JP6792359B2 - Image coding device - Google Patents

Description

The present invention relates to the coding of an image, particularly a technique of applying frequency conversion to an image and encoding the generated coefficients.

Recently, with the development of digital imaging devices such as digital cameras and digital camcorders, various image data compression coding methods are being studied. JPEG (Joint Photographic Experts Group) 2000 has been proposed as one of the compression coding methods. One of the features of JPEG2000 is that it uses wavelet transform (hereinafter referred to as DWT transform) instead of the discrete cosine transform (DCT) used in JPEG. By using the DWT conversion, it is possible to apply frequency conversion without dividing the image into a plurality of blocks even if the image is large in size. As a result, JPEG2000 can eliminate the deterioration of image quality that may occur at the block boundary when the image is divided into a plurality of blocks.

In order to perform the DWT conversion, it is necessary to filter all the images in the vertical direction using the filter for DWT, and then filter in the horizontal direction. Therefore, as in Patent Document 1, a large-scale memory (buffer) is required for filtering.

In response to this point, Patent Document 2 proposes to reduce the memory capacity. However, according to Patent Document 2, there arises a problem that the timings at which a plurality of frequency domains (subbands) obtained by performing DWT conversion are output are substantially the same. That is, the circuit scale of the subsequent quantization unit and coding unit becomes large.

Japanese Patent Application Laid-Open No. 2003-274185 JP 2001-285634

The present invention has been made in view of the above problems, and an object of the present invention is to provide a technique for leveling the quantization and coding processes after the discrete wavelet transform and reducing the circuit scale.

In order to solve this problem, for example, the image coding apparatus of the present invention has the following configuration. That is,
An image coding device that encodes image data.
A conversion means for inputting image data to be encoded, wavelet transforming the input image data, and generating conversion coefficient data for a plurality of subbands.
A coding means that quantizes and encodes the coefficient data of each subband of the plurality of subbands generated by the conversion means, and
It has a buffer means for temporarily holding the subband coefficient data to be supplied to the coding means, which is the conversion coefficient data generated by the conversion means.
Of the plurality of subbands generated by the conversion means, the conversion coefficient data of the preset subbands is supplied to the coding means without going through the buffer means.
The conversion coefficient data of the other sub-bands among the plurality of sub-bands is temporarily held by the buffer means, and the conversion coefficient data held by the buffer means is supplied to the coding means. And.

According to the present invention, it is possible to reduce the circuit scale of the quantization and coding processing after the wavelet transform.

The block diagram which shows the structural example of the image coding apparatus which concerns on 1st Embodiment. The figure for demonstrating the timing of performing a reversible 5-3DWT conversion. The block diagram which shows the structure of the image coding apparatus for explaining a problem. The figure which shows the processing timing of the image coding apparatus of FIG. The figure which shows the processing timing of the image coding apparatus which concerns on 1st Embodiment. The figure which shows the relationship of each subband generated by a wavelet transform. The figure which shows the processing timing at the time of executing the wavelet transform up to decomposition level 3 in the image coding apparatus which concerns on 1st Embodiment. The block diagram which shows the structural example of the image coding apparatus which concerns on 2nd Embodiment. The figure which shows the processing timing of the image coding apparatus which concerns on 2nd Embodiment. The figure for demonstrating the relationship of the number of horizontal DWT coefficients of each subband generated by wavelet transform when the number of pixels in a horizontal direction is an odd number.

First, prior to the description of the embodiment, the wavelet coefficient (hereinafter referred to as DWT coefficient) in the DWT transform using the reversible 5-3 tap filter proposed in JPEG2000 (hereinafter referred to as reversible 5-3DWT transform) is output. The timing will be described in detail with reference to FIGS. 2 to 6. FIG. 2 is a diagram for explaining the timing at which the DWT coefficient is output when the reversible 5-3 DWT conversion is realized by using the lifting structure.

In FIG. 2A, reference numerals a to e indicate input pixel data arranged in the horizontal direction. Further, in FIG. 2A, b'is the DWT coefficient of the high frequency component of the decomposition level 1 generated using the pixel data a, b, c, and d'is the decomposition level 1 generated using the pixel data c, d, e. It is the DWT coefficient of the high frequency component of. c "is the DWT coefficient of the low frequency component generated by using b', d', and c. Each DWT coefficient is as shown in the following equations (1) to (3).
b'= b-(a + c) / 2… (1)
d'= d-(c + e) / 2… (2)
c "= c + (b'+ d'+ 2) / 4… (3)
As shown in equations (1) and (2), both b'and d'are DWT coefficients of high-frequency components, so that the types of equations are the same except for the pixel data used.

FIG. 2B is a diagram showing a state in which pixel data f is newly input in the horizontal direction with respect to FIG. 2A. As shown in FIG. 2B, even if the pixel data f is input, there are not enough input pixels to newly execute the DWT conversion, so a new DWT coefficient is output at the timing when the pixel data f is input. I can't.

On the other hand, as shown in FIG. 2 (c), when pixel data g is further input to FIG. 2 (b), DWT conversion can be executed, and the DWT coefficient f'of the high frequency component and the DWT coefficient of the low frequency component can be executed. e "can be output.

