JP2000125298A - Device and method for coding image data - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize coding of image data where high image quality can be obtained without degraded by a quantization error. SOLUTION: An image memory 10 stores two-dimensional image data being a coding object when the data are received. Then the image-data G stored in the image memory 10 are coded by acting a block processing section 11, a DCT section 12, a quantization section 13 and a variable length coding section 14 like a coding means. An error compensation section 20 applies inverse quantization to a quantization level Qj obtained by the quantization section 13 and an IDCT section 22 generates decoded image data gk. A subtractor 31 obtains a difference between image data Gk and the decoded image data gk to introduce a quantization error ek. An error distribution section 23 distributes the quantization error ek to pixels of an unprocessed line so as to reflect the quantization error ek on the unprocessed line.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an apparatus and a method for encoding image data.

[0002]

2. Description of the Related Art Conventionally, when an image is handled as image data, the image data is compressed by encoding it, the amount of information is reduced, and the image data is efficiently handled.

[0003] Such conventional image data encoding devices include transform encoding, predictive encoding, block encoding, and the like.
Data compression is performed by performing known encoding processing such as encoding by vector quantization and subband encoding. In each of these encoding processes, the amount of information is reduced by performing predetermined quantization on the input signal.

[0004] Taking the case of transform coding as an example, the structure of a conventional coding apparatus is as shown in FIG. FIG. 13 is a block diagram of a conventional encoding device that encodes image data using a discrete cosine transform (hereinafter, referred to as “DCT”), which is one of orthogonal transforms. When two-dimensional image data to be encoded is input, it is stored in the image memory 10. Then, the image data G stored in the image memory 10 is sent to the DCT unit 12. DC
The T unit 12 derives a plurality of transform coefficients F by applying DCT to the image data G,
Send to The quantization unit 13 performs quantization at a predetermined step width, and generates a quantization level value Q for each transform coefficient F. Then, the variable-length encoding unit 14 performs variable-length encoding according to the occurrence probability of the quantization level value Q, and outputs this as encoded data S.

[0005]

However, in the conventional coding apparatus, the information originally contained in the image data is damaged by performing the coding. In other words, a predetermined quantization is applied to the input signal to generate a quantization error, and image information corresponding to the quantization error is lost.

Therefore, even if the encoded data generated by the conventional encoding device is restored, the image information lost during the encoding is not reproduced, and the quality of the restored image deteriorates.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and has as its object to provide an image data encoding apparatus and method capable of compensating a quantization error and obtaining high-quality image quality. I do.

[0008]

According to one aspect of the present invention, there is provided an apparatus for encoding image data in order to perform image compression, comprising: An image memory for storing image data for each pixel of a two-dimensional image, and (b) sequentially reading out the image data along a predetermined line direction, and the image for a plurality of pixels continuous along the line direction. Encoding means for generating encoded data by performing one-dimensional predetermined encoding based on the data; and (c) calculating an error of the processing line generated by the encoding by using a pixel of an unprocessed line. Error compensating means for compensating the error by reflecting the error in encoding.

According to a second aspect of the present invention, in the image data encoding apparatus according to the first aspect, the encoding means includes the image data of a processing target line read from the image memory and one line. The present invention is characterized in that predetermined encoding is performed on a difference from decoded image data obtained by decoding the preceding encoded data.

According to a third aspect of the present invention, in the image data encoding apparatus according to the first or second aspect, the encoding means sequentially reads out the image data along a predetermined line direction, After performing block division for each predetermined number of pixels that are continuous along the line direction and performing a predetermined orthogonal transformation on the image data for each block,
It is characterized in that predetermined coding is performed.

According to a fourth aspect of the present invention, in the image data encoding apparatus according to the first aspect, the encoding means sequentially reads the image data along a predetermined line direction, and sequentially reads the image data along the line direction. Block division is performed for every predetermined number of pixels along the block, and predetermined vector quantization is performed on the image data for each block.

According to a fifth aspect of the present invention, in the image data encoding apparatus according to the first aspect, the encoding means reads out the image data successively and sequentially along a predetermined line direction, and It is characterized by performing band coding.

According to a sixth aspect of the present invention, there is provided a method for encoding image data in order to perform image compression, comprising: (a) converting image data for each pixel of a two-dimensional image to be encoded to a predetermined value; Sequentially reading along the line direction, and (b) generating coded data by performing one-dimensional predetermined coding based on the image data of a plurality of pixels continuous along the line direction. And (c) by reflecting the error on the processed line generated in the step (b) in the coding of the pixels of the unprocessed line,
Performing the error compensation.

According to a seventh aspect of the present invention, in the image data encoding method according to the sixth aspect, the step (b) comprises:
A predetermined encoding is performed on a difference between the image data of the processing target line read from the image memory and the decoded image data obtained by decoding the encoded data one line before.

