A described in the first embodiment in the loop filter unit 505 provided therein according to the filter parameter transmitted from the filter parameter generation unit 504.

FIG. 8 is a schematic block diagram illustrating a configuration of an encoding function including the loop filter unit 505 in the encoding unit 503.
The encoding unit 503 includes adders 5030 and 5036, a DCT unit 5031, a quantization unit 5032, an entropy encoding unit 5033, an inverse quantization unit 5034, an IDCT unit 5035, a loop filter unit 505, and a motion compensation unit 5037. .

  How the input image data is encoded by the encoding unit 503 will be described separately when the picture type is an intra-picture encoded picture and when the picture type is an inter-picture encoded picture. . The intra-picture encoded picture is encoded without using the image data of the previous and subsequent frames. The inter-picture coded picture is encoded by generating a predicted image based on the image data of the previous and subsequent frames and taking a difference from the predicted image.

  First, a case where the picture type is an intra-picture coded picture will be described. When the picture type is an intra-picture coded picture, the DCT unit 5031 obtains DCT coefficients by performing discrete cosine transform on the input image data. The quantization unit 5032 quantizes the DCT coefficient to obtain a quantization coefficient. Further, the entropy encoding unit 5033 entropy-encodes the quantized coefficient and the filter parameter given from the filter parameter generation unit 504 to the loop filter unit 505 to convert it into re-encoded data.

  On the other hand, the inverse quantization unit 5034 inversely quantizes the quantization coefficient converted by the quantization unit 5032 to obtain a DCT coefficient. The IDCT unit 5035 performs inverse discrete cosine transform on the DCT coefficient to convert it into image data. The loop filter unit 505 suppresses deterioration in image quality of the converted image data before using the converted image data as a reference image. That is, the loop filter unit 505 filters the image data in accordance with the filter parameter specified by the filter parameter generation unit 504. The entropy encoding unit 5033 entropy-encodes the filter parameter, and outputs it as re-encoded data together with the entropy-encoded quantization coefficient. Note that the image filtered by the loop filter unit 505 is called a locally decoded image because it is an image obtained by decoding image data once encoded by the encoding unit 503 (local).

  Next, a case where the picture type is an inter-picture coded picture will be described. When the picture type is an inter-picture coded picture, the input image data is not directly subjected to discrete cosine transform, but the motion compensation unit 5037 generates a predicted image of the input image, and the addition unit 5030 A difference (prediction error) from the predicted image is generated, and the DCT unit 5031 performs discrete cosine transform on the prediction error to obtain a DCT coefficient. Further, the quantization unit 5032 quantizes the DCT coefficient to obtain a quantization coefficient. The entropy encoding unit 5033 entropy-encodes the quantization coefficient and the filter parameter specified by the filter parameter generation unit 504, generates re-encoded data, and outputs it. In this way, the filter parameter is output by being superimposed on the re-encoded data.

On the other hand, the inverse quantization unit 5034 inversely quantizes the quantization coefficient generated by the quantization unit 5032 to obtain a DCT coefficient. The IDCT unit 5035 performs inverse discrete cosine transform on this DCT coefficient. The addition unit 5036 adds the conversion result of the IDCT unit 5035 and the predicted image generated by the motion compensation unit 5037. The loop filter unit 505 filters the addition result of the addition unit 5036 according to the filter parameter specified by the filter parameter generation unit 504, and generates a reference image.
The motion compensation unit 5037 generates a predicted image by performing motion compensation on the reference image that is the output of the loop filter unit 505. That is, the motion compensation unit 5037 performs a motion vector search by comparing the reference image with the input image data, and generates a predicted image by performing motion compensation on the reference image based on the resultant motion vector.

Next, the operation of the decoding device 510 will be described. The decoding device 510 decodes the input re-encoded data. The decoding unit 511 has a loop filter unit 512 therein, and the output image of the decoding device is filtered.
The decoding unit 511 decodes the re-encoded data to obtain image data and filter parameters. The loop filter unit 512 inside the decoding unit 511 performs filtering similar to the filter A described in the first embodiment on the decoded image data according to the decoded filter parameters. The filtered image data is output from the decoding unit 511. The filtered image becomes a reference image used when decoding the inter-picture coded picture. Note that ISO / IEC 14496-10 may be used as an encoding technique using a loop filter.