In this way, the timing for executing the DWT conversion in the horizontal direction is once for every two pixel inputs of the pixel data. When performing the DWT conversion hierarchically to the decomposition level 2, it is necessary to use the DWT coefficient of the low frequency component of the decomposition level 1. As described above, the timing at which the DWT coefficient of the low frequency component of the decomposition level 1 is output is once for each pixel data 2 pixel input, so the timing at which the decomposition level 2 DWT coefficient is output is once every 4 pixel input. It becomes. From the above relationship, the timing at which the DWT coefficient of the decomposition level N is output is once for each X pixel using X represented by the following equation (4).
X = 2 ^N … (4)

Since the same relationship is applied to the vertical direction, the timing at which the DWT coefficient in the vertical direction is output is once per pixel data X-line input according to the equation (4).

FIG. 6 shows the relationship between each subband when the input image is subjected to vertical / horizontal DWT transform to decomposition level 2. In FIG. 6, L indicates a low frequency region and H indicates a high frequency region. For example, 1HH indicates a subband in the high frequency region in both the horizontal and vertical directions of decomposition level 1. As shown in FIG. 6, the size of each subband of decomposition level 1 is half of the input image data in the horizontal and vertical directions. That is, the number of coefficients of each subband is 1/4 of the number of pixels of the original image. Further, each subband of decomposition level 2 is half of each subband of decomposition level 1 in the horizontal and vertical directions.

Based on the above description, a series of processing timings for DWT conversion and encoding of image data will be described with reference to FIGS. 3 and 4.

FIG. 3 is a block diagram showing a configuration example of the image coding apparatus considered by the present inventor. As shown in FIG. 3, the image coding apparatus includes a wavelet conversion unit 300 and a quantization / coding unit 303. The wavelet transform unit 30 includes a Lev1-DWT transform unit 301 that performs DWT conversion of decomposition level 1 and a Lev2-DWT transform unit 302 that performs DWT conversion of decomposition level 2 inside. Although JPEG2000 will be described as an example of the coding method, the coding method is not particularly limited.

Further, FIG. 4 shows the conversion timing of the DWT coefficient of each subband of the decomposition levels 1 and 2 when the image data is input to the image encoding device of FIG. 3 in raster order and DWT converted to the decomposition level 2, and the DWT after the conversion. It is a figure which shows the timing when a coefficient is quantized and coded.

The viewpoint of FIG. 4 will be described. In FIG. 4, each section from t0 to t14 on the horizontal axis indicates a period during which pixel data of one line is input. It is assumed that the input of image data is started from the t0 period.

The rectangular bar shown in "Pixel data" indicates that pixel data has been input during that period. In FIG. 4, since there are bars in all periods, it is shown that pixel data for one line is input in all periods.

On the other hand, the bar of the subband of each decomposition level of "DWT transform" indicates whether or not the DWT transform is executed in each period and the data of each subband is output to the quantization / coding unit 303. For example, in the t0 period, there is no bar in the subband of any decomposition level, which indicates that the DWT transform is not executed. Also, since the bars of Lev1-LH, Lev1-HL, and Lev1-HH exist in the t2 period, the decomposition level 1DWT is executed and the coefficients of the level 1 subbands LH, HH, and LH are output. Shown.

The "quantization / coding" bar indicates at what timing the DWT coefficient output by DWT conversion in each period is quantized / coded. For example, in the t0 period, there is no bar in any decomposition level subband, which indicates that the coefficients of all subbands are not quantized and encoded. Moreover, since the bars of Lev1-LH, Lev1-HL, and Lev1-HH are present in the t2 period, it is shown that the subbands HL, LH, and HH of decomposition level 1 are quantized and coded.

As shown in the figure, in this image coding apparatus, the DWT conversion bar and the quantization / coding bar exist in the same period. This indicates that the DWT coefficient output by the DWT transform is immediately quantized and encoded in the same line input period. In addition, the length of the bar indicates the period during which the quantization / coding is executed.

As shown in FIG. 6, the DWT coefficient (Lev1-LH, HL, HH) of each subband of the decomposition level 1 is output only 0.5 line in the period when one line of pixel data is input. However, as explained with reference to FIG. 2, the coefficient input to the quantization / coding unit is also only one for every two input pixels. Therefore, as shown in FIG. 4, the DWT coefficient is quantized over one line input period. It is encoded.

Next, the coding process executed in each section of FIG. 4 will be described with reference to FIG. In the period from t0 to t1, pixel data for two lines is input to the Lev1-DWT transform unit 301. However, at this point, the number of lines sufficient to perform the reversible 5-3DWT conversion has not yet been input. Therefore, the Lev1-DWT conversion unit 301 only holds the input pixel data of the two lines in the line buffer in the Lev1-DWT conversion unit 301, and does not execute the DWT conversion during this period.

In the t2 period, the Lev1-DWT transform unit 301 executes DWT conversion using the pixel data of two lines held in the line buffer in the t0 and t1 periods and the input pixel data of the third line. As a result, coefficient data of subbands LL, LH, HH, and LH of decomposition level 1 are generated. Pixel data that is insufficient to execute the 5-tap filter process is generated by mirroring. Of the generated DWT coefficients, the DWT coefficient of the subband LL of decomposition level 1 is used for the DWT conversion of decomposition level 2, and is therefore transmitted to the Lev2-DWT transform unit 302. The coefficient data of the other subbands, that is, the subbands LH, HH, and LH of the decomposition level 1, are output to the quantization / coding unit 303 in the subsequent stage.