[0015]

DESCRIPTION OF THE PREFERRED EMBODIMENTS <1. First Embodiment> First,
A first embodiment will be described. FIG. 1 is a block diagram showing an image data encoding apparatus 100 according to the first embodiment of the present invention. Encoding device 1 for this image data
In 00, an error compensating unit 20 for compensating for a quantization error generated by quantization is provided.

When two-dimensional image data to be encoded is input, it is stored in the image memory 10. Then, the blocking unit 11, the DCT unit 12, the quantization unit 13, and the variable length coding unit 14 function as coding means, so that coding is performed on the image data stored in the image memory 10. .

The image data G stored in the image memory 10
Are sequentially read out pixel by pixel along the main scanning direction X as shown in FIG. When the reading of the first line along the main scanning direction X is completed, the line to be read is advanced by one pixel in the sub-scanning direction Y, and similarly, the pixels are sequentially read out along the main scanning direction X for each pixel. Is read out. Thereafter, by performing the same reading, the image data G for each pixel is sequentially read in the line direction along the main scanning direction X. The image data G thus read is sent to the blocking unit 11.

The blocking unit 11 blocks one line of image data G for every 16 consecutive pixels. That is, as shown in FIG. 3, one line of image data is formed into one-dimensional blocks for every 16 continuous pixels, and a plurality of blocks BC1, BC2,..., BCn (n is an arbitrary integer) is generated. Then, each block BC1, BC
Image data Gk corresponding to 16 pixels in BCn
(That is, G0 to G15) are sequentially transmitted to the DCT unit 12. The blocking unit 11 may transmit the image data G to the DCT unit 12 as image data G0, G1, G2,..., G15 every time 16 image data G are input from the image memory 10.

In the DCT section 12, the image data G0, G1
, G2, ..., G15,

[0020]

(Equation 1)

A DCT operation, which is one of the orthogonal transformations based on the above, is performed to derive a plurality of DCT coefficients Fj of F0 to F15 (that is, F0 to F15). Generally, image data G0, G1, G2,..., G15 for each pixel have a high correlation with adjacent pixels. Therefore, if the image data G0 to G15 are converted into frequency domain data (DCT coefficients), the data is biased to DCT coefficients (those having a relatively small value of j) corresponding to the low frequency band, and correspond to the high frequency band. The variance of DCT coefficients (those having a relatively large value of j) is small. Then, the DCT coefficient Fj obtained as described above is sent to the quantization unit 13.

In the quantization unit 13, 15 DCTs input from the DCT unit 12 are realized in order to realize high coding efficiency.
The quantization is performed on each of the coefficients Fj to generate a quantization level value Qj. The quantization performed here is shown in FIG.
As shown in FIG. 4, a corresponding quantization level Qj is set for each predetermined step width (in FIG. 4, the step width is 16), and when the DCT coefficient Fj is input, the DCT coefficient Fj
A quantization level value Qj is output according to the value of j. It is needless to say that an arbitrary step width can be applied in the quantization unit 13.

For example, as shown in FIG.
6, when the DCT coefficient Fj is "8.ltoreq.Fj <24", the quantization level value Qj = 1, so that the DCT coefficient Fj
Regardless of whether j is "8" or "23", the quantization level value Qj = 1, and a quantization error occurs here.

In the quantization section 13, the quantization level value Qj (the value of j is relatively small) corresponding to the low-frequency range may be a value other than "0". Although the number increases, the quantization level value Qj corresponding to the high frequency (the value of j is relatively large) has a relatively large value of "0".

Then, the quantization level Qj (j = 0 to 15) for the block obtained by the quantization unit 13 is obtained.
Is sent to the variable-length encoding unit 14 and also to an error compensating unit 20 provided to eliminate the quantization error.

The variable length coding unit 14 performs variable length coding according to the occurrence probability of the quantization level value Qj obtained from the quantization unit 13 to generate and output coded data S. As a result, encoded data S having high encoding efficiency can be generated.

The error compensating section 20 includes an inverse quantizing section 21 and an inverse discrete cosine transform (hereinafter referred to as “IDCT”) section 22.
, A subtractor 31, and an error distribution unit 23. The quantization level value Qj output from the quantization unit 13 is input to the inverse quantization unit 21.

In the inverse quantization unit 21, the quantization level value Qj
The inverse quantization level value qj is generated by selecting the median value of the step width adopted by the quantization unit 13 according to. For example, as a result of the quantization performed by the quantization unit 13 with the step width 16 as shown in FIG.
When j = 0, the inverse quantization level value qj = 0, and when the quantization level value Qj = 1, the inverse quantization level value qj = 16, and the quantization level value Qj = 2. In this case, the inverse quantization level value qj is 32. The correspondence between such a quantization level value Qj and the inverse quantization level value qj is stored in a memory or the like in advance, and the inverse quantization unit 2
1 reads out the inverse quantization level value qj corresponding to the value of the quantization level value Qj and sends it to the IDCT unit 22.