  FIG. 9 is a schematic block diagram illustrating an internal configuration of the decoding unit 511. The decoding unit 511 includes an entropy encoding / decoding unit 5110, an inverse quantization unit 5111, an IDCT unit 5112, an addition unit 5113, a loop filter unit 512, and a motion compensation unit 5115. The entropy encoding / decoding unit 5110 performs entropy encoding / decoding on the input re-encoded data to obtain quantization coefficients and filter parameters. The entropy encoding / decoding unit 5110 transmits the decoded quantization coefficient to the inverse quantization unit 5111 and transmits the decoded filter parameter to the loop filter unit 512. The inverse quantization unit 5111 inversely quantizes the transmitted quantization coefficient to obtain a DCT coefficient. The IDCT unit 5112 performs inverse discrete cosine transform on this DCT coefficient to obtain difference image data. The adding unit 5113 adds the difference image data and the predicted image generated by the motion compensation unit 5115. Based on the filter parameter transmitted from the entropy encoding / decoding unit 5110, the loop filter unit 512 performs filtering on the addition result to obtain image data of a decoded image. The obtained image data of the decoded image is output to the outside. The motion compensation unit 5115 uses the image data of the decoded image as a reference image. The motion compensation unit 5115 performs motion compensation on the reference image and transmits it to the adding unit 5113 as a predicted image in inter-screen prediction.

  A characteristic part in the present embodiment is that the filter parameter generation unit 504 determines a filter parameter for controlling the filter strength of the loop filter from information obtained when the encoded data transmitted from the decoding unit 501 is decoded. Is a point. The operation of the filter parameter generation unit 504 of the present embodiment may be the same as that of the filter parameter generation unit 104 of the first embodiment.

  As described above, according to the technique of the present embodiment, the strength of the loop filter in encoding is determined according to information obtained when decoding encoded data input to the re-encoding device. Compared with the conventional method in which the strength of the loop filter is set using only the information at the time of encoding, without using the information obtained at the time of decoding, image quality degradation when re-encoding is suppressed can do.

[Third Embodiment]
A third embodiment of the present invention will be described. FIG. 10 is a schematic block diagram showing the configuration of the re-encoding device according to the third embodiment of the present invention. The re-encoding device includes a decoding unit 601, a filter unit 602, an encoding unit 603, and a quantization parameter generation unit 604.

The decoding unit 601 decodes the input moving image encoded data. The decoded image data is transmitted to the filter unit 602. The decoding unit 601 also transmits the quantization parameter obtained when decoding the moving image encoded data to the quantization parameter generation unit 604.
The filter unit 602 filters image data and suppresses deterioration in image quality caused by encoding. The filtered image data is transmitted to the encoding unit 603.
A quantization parameter generation unit (quantization step generation unit) 604 obtains a lower limit of the quantization parameter used in the encoding unit 603.
The encoding unit 603 encodes and outputs the image transmitted from the decoding unit 601 with a quantization parameter larger than the lower limit of the quantization parameter obtained by the quantization parameter generation unit 604.

  Next, the relationship between the quantization parameter and the quantization step will be described. Similar to the first and second embodiments, the quantization parameter is information regarding the quantization step embedded in the encoded data. When decoding, a quantization parameter is obtained from the encoded data, and a quantization step can be obtained from the quantization parameter. FIG. 11 shows the relationship between the quantization parameter and the quantization step. As shown in FIG. 11, there are various relationships between the quantization parameter and the quantization step as shown in FIGS. 11A and 11B. In this embodiment, the quantization parameter There is a relationship that the quantization step increases as the value increases. The quantization parameter takes an integer in a predetermined range.

An operation of determining a quantization parameter in the encoding unit 603 will be described. The encoding unit 603 includes a rate control unit (not shown). The rate control unit determines the quantization parameter QPrate so as to match the bit rate specified from the outside. In addition, the encoding unit 603 receives the quantization parameter QPmin transmitted from the quantization parameter generation unit 604. The encoding unit 603 compares the quantization parameter QPrate of the rate control unit with the quantization parameter QPmin transmitted from the quantization parameter generation unit 604, and selects the larger one. Expressed as an equation, equation (20) is obtained.
QP = MAX (QPrate, QPmin) (20)
Here, MAX (x, y) is an operation for selecting the larger of x and y. The encoding unit 603 performs quantization using the selected quantization parameter QP. The quantization parameter used in the quantization in the encoding unit 603 is QP2, and the quantization step is QSTEP2. Note that the quantization parameter used in inverse quantization in the decoding unit 601 is QP1, and the quantization step is QSTEP1.