In the t2 period, the DWT coefficient of the subband LL of decomposition level 1 is still output for only one line, and the DWT conversion of decomposition level 2 cannot be executed. Therefore, the Lev2-DWT transform unit 302 does not execute the DWT transform and only holds the coefficient data in the internal line buffer. Further, the number of coefficients in each subband of the decomposition level 1 is half of the original pixel data. Therefore, the DWT coefficients for 1.5 lines (0.5 × 3) are output based on the original pixel data by combining the three subbands LH, HH, and LH of the decomposition level 1. The quantization / coding unit 303 quantizes the DWT coefficients for a total of 1.5 lines sent from the wavelet transform unit 300 using the quantization parameters. Then, in the coding, entropy coding such as EBCOT (Embedded Block Coding with Optimized Truncation) is applied to each subband and output as coded data.

In the t3 period, as in the t0 to t1 periods, data for the number of lines sufficient for applying the reversible 5-3DWT conversion is not input. Therefore, the input pixel data is held in the line buffer in the Lev1-DWT transform unit 301, and the DWT coefficient data is not output during this period.

In the t4 period, DWT conversion is performed using the pixel data for two lines input in the t2 and t3 periods and held in the line buffer and the pixel data for the third line, and the decomposition level 1 subbands LL, LH, and HH are used. , LH coefficient data is calculated. However, even at this time, since the DWT coefficient data of the subband LL of the decomposition level 1 is output for only two lines including the amount held in the line buffer in the t2 period, the DWT conversion of the decomposition level 2 cannot be executed. After that, the DWT transform of the decomposition level 1 is executed once every two lines (that is, in the period of t6, t8, t10, t12, t14 ...) as shown in the above equation 4. The quantization / coding unit 303 quantizes the DWT coefficients for a total of 1.5 lines sent from the wavelet transform unit 300 using the quantization parameters, and applies entropy coding such as EBCOT to each subband to encode the code. Output as quantized data.

During the t5 period, the DWT conversion of decomposition levels 1 and 2 is not executed, and the input data is held in the line buffer in the Lev1-DWT conversion unit 301.

During the t6 period, the Lev1-DWT transform unit 301 executes the DWT transform of the decomposition level 1. In addition, the DWT coefficient data of the decomposition level 1 subband LL output at that time and the two lines of the decomposition level 1 subband LL held in the line buffer in the Lev2-DWT converter 302 during the t2 and t4 periods. The Lev2-DWT transform unit 302 calculates the coefficient data of the subbands LL, LH, HH, and LH of the decomposition level level 2 using the DWT coefficient data of the minute.

As shown in FIG. 4, since the number of coefficients of decomposition level 2 is half the coefficient of decomposition level 1, the four subbands of decomposition level 2 LL, LH, HH, and LH are based on the original pixel data. In total, the DWT coefficient for one line (0.25 × 4) is output to the quantization / coding unit 303 in the subsequent stage. This and the 1.5 lines of the three sub-bands LH, HH, and LH of decomposition level 1 are combined, and the DWT coefficient of 2.5 lines (1 + 1.5) based on the original pixel data is output during this period. To.

The DWT transform of the decomposition level 2 is executed once every four lines (that is, t10, t14 ... Period) as shown in the above equation 4. The quantization / coding unit 303 quantizes the DWT coefficients for a total of 2.5 lines sent from the wavelet transform unit 300 using the quantization parameters, and performs entropy coding such as EBCOT for each subband. Output as encoded data. After that, the same process is repeated until the final line of the image data.

As can be seen from the above explanation, when the DWT transform is executed up to the decomposition level 2, the period during which the DWT coefficient for a maximum of 2.5 lines is output during the period when one input pixel is input (t6, t10 in FIG. 4). , T14) exists. Therefore, if the quantization and coding are to be executed immediately at the timing of executing the DWT conversion, the performance capable of processing the DWT coefficient for 2.5 lines in the period for inputting one line of pixel data is required, and the quantization and coding are required. It is understandable that the circuit scale of the coding unit becomes large. Hereinafter, embodiments according to the present invention will be described based on these points.

[First Embodiment]
FIG. 1 is a block diagram showing a configuration example of an image coding apparatus according to an embodiment of the present invention in the first embodiment. In order to simplify the description, the discrete wavelet transform will be performed twice in this embodiment as well.