In the IDCT unit 22, the inverse quantization level value q
When 0, q1, q2, ..., q15 are input,

[0030]

(Equation 2)

An IDCT operation is performed to derive 16 decoded image data gk (that is, g0 to g15) corresponding to 16 pixels included in the block. Therefore,
The decoded image data gk is image data containing a quantization error.

Then, the subtractor 31 obtains the difference between the image data Gk corresponding to 16 pixels of the block obtained from the blocking unit 11 and the decoded image data gk decoded by the IDCT unit 22, and calculates the quantum in the image data. The conversion error ek is determined. FIG. 5 shows this quantization error ek.
FIG. That is, a value obtained by subtracting the decoded image data g0 from the image data G0 is a quantization error e0, and a value obtained by subtracting the decoded image data g1 from the image data G1 is a quantization error e1. Hereinafter, the same operation is performed for each of the 16 corresponding pixels, and a quantization error ek (i.e., e0 to e15) occurring in each pixel is obtained. The quantization error ek for each pixel thus obtained is guided to the error distribution unit 23.

The error distribution section 23 distributes the quantization error ek for each pixel to the pixels near the unprocessed line.
More specifically, the quantization error ek generated for each pixel of the current encoding target line is distributed to neighboring pixels in the next line to be encoded. FIG. 6 is a diagram illustrating the concept of error distribution. As shown in FIG. 6, assuming that the current line to be encoded is the m-th line and a quantization error ek is detected at the k-th pixel Px in the block, this quantization error ek is calculated as (m + 1) Pixel Pk of the th line
Are distributed to three pixels, i.e., a (k-1) th pixel Pa, a kth pixel Pb, and a (k + 1) th pixel Pc located in the vicinity of.

In this distribution, a quarter of the quantization error ek, that is, "ek / 4" is distributed to the (k-1) th pixel Pa, and two minutes to the kth pixel Pb. And a 1/4 quantization error ek, ie, “ek / 4”, is distributed to the (k + 1) th pixel Pc. Accordingly, the k-th pixel Pb on the (m + 1) -th line receives the distribution of a quarter quantization error from the (k-1) -th pixel and the (k + 1) -th pixel on the m-th line, respectively. The k-th pixel Px receives a distribution of a half quantization error. Such distribution is performed in the same manner for all 16 pixels included in the block.

The quantization error ek distributed to the neighboring pixels of the unprocessed line is sent to the image memory 10 and is cumulatively added to the image data G of the distributed pixels. As a result, the image data G stored in the image memory 10 is sequentially updated to a value reflecting the quantization error ek as the encoding for each line proceeds.

As a result, the image data G of the second and subsequent lines
Is read from the image memory 10 and the encoding process is performed, the quantization error ek caused by the previous line is
Has been reflected. In other words, since the quantization error ek caused by the previous line is included when the current line is encoded, the quantization error has been eliminated.

Therefore, according to the image data encoding apparatus 100 shown in this embodiment, only the quantization error ek is extracted by the subtractor 31 of the error compensating section 20, and this quantization error ek is converted to the unprocessed line. Since it is configured to be distributed to neighboring pixels, encoding can be performed without losing information originally included in image data, so that a code with high quality image quality is maintained. Can be performed.

In this embodiment, a case has been exemplified in which 16 consecutive pixels for each line are divided into one block. However, any number of pixels can be applied to one block. This is the same in the following embodiments.

<2. Second Embodiment> Next, a second embodiment will be described. FIG. 7 is a block diagram showing an image data encoding device 100a according to the second embodiment of the present invention. The blocks having the same functions as those in the first embodiment shown in FIG. 1 are denoted by the same reference numerals.

The image data encoding apparatus 100a is also provided with an error compensator 20 for compensating for a quantization error generated by quantization. Then, when two-dimensional image data to be encoded is input, it is stored in the image memory 10, and similarly to the first embodiment, the blocking unit 11, the DCT unit 12, the quantization unit 13, and the variable-length code When the encoding unit 14 functions as an encoding unit, the encoding is performed on the image data stored in the image memory 10.

The error memory 24 of the error compensator 20
Is initialized to "0" prior to encoding of image data.