  Next, image quality degradation in the re-encoding system will be described again with reference to FIG. In the re-encoding system, the image quality of the image I1> the image quality of the image I2 in the image quality compared with the original image I0. That is, the image quality of the image I2 is inferior to that of the image I1. Considering that it is preferable to apply a stronger filter strength to an image with lower image quality, the filter strength to be applied to the image I2 is stronger than the filter strength to be applied to the image I1.

  By the way, in the technique described in ISO / IEC14496-2 and ISO / IEC14496-10, the strength of the filter applied to the decoded image is determined according to the quantization roughness (quantization step) at the time of encoding. Is obtained from the quantization parameter. That is, if the quantization step at the time of encoding is determined, the strength of the filter is determined. In the case of an apparatus using such a technique for the encoding unit 603, in order to apply a filter having a strength higher than a predetermined level in a decoded image, a value higher than the predetermined level is set as a quantization parameter in re-encoding. There is a need. Based on the above idea, the quantization parameter generation unit 604 of the present embodiment sets a lower limit of the quantization parameter based on the filter strength to be applied.

A method in the present embodiment in which the quantization parameter generation unit 604 sets the lower limit of the quantization parameter will be described.
In order to make the strength of the filter determined according to the quantization step QSTEP2 in the encoding unit 603 equal to or higher than the strength of the filter determined according to the quantization step QSTEP1 in the decoding unit 601, quantization is performed. A quantization step QSTEP2 such that step QSTEP2 ≧ quantization step QSTEP1 may be used in the quantization in the encoding unit 603.

  If the encoding method used for decoding in the decoding unit 601 and the encoding method used for encoding in the encoding unit 603 are the same, the quantization parameter QP2 ≧ decoding in the encoding unit 603 The quantization parameter QP1 in the unit 601 may be used. Therefore, when the encoding method in the decoding unit 601 is the same as the encoding method in the encoding unit, the quantization parameter generation unit 604 sets the quantization parameter QP1 in the decoding unit 601 to the quantization parameter QPmin. To do.

  However, when the encoding method used for decoding by the decoding unit 601 is different from the encoding method used for encoding by the encoding unit 603, the relationship between the quantization parameter and the quantization step is different from each other. Therefore, even if QP2 satisfying QP2 ≧ QP1 is used, QSTEP2 ≧ QSTEP1 is not always satisfied. Therefore, the quantization parameter generation unit 604 needs to obtain a quantization parameter QP2 that satisfies QSTEP2 ≧ QSTEP1.

Here, the operation of the quantization parameter generation unit 604 when the relationship between the quantization parameter and the quantization step in the encoding scheme used in the decoding unit 601 and the encoding unit 603 is given will be described.
First, in the encoding method used in the decoding unit 601, the relationship between the quantization parameter QP1 and the quantization step QSTEP1 is expressed by Expression (21).
QSTEP1 = 2QP1 (21)

FIG. 12 shows the relationship between the quantization parameter and the quantization step in the encoding unit 603 of the present embodiment.
After obtaining QSTEP1 according to equation (21), the quantization parameter generation unit 604 obtains a quantization parameter to be set by the method described below. FIG. 13 is a flowchart showing how to obtain the quantization parameter of the quantization parameter generation unit 604 in the present embodiment. When the relationship between the quantization parameter and the quantization step in the encoding unit 603 is that of FIG. 12, the quantization parameter can be obtained according to the flowchart shown in FIG.

In step S401, the quantization parameter generation unit 604 sets the quantization parameter QP as the minimum value of the quantization parameter that can be set in the encoding unit 603. In the present embodiment, the minimum value of the quantization parameter is 0.
In S402, the quantization parameter generation unit 604 obtains a quantization step QSTEP from the quantization parameter QP. In the case of this embodiment, it is obtained by drawing a table showing the relationship between the quantization parameter and the quantization step shown in FIG.
In step S403, the quantization parameter generation unit 604 determines whether the obtained quantization step QSTEP is equal to or higher than the previously obtained QSTEP1. If it is equal to or greater than QSTEP1, the process ends because the necessary quantization parameter has been obtained. If it is less than QSTEP1, the process proceeds to S404.
In S404, it is determined whether QP is equal to the maximum value of the set quantization step. If it is equal to the maximum value, a larger quantization step cannot be set, and the process ends here. If it is equal to or less than the maximum value, the process proceeds to S405. In S405, 1 is added to the quantization parameter by QP = QP + 1.