As shown in the figure, the image coding apparatus according to the first embodiment includes a wavelet transform unit 300, a subbandline buffer 103, and a quantization / coding unit 104. As for the pixel data of the image data to be encoded, it is assumed that the pixel data is supplied to the wavelet transform unit 300 in the order of raster scan. The wavelet transform unit 300 is composed of a Lev1-DWT transform unit 301 that performs discrete wavelet transform of decomposition level 1 (hereinafter, also referred to as DWT transform in the embodiment) and a Lev2-DWT transform unit 302 that performs DWT transform of decomposition level 2. To. In this embodiment, JPEG2000 will be used as an example of the coding method, but the type of the coding method is not particularly limited. Further, although the wavelet transform unit 300, the Lev1-DWT transform unit 301, and the Lev2-DWT transform unit 302 use the same configuration as the encoding device of FIG. 2 described above, the coefficient data obtained by the DWT transform is output. The destination is different. Further, it is assumed that the quantization / coding unit 104 has a performance capable of quantizing / coding the DWT coefficient for one line during the period in which one line of pixel data is input.

FIG. 5 shows the timing of conversion of image data into DWT coefficient data of each subband of decomposition level 1 and level 2 when the image data is DWT-converted to decomposition level 2 in raster order, and the converted DWT coefficient data is quantized. -It is a timing diagram for explaining the timing to be encoded. Since the view of FIG. 5 is the same as that of FIG. 4, the explanation is omitted.

Hereinafter, the coding process executed in each section of FIG. 5 will be described with reference to FIG.

In the period from t0 to t1, pixel data for two lines is input to the Lev1-DWT transform unit 301. However, at this point, data for the number of lines sufficient for performing the reversible 5-3DWT conversion has not yet been input. Therefore, the Lev1-DWT conversion unit 301 only holds the input pixel data of two lines in the line buffer in the Lev1-DWT conversion unit 301, and does not execute the DWT conversion during this period.

In the t2 period, the Lev1-DWT transform unit 301 executes DWT conversion using the pixel data for two lines held in the line buffer in the t0 and t1 periods and the input pixel data for the third line, and performs DWT conversion, and is a sub of decomposition level 1. Outputs coefficient data for bands LL, LH, HH, and LH. At that time, since the DWT coefficient data of the subband LL of the decomposition level 1 is used for the DWT conversion of the level 2, it is supplied to the Lev2-DWT transform unit 302. Further, the DWT coefficient data of the subband LH and HL of the decomposition level 1 is directly output to the quantization / coding unit 104 without the intervention of the subband line buffer 103, but the DWT of the subband HH of the level 1 is output. The coefficient data is transmitted to the subbandline buffer 103 and temporarily held.

In the t2 period, the DWT coefficient data of the subband LL of the decomposition level 1 is still output for only one line, and the DWT conversion of the decomposition level 2 cannot be executed. Therefore, the Lev2-DWT transform unit 302 does not execute the DWT transform, but only holds the coefficient data in the internal line buffer. The quantization / coding unit 104 quantizes the DWT coefficient data for a total of 1.0 lines of the decomposition level 1 subbands LH and HL sent from the wavelet transform unit 300 using the quantization parameters alternately every cycle. To do. Then, in coding, entropy coding such as EBCOT (Embedded Block Coding with Optimized Truncation) is applied to each subband and output as coded data.

In the t3 period, as in the t0 to t1 periods, data for the number of lines sufficient for applying the reversible 5-3DWT conversion is not input. Therefore, the input pixel data is held in the line buffer in the Lev1-DWT transform unit 301, and the DWT coefficient data is not output during this period. However, the quantization / coding unit 104 reads out the DWT coefficient data of the decomposition level 1 subband HH held in the subbandline buffer 103 in the t2 period, quantizes it using the quantization parameter, and performs entropy coding. And output as encoded data. In this period, only the DWT coefficient data of the subband HH of the decomposition level 1 is encoded, so only 0.5 lines of DWT coefficient data are quantized and encoded based on the pixel data. Therefore, the quantization / coding can be executed in half the period in which the pixel data is input in one line.

In the t4 period, the Lev1-DWT transform unit 301 performs DWT conversion using the pixel data input in the t2 and t3 periods and held in the line buffer and the pixel data input this time, and performs DWT conversion to decompose level 1 subband LL. , LH, HH, LH coefficient data is generated. At that time, since the DWT coefficient data of the subband LL of the decomposition level 1 is used for the DWT conversion of the decomposition level 2, it is transmitted to the Lev2-DWT conversion unit 302. However, even at this time, since the DWT coefficient of the level 1 subband LL is output for only two lines including the amount held in the line buffer in the t2 period, the level 2 DWT conversion cannot be executed. Further, as in the t2 period, the DWT coefficient data of the decomposition level 1 subband LH and HL is output to the quantization / coding unit 104, but the DWT coefficient data of the decomposition level 1 subband HH is the subband line buffer 103. It is sent to and temporarily retained (delayed by one cycle). The quantization / coding unit 104 quantizes the DWT coefficient data for a total of 1.0 lines of the decomposition level 1 subbands LH and HL sent from the wavelet transform unit 300 using the quantization parameters alternately every cycle. , Entropy encoding is applied and output as encoded data.

During the t5 period, the DWT conversion of decomposition levels 1 and 2 is not executed, and the input data is held in the line buffer in the Lev1-DWT conversion unit 301. However, as in the t3 period, the quantization / coding unit 104 reads out the DWT coefficient data (stored in the t4 period) of the decomposition level 1 subband HH held in the subband line buffer 103, and the quantization parameter. Quantized using, entropy-encoded, and output as encoded data.