  The quantization parameter QP obtained by the operation shown in the flowchart of FIG. 13 is set as a quantization parameter QPmin that is output from the quantization parameter generation unit 604. In the above flowchart, the quantization parameter is obtained by a linear search. However, a binary search may be performed by using the fact that the quantization step is monotonically increasing with respect to the quantization parameter.

  As described above, since the encoding unit 604 performs quantization with a quantization parameter equal to or greater than the quantization parameter QPmin, in the case of filtering according to the size of the quantization parameter when decoding, it is inappropriate. Weak filters are no longer applied.

  As described above, according to the technique of the present embodiment, the quantization parameter in the encoding is determined according to the quantization parameter obtained when decoding the moving image encoded data input to the re-encoding device. Determine the lower limit. Accordingly, it is possible to prevent an improperly weak filter from being applied in the post filter or loop filter during decoding, and it is possible to suppress deterioration in image quality that occurs during re-encoding.

  Also, the decoding unit 101, the filter unit 102, the encoding unit 103, the filter parameter generation unit 104, the decoding unit 111, the filter unit 112 in FIG. 1, and the decoding unit 101, the filter unit 102, the encoding unit 503 in FIG. To realize the functions of the filter parameter generation unit 504, the loop filter unit 505, the decoding unit 511, the loop filter unit 512, and the decoding unit 601, the filter unit 602, the encoding unit 603, and the quantization parameter generation unit 604 in FIG. The above program may be recorded on a computer-readable recording medium, and the program recorded on the recording medium may be read into a computer system and executed to perform the processing of these units. The “computer system” here includes an OS and hardware such as peripheral devices.

Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.
The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system serving as a server or a client in that case is also used to hold a program for a certain period of time. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes design and the like within a scope not departing from the gist of the present invention.

  The present invention is most suitable for use in a re-encoding device that receives encoded digital moving image data such as an HDD recorder, re-encodes and records, or re-distributes, but is not limited thereto. Also, it may be used for a re-encoding device such as a broadcasting facility or a network distribution facility.

It is the schematic block diagram which showed the structure of the codec system using the re-encoding apparatus 100 in 1st Embodiment of this invention. 4 is a schematic block diagram illustrating a configuration of a function of encoding an input moving image in the encoding unit 103 according to the same embodiment. FIG. Outline of configuration of function of decoding moving image encoded data encoded using encoding technique similar to encoding technique used in encoding section 103 in decoding section 101 in the same embodiment It is a block diagram. It is a figure which shows the area | region to which the filter in the same embodiment is applied, showing two blocks of vertical and horizontal 8 pixels × 8 pixels. It is the schematic block diagram which showed the structure of the re-encoding system in the embodiment. It is the graph which showed the relationship between the quantization step in the encoding part 301 of the re-encoding system in the same embodiment, and SNR of the original image I0 and the image I2. It is the schematic block diagram which showed the structure of the codec system using the re-encoding apparatus 500 in 2nd Embodiment of this invention. It is the schematic block diagram which showed the structure of the encoding function containing the loop filter part 505 in the encoding part 503 in the embodiment. It is the schematic block diagram which showed the structure inside the decoding part 511 in the same embodiment. It is the schematic block diagram which showed the structure of the re-encoding apparatus in 3rd Embodiment of this invention. It is the graph which showed the kind of relationship between a quantization parameter and a quantization step. It is the table | surface which showed the relationship between the quantization parameter in the same embodiment, and a quantization step. It is a flowchart which shows the operation | movement at the time of calculating | requiring the quantization parameter of the quantization parameter production | generation part 604 in the embodiment. It is the schematic block diagram which showed the structure of the codec system using the conventional re-encoding apparatus. It is the schematic block diagram which showed the structure of the conventional re-encoding apparatus. It is a graph for demonstrating the 1st method of determining the conventional filter parameter. It is a graph for demonstrating the 2nd method of determining the conventional filter parameter. It is a graph which shows the relationship between a quantization parameter and the optimal filter parameter in the quantization parameter. It is the schematic block diagram which showed the structure of the codec system using the conventional re-encoding apparatus provided with a loop filter.