During the t6 period, the Lev1-DWT transform unit 301 executes the DWT transform of the decomposition level 1. At that time, the DWT coefficient data of the output decomposition level 1 subband LL and the DWT coefficient of the level 1 subband LL held in the line buffer in the Lev2-DWT transform unit 302 during the t2 and t4 periods are used. Then, the Lev2-DWT transform unit 302 executes the DWT transform to calculate and output the coefficient data of the subbands LL, LH, HH, and LH of the decomposition level 2.

As shown in FIG. 4, since the number of coefficient data of decomposition level 2 is half of the coefficient data of decomposition level 1, the subbands LL, LH, HH, and LH of decomposition level 2 are 4 based on the original pixel data. A total of one line (0.25 × 4) of DWT coefficient data is transmitted to the subband line buffer 103 and temporarily held. In addition, if the 1.5 lines of the three subbands LH, HH, and LH output by the DWT conversion of decomposition level 1 are included, the DWT coefficient data of 2.5 lines (1 + 1.5) based on the original pixel data is obtained. It is output during this period.

The quantization / coding unit 104 quantizes the DWT coefficient data for a total of 1.0 lines of decomposition level 1 subbands LH and HL sent from the wavelet transform unit 300 using the quantization parameters alternately every cycle. , Entropy encoding is applied and output as encoded data.

During the t7 period, the DWT conversion of decomposition levels 1 and 2 is not executed, and the input data is held in the line buffer in the Lev1-DWT conversion unit 301. However, here, the quantization / coding unit 104 is held in the subband line buffer 103 during the t6 period and the DWT coefficient data of the decomposition level 1 subband HH held in the subband line buffer 103. The DWT coefficient data of the subbands LH and HL of the decomposition level 2 are quantized and encoded. The quantization / coding unit 104 first quantizes / encodes the DWT coefficient data of the subband HH of the decomposition level 1 corresponding to 0.5 lines based on the pixel data. Subsequently, the quantization / coding unit 104 quantizes / encodes the DWT coefficient data of the subband LH of the decomposition level 2 corresponding to 0.5 lines based on the pixel data. Finally, the quantization / coding unit 104 quantizes / encodes the DWT coefficient of the subband HL of the decomposition level level 2 corresponding to 0.5 line based on the pixel data. As a result, the DWT coefficient for 1.0 line is quantized and encoded using the entire period during which one line of pixel data is input. If the decomposition level 1 subband HH is quantized / coded in this period, the processing order of the DWT coefficient to be quantized / coded is not particularly limited. For example, instead of processing decomposition level 1 subband HH, decomposition level 2 subband LH, and decomposition level 1 subband HL in that order, decomposition level 2 subband LH, decomposition level 2 subband HL, and decomposition level 1 subband HL. The processing order may be changed, such as processing in the order of subband HH. Further, instead of the decomposition level 2 subband LH and the decomposition level 2 subband HL, the decomposition level 2 subband HH and the decomposition level 2 subband LL may be quantized / encoded during this period. In this case, the decomposition level 2 subband LH and the decomposition level 2 subband HL that were not quantized / encoded in this period may be quantized / encoded in the t9 period shown later.

In the t8 period, the DWT coefficient of the decomposition level 1 subband LL required to perform the next decomposition level 2 DWT transform is insufficient. Therefore, only decomposition level 1 1DWT conversion is performed. Further, as in the t2, t4, and t6 periods, the DWT coefficient data of the subband LH and HL of the decomposition level 1 is output to the quantization / coding unit 104, but the DWT coefficient data of the subband HH of the decomposition level 1 is sub. It is transmitted to the band line buffer 103 and temporarily held. The quantization / coding unit 104 quantizes the DWT coefficient data for a total of 1.0 lines of the decomposition level 1 subbands LH and HL sent from the wavelet transform unit 300 using the quantization parameters alternately every cycle. , Entropy encoding is applied and output as encoded data.

During the t9 period, DWT conversion of decomposition levels 1 and 2 is not executed, and the input data is held in the line buffer in the Lev1-DWT conversion unit 301. However, in this t9 period, the quantization / coding unit 104 uses the DWT coefficient data of the decomposition level 1 subband HH held in the subbandline buffer 103 in the t8 period and the subbandline buffer in the t6 period. The DWT coefficient data of the decomposition level 2 subbands HH and LL held in 103 are quantized and encoded. Specifically, the quantization / coding unit 104 first quantizes / encodes the DWT coefficient data of the subband HH of the decomposition level 1 corresponding to 0.5 lines based on the pixel data. The quantization / coding unit 104 subsequently quantizes / encodes the DWT coefficient data of the decomposition level 2 subband HH corresponding to 0.5 lines based on the pixel data. Finally, the quantization / coding unit 104 quantizes / encodes the DWT coefficient data of the subband LL of the decomposition level 2 corresponding to 0.5 lines based on the pixel data. As a result, the DWT coefficient for 1.0 line is quantized and encoded using the entire period during which one line of pixel data is input.

After that, the above process is repeated until the final line of the image data.

By executing the quantization / coding at the configuration and processing timing as described above, the quantization / coding unit 104 has the same number during the period of inputting the pixel data for one line from the image data to be encoded. It only needs to have the ability to process the DWT coefficient data of. That is, the total number of conversion coefficient data input by the quantization / coding unit 104 can be set to the number of pixels in the horizontal direction of the image data to be encoded at the maximum. In the case of the configuration shown in FIG. 3, the quantization / coding unit needs to have the ability to process 2.5 lines of DWT coefficient data during the period of inputting pixel data for one line of the image data to be encoded. It can be understood that the increase in circuit scale can be suppressed compared to the previous case.

In this embodiment, a reversible 5-3 tap filter is used, but any filter such as the lossy 9-7 tap filter proposed in JPEG2000 that holds the relationship of Equation 4 has the same configuration. , It is possible to execute at the processing timing.

Further, in the present embodiment, the case where the wavelet coefficient is executed up to the decomposition level 2 has been described as an example, but when the wavelet coefficient is executed up to a higher level after the decomposition level 3, the same idea can be executed. Is.

For example, in the case of decomposition level 3, a new buffer for temporarily storing the decomposition level 3 subband is provided in the subband line buffer 103, and the retained DWT coefficient is set to a low level subband as shown in the timing diagram of FIG. This can be achieved by preferentially reading from.

Here, for example, if the case of decomposing and coding into subbands up to the decomposition level N is generalized, the wavelet transform unit of the i (i ≧ 2) is the decomposition level generated by the wavelet transform unit of the i-1. The conversion coefficient data of the subband LL of i-1 is input, and the conversion coefficient data of the subbands LL, HL, LH, and HH of the decomposition level i is output. Then, the first wavelet transform unit located first supplies two of the generated decomposition level 1 subbands HL, LH, and HH to the quantization / coding unit without the intervention of a buffer memory. At the same time, the remaining one is supplied to the buffer memory. Then, the other wavelet transform units other than the first wavelet transform unit supply the generated three subbands HL, LH, and HH of each decomposition level to the buffer memory. Then, the quantization / coding unit gives priority to the lower resolution level from the first wavelet transform unit or the buffer memory, and has the same number of conversion coefficients as the number of pixels in the horizontal direction of the image data to be encoded. Data may be input and quantized and encoded.

[Second Embodiment]
Subsequently, a second embodiment of the present invention will be described. FIG. 8 shows a block configuration diagram of the image coding apparatus according to the second embodiment, and FIG. 9 shows a timing chart related to the coding process. The apparatus configuration of the second embodiment is almost the same as that of the first embodiment, but the target held in the subband line buffer 103 is the DWT conversion coefficient data of the subband LH of the decomposition level 1. Is different. Hereinafter, the processing of the image coding apparatus according to the second embodiment shown in FIG. 8 will be described with reference to the timing chart of FIG.

In the period from t0 to t1, pixel data for two lines in total from the image data to be encoded is supplied to the Lev1-DWT transform unit 301. However. At this point, enough data for the number of lines to perform the reversible 5-3DWT conversion has not yet been input. Therefore, the Lev1-DWT conversion unit 301 only holds the input pixel data of two lines in the line buffer in the Lev1-DWT conversion unit 301, and does not execute the DWT conversion during this period.

In the t2 period, the Lev1-DWT transform unit 301 executes DWT conversion using the pixel data for two lines of the coded image held in the line buffer in the t0 and t1 periods and the pixel data for the third line, and decomposes them. Outputs the coefficient data of level 1 subbands LL, LH, HH, and LH. At that time, since the DWT coefficient data of the subband LL of the decomposition level 1 is used to generate the DWT conversion data of the decomposition level 2, it is supplied to the Lev2-DWT conversion unit 302. Further, the DWT coefficient data of the subband HH and HL of the decomposition level 1 is output to the quantization / coding unit 104, but the DWT coefficient data of the subband LH of the decomposition level 1 is supplied to the subband line buffer 103 once. , Retained (delayed).

In the t2 period, the DWT coefficient data of the subband LL of the decomposition level 1 is still output for only one line, and the DWT conversion of the decomposition level 2 cannot be executed. Therefore, the Lev2-DWT transform unit 302 does not execute the DWT transform for the subband LL of the decomposition level 1, and only holds the coefficient in the internal line buffer. The quantization / coding unit 104 quantizes the DWT coefficient data for a total of 1.0 lines of HH and HL subbands of decomposition level 1 sent from the wavelet transform unit 300 using the quantization parameters alternately every cycle. Will be done. Then, in coding, each subband is subjected to entropy coding such as EBCOT (Embedded Block Coding with Optimized Truncation) and output as coded data.

In the t3 period, as in the t0 to t1 periods, data for the number of lines sufficient for applying the reversible 5-3DWT conversion is not input. Therefore, the input pixel data is retained in the line buffer in the Lev1-DWT transform unit 301, and the DWT coefficient is not output during this period. However, the quantization / coding unit 104 reads out the DWT coefficient data of the subband LH of the decomposition level 1 held in the subband line buffer 103 in the t2 period, quantizes it using the quantization parameter, and performs entropy coding. Is applied and output as encoded data. In this period, only the DWT coefficient data of the subband LH of the decomposition level 1 is encoded, so only the DWT coefficient data of 0.5 lines is quantized and encoded based on the pixel data. Therefore, the quantization / coding can be executed in half the period in which the pixel data is input in one line.