Explanation of symbols

100: Re-encoding device 101: Decoding unit
102: Filter unit, 103: Encoding unit,
104: Filter parameter generation unit 110: Decoding device
111: Decoding unit 112: Filter unit
301: Encoding unit 302: Decoding unit
303: Re-encoding unit, 304 ... Decoding unit,
500 ... Re-encoding device, 503 ... Encoding unit,
504 ... Filter parameter generation unit, 505 ... Loop filter unit,
510 ... Decoding device, 511 ... Decoding unit,
512 ... Loop filter unit, 601 ... Decoding unit,
602: Filter unit, 603: Encoding unit,
604 ... Quantization parameter generation unit, 1011 ... Inverse quantization unit,
1012 ... IDCT section, 1031 ... DCT section,
1032 ... Quantization unit, 1300 ... Re-encoding device,
1301 ... Decoding unit 1302 ... Filter unit,
1303 ... Encoding unit, 1310 ... Decoding device,
1311 ... Decoding unit, 1312 ... Filter unit,
1401 ... Inverse quantization unit, 1402 ... IDCT unit,
1404 ... DCT section, 1405 ... quantization section,
1500 ... re-encoding device, 1501 ... decoding unit,
1502 ... Filter unit, 1503 ... Encoding unit,
1504 ... Loop filter unit, 1510 ... Decoding device,
1511 ... Decoding unit, 1512 ... Loop filter unit,
5030 ... Adder, 5031 ... DCT,
5032 ... Quantization unit, 5033 ... Entropy encoding unit,
5034 ... Inverse quantization unit, 5035 ... IDCT unit,
5036 ... addition unit, 5037 ... motion compensation unit,
5110 ... Entropy encoding / decoding unit, 5111 ... Inverse quantization unit,
5112 ... IDCT section, 5113 ... Addition section,
5115: Motion compensation unit

Claims (13)

In a re-encoding device that generates re-encoded data to be input to a decoding device having filter means,
Decoding means for decoding the input encoded data to generate decoded data and detecting information relating to the encoded data;
Filter parameter generation means for generating a filter parameter for controlling the filter means based on the information detected by the decoding means;
Encoding means for encoding the decoded data generated by the decoding means to generate re-encoded data, and superimposing the filter parameter generated by the filter parameter generating means on the re-encoded data, A re-encoding device.   The re-encoding apparatus according to claim 1, wherein the information detected by the decoding unit is a quantization step.   The re-encoding apparatus according to claim 1, wherein the information detected by the decoding unit is a bit rate.   The re-encoding apparatus according to claim 1, wherein the information detected by the decoding unit is a frame rate.   The re-encoding apparatus according to claim 1, wherein the information detected by the decoding unit is a bit per pixel.   The re-encoding apparatus according to claim 1, wherein the information detected by the decoding unit is a picture type.   The re-encoding apparatus according to claim 1, wherein the information detected by the decoding unit is a block type. The information detected by the decoding means is a quantization step,
The filter parameter generation unit generates a filter parameter for controlling the filter unit based on a quantization step detected by the decoding unit and a quantization step when the encoding unit encodes the decoded data. The re-encoding device according to claim 1. The filter parameter generation means controls the filter means according to the difference between the quantization step detected by the decoding means and the quantization step when the encoding means encodes the decoded data. The re-encoding device according to claim 8, wherein: Decoding means for decoding the input encoded data to generate decoded data and detecting a quantization step of the encoded data;
Quantization step generation means for detecting a minimum value of the quantization step detected by the decoding means;
A re-encoding device comprising: encoding means for encoding the decoded data generated by the decoding means in a quantization step having a value larger than the minimum value detected by the quantization step generating means. In a re-encoding method in a re-encoding device that generates encoded data to be input to a decoding device having filter means,
A first step in which a re-encoding device decodes input encoded data to generate decoded data, and detects information about the encoded data;
A second step in which the re-encoding device generates a filter parameter for controlling the filter means based on the information detected by the decoding means;
The re-encoding device encodes the decoded data generated in the first process, generates re-encoded data, and superimposes the filter parameter generated in the second process on the re-encoded data. And a third process for outputting the recoding method. In the re-encoding method in the re-encoding device,
A first step in which a re-encoding device decodes input encoded data to generate decoded data and detects a quantization step of the encoded data;
A second process in which the re-encoding device detects a minimum value of the quantization step detected in the first process;
A re-encoding device comprising: a third step of encoding the decoded data generated in the first step in a quantization step having a value larger than the minimum value detected in the second step. A characteristic re-encoding method. The program for functioning a computer as a re-encoding apparatus in any one of Claims 1-10.

Images