In the t4 period, the Lev1-DWT transform unit 301 executes DWT conversion using the pixel data input in the t2 and t3 periods and held in the line buffer and the newly input pixel data, and performs DWT conversion, and the sub-decomposition level 1 sub. Generate coefficient data for bands LL, LH, HH, and LH. At that time, since the DWT coefficient data of the subband LL of the decomposition level 1 is used to generate the DWT conversion data of the decomposition level 2, it is supplied to the Lev2-DWT conversion unit 302. However, at this time, even if the DWT coefficient data of the subband LL of the decomposition level 1 and the amount held in the line buffer in the t2 period are combined, the amount is not divided into two lines. Therefore, the Lev2-DWT converter 302 does not perform the DWT transform. Further, as in the t2 period, the DWT coefficient data of the decomposition level 1 subband HH and HL is output to the quantization / coding unit 104, but the DWT coefficient data of the decomposition level 1 subband LH is the subband line buffer. It is sent to 103 and temporarily retained. The quantization / coding unit 104 quantizes the DWT coefficient data for a total of 1.0 lines of decomposition level 1 subbands HH and HL sent from the wavelet transform unit 300 using the quantization parameters alternately every cycle. , Entropy encoding is applied and output as encoded data.

During the t5 period, the DWT conversion of the decomposition levels 1 and 2 is not executed, and the input data is held in the line buffer in the Lev1-DWT conversion unit 301. However, as in the t3 period, the quantization / coding unit 104 reads out the DWT coefficient data of the decomposition level 1 subband LH held in the subband line buffer 103 in the t4 period, and quantizes using the quantization parameter. It is converted, entropy-encoded, and output as encoded data.

In the t6 period, the Lev1-DWT transform unit 301 executes the DWT transform, generates and outputs the coefficient data of the subbands LL, LH, HH, and LH of the decomposition level 1. At that time, the output DWT coefficient data of the decomposition level 1 subband LL and the DWT of the decomposition level 1 subband LL held in the line buffer in the Lev2-DWT transform unit 302 during the t2 and t4 periods. The Lev2-DWT transform unit 302 executes the DWT transform using the coefficient data. As a result, coefficient data of subbands LL, LH, HH, and LH of decomposition level 2 are generated.

As shown in FIG. 4, since the number of coefficient data of decomposition level 2 is half that of the coefficient data of decomposition level 1, the subbands LL, LH, HH, and LH of decomposition level 2 are 4 based on the original pixel data. In total, one line (0.25 × 4) of DWT coefficient data is transmitted to the subband line buffer 103 and temporarily held. Since the coefficient data is for three 1.5 lines of decomposition level 1 subbands LH, HH, and LH, the number of coefficient data generated in this t6 period is 2.5 lines (1 + 1.5) based on the original pixel data. Become.

The quantization / coding unit 104 quantizes the DWT coefficient data for a total of 1.0 lines of the decomposition level 1 subbands HL and HH sent from the wavelet transform unit 300 using the quantization parameters alternately every cycle. , Entropy encoding is applied and output as encoded data.

In the t7 period, the DWT conversion of the decomposition levels 1 and 2 is not executed, and the input data is held in the line buffer in the Lev1-DWT conversion unit 301. However, in this t7 period, the quantization / coding unit 104 performs the DWT coefficient data of the decomposition level 1 subband LH held in the subbandline buffer 103 in the t6 period, and the subbandline buffer in the t6 period. The DWT coefficient data of the decomposition level 2 subbands LH and HL held in 103 are quantized and encoded. The quantization / coding unit 104 first quantizes / encodes the DWT coefficient data of LH of decomposition level 1 corresponding to 0.5 lines based on the pixel data. The quantization / coding unit 104 subsequently quantizes / encodes the DWT coefficient data of the decomposition level 2 subband LH corresponding to 0.25 lines based on the pixel data. Finally, the quantization / coding unit 104 quantizes / encodes the DWT coefficient data of the subband HL of the decomposition level 2 corresponding to 0.25 lines based on the pixel data. As a result, the DWT coefficient data for 1.0 line is quantized and encoded using the entire period during which one line of pixel data is input.

In the t8 period, the DWT coefficient data of the decomposition level 1 subband LL required to perform the next decomposition level 2 DWT transform is insufficient. Therefore, only the decomposition level 1 DWT transform is performed. Further, as in the t2, t4, and t6 periods, the DWT coefficient data of the subband HH and HL of the decomposition level 1 is output to the quantization / coding unit 104, but the DWT coefficient data of the subband LH of the decomposition level 1 is It is transmitted to the subbandline buffer 103 and temporarily held. The quantization / coding unit 104 quantizes the DWT coefficient data for a total of 1.0 lines of decomposition level 1 subbands HH and HL sent from the wavelet transform unit 300 using the quantization parameters alternately every cycle. , Entropy encoding is applied and output as encoded data.

During the t9 period, the DWT conversion of the decomposition levels 1 and 2 is not executed, and the input data is held in the line buffer in the Lev1-DWT conversion unit 301. However, in this t9 period, the quantization / coding unit 104 contains the DWT coefficient data of the decomposition level 1 subband LH held in the subbandline buffer 103 in the t8 period, and the subbandline buffer in the t6 period. The DWT coefficient data of the decomposition level 2 subbands HH and LL held in 103 are quantized and encoded. Specifically, the quantization / coding unit 104 first quantizes / encodes the DWT coefficient data of the subband LH of the decomposition level 1 corresponding to 0.5 lines based on the pixel data. The quantization / coding unit 104 subsequently quantizes / encodes the DWT coefficient data of the subband HH of the decomposition level 2 corresponding to 0.25 lines based on the pixel data. Finally, the quantization / coding unit 104 quantizes / encodes the DWT coefficient data of the subband LL of the decomposition level 2 corresponding to 0.25 lines based on the pixel data. As a result, the quantization / coding of the DWT coefficient data for 1.0 line is executed using the entire period during which one line of pixel data is input.

After that, the same process is repeated until the final line of the image data.

Here, the reason why the coefficient held in the subband line buffer 103 is used as the coefficient data of the subband LH of the decomposition level 1 will be described with reference to FIG.

FIG. 10 is a diagram showing the number of horizontal DWT coefficients of each subband when an image having a horizontal resolution (number of pixels) of 9 is DWT-converted to resolution level 2.

As shown in FIG. 10, when the horizontal resolution is odd, whether the number of horizontal coefficients of each subband of decomposition level 1 belongs to the subband of the low frequency component (LH = 5 of decomposition level 1) or high frequency. It depends on whether it belongs to a component (HL, HH = 4 at decomposition level 1). Therefore, if the level 1 DWT coefficient data input to the quantization / coding unit 104 in the same period is subband LH, HH or LH, HL, the number of coefficients to be quantized / encoded in one line period is Different timing adjustment is required. Therefore, by setting the DWT coefficients of decomposition level 1 input to the quantization / coding unit 104 in the same period to the subbands HL and HH, the number of coefficients to be quantized / encoded in one line period can be made the same. Because.

103 ... Subband line buffer, 104 ... Quantization / coding unit, 300 ... Wavelet transform unit, 301 ... Lev1-DWT transform unit, 302 ... Lev2-DWT transform unit

Claims (7)

An image coding device that encodes image data.
A conversion means for inputting image data to be encoded, wavelet transforming the input image data, and generating conversion coefficient data for a plurality of subbands.
A coding means that quantizes and encodes the coefficient data of each subband of the plurality of subbands generated by the conversion means, and
It has a buffer means for temporarily holding the subband coefficient data to be supplied to the coding means, which is the conversion coefficient data generated by the conversion means.
Of the plurality of subbands generated by the conversion means, the conversion coefficient data of the preset subbands is supplied to the coding means without going through the buffer means.
The conversion coefficient data of the other subbands among the plurality of subbands is temporarily held by the buffer means, and the conversion coefficient data held by the buffer means is supplied to the coding means. Image coding device. Among the plurality of subbands generated by the conversion means, the conversion coefficient data of the subband having the lowest decomposition level and containing a high frequency component is supplied to the coding means without going through the buffer means. The image coding apparatus according to claim 1. Of the subbands LL, HL, LH, and HH generated by the conversion means, the coefficient data of the subband HH is supplied to the coding means without going through the buffer means, and the coefficient data of the subband LL is The image coding apparatus according to claim 1 or 2, wherein the image coding device is temporarily held by the buffer means and then supplied to the coding means . The coefficient data of the subbands HL and HH are supplied to the coding means without going through the buffer means, and the coefficient data of the subbands LH and LL are temporarily held by the buffer means and then sent to the coding means . The image coding apparatus according to claim 3, wherein the image coding apparatus is supplied. The coding means reads out the coefficient data so that the total number of conversion coefficient data read from the buffer means is the number of pixels in the horizontal direction of the image to be encoded, and performs quantization and coding. The image coding apparatus according to any one of claims 1 to 4, wherein the image coding apparatus is characterized by the above. Any one of claims 1 to 5, wherein the coding means preferentially reads, quantizes, and encodes the coefficient data of the subband having a small decomposition level of the wavelet transform from the buffer means. The image coding apparatus according to the section. The transform means has a plurality of wavelet transform means and has a plurality of wavelet transform means.
Here, the i (i ≧ 2) wavelet transform means inputs the conversion coefficient data of the subband LL of the decomposition level i-1 generated by the wavelet transform means of the i-1, and the subband of the decomposition level i-1. Output conversion coefficient data of LL, HL, LH, HH,
The first wavelet transform means, which is the first of the plurality of wavelet transform means, encodes two of the generated decomposition level 1 subbands HL, LH, and HH without the intervention of the buffer means. And the remaining one is supplied to the buffer means.
The other wavelet transform means other than the first wavelet transform means supply the three generated subbands HL, LH, and HH of each decomposition level to the buffer means.
The coding means is
Quantization and coding are performed by inputting the same number of conversion coefficient data as the number of pixels in the horizontal direction of the image data to be encoded while giving priority to the one having the lower decomposition level from the conversion means or the buffer means. The image coding apparatus according to any one of claims 1 to 6, wherein the image coding apparatus is characterized by performing the above.

Images