The image data G stored in the image memory 10
Are read out line by line along the main scanning direction X as in the first embodiment. The read image data G is sent to the blocking unit 11. As described above, the blocking unit 11 divides the image data for one line into one-dimensional blocks for every 16 continuous pixels, and stores the image data Gk (that is, G0 to G15) corresponding to the 16 pixels in each block.
Are sequentially transmitted to the DCT unit 12. The DCT unit 12 performs the DCT operation of Equation 1 based on the image data Gk,
The DCT coefficients Fj (ie, F0 to F15) are derived.
The DCT coefficient Fj is sent to the adder 32.

In the adder 32, the DCT coefficient Fj and the quantization error Ej to be added to the DCT coefficient derived from the error compensator 20 are added, and a compensation value F'j of the DCT coefficient is derived. Then, the compensation value F′j of the DCT coefficient is sent to the quantization unit 13.

The quantization section 13 quantizes each of the DCT coefficient compensation values F'j to generate a quantization level value Qj. The quantization performed here is also the first
This is the same mode as the embodiment. Therefore, a quantization error occurs in the quantization unit 13. Then, the variable length coding unit 14 performs variable length coding on the quantization level value Qj and outputs coded data S.

In the error compensating section 20, the inverse quantizing section 2
The 1 and the subtractor 33 and the error memory 24 compensate for the quantization error generated by the quantization.

The inverse quantization section 21 performs inverse quantization based on the quantization level value Qj to generate an inverse quantization level value qj. Then, the obtained inverse quantization level value qj is guided to the subtractor 33.

Then, the difference between the compensation value F'j of the DCT coefficient and the inverse quantization level value qj is obtained in the subtractor 33.
The quantization error ej in the DCT coefficient Fj is obtained. The quantization error ej in the DCT coefficient Fj obtained here is sent to the error memory 24.

The error memory 24 has a storage area capable of storing at least one line of image data G along the main scanning direction X. A subtractor is provided in the area where the quantization error Ej is stored. 33, the quantization error ej
Overwrite. As a result, the quantization error e stored in the error memory 24 as the encoding for each line progresses.
j will be updated sequentially. Then, the encoding for the next line is started, and when the DCT coefficient Fj at the position corresponding to the quantization error ej is obtained, the quantization error e
j is sent to the adder 32 as Ej. The 16 DCTs derived from this block are represented by such a distribution.
The same is done for all coefficients.

At the time of the encoding process of the first line of the image data G, the error memory 24 has been initialized, so that the DCT coefficient compensation value F′j obtained by the adder 32 is
j equals When the image data G on the second and subsequent lines is read from the image memory 10 and subjected to the encoding process, the quantization error e caused by the previous line is obtained.
Since j is stored in the error memory 24, the DCT coefficient compensation value F'j is a value reflecting the error caused by the previous line. In other words, the quantization error ej caused by the previous line is reflected on the DCT coefficient Fj when the current line is encoded, so that the quantization error can be eliminated.

Therefore, according to the image data encoding apparatus 100a shown in this embodiment, only the quantization error ej is extracted by the subtracter 33 of the error compensating unit 20, and this quantization error ej is used as the unprocessed line. Since it is configured to be distributed to neighboring pixels, encoding can be performed without losing information originally included in image data, so that a code with high quality image quality is maintained. Can be performed.

Further, according to the image data encoding apparatus 100a shown in this embodiment, it is not necessary to provide the IDCT section 22 as shown in the first embodiment, so that the apparatus configuration can be simplified. In addition to the above, the processing efficiency can be improved.

<3. Third Embodiment> Next, a third embodiment will be described.

In each of the above embodiments, the one-dimensional D
Since the image data is encoded by performing the CT, the correlation between the lines (that is, the sub-scanning direction Y
Is not used. For this reason, the encoding efficiency is lower than the encoding by the two-dimensional DCT.

Therefore, in this embodiment, a configuration will be described in which the coding efficiency is improved by one-dimensional DCT by coding the difference between adjacent pixels for each pixel.

FIG. 8 is a block diagram showing an image data encoding apparatus 100b according to a third embodiment of the present invention. The blocks having the same functions as those in the first embodiment shown in FIG. 1 are denoted by the same reference numerals.

The image data encoding device 100b is also provided with an error compensator 20 for compensating for a quantization error generated by quantization. Also, a decoded image memory 41 is provided to obtain a correlation between adjacent lines.

When two-dimensional image data to be encoded is input, it is stored in the image memory 10 and, as in the above embodiments, is divided into the blocking unit 11 and the DCT unit 1.
2. The image data stored in the image memory 10 is encoded by the quantization unit 13 and the variable length encoding unit 14 functioning as encoding means.

It is assumed that the decoded image memory 41 has been initialized to "0" prior to encoding of image data.

The image data G stored in the image memory 10
Are read out line by line along the main scanning direction X as in the first embodiment. The read image data G is guided to the subtractor 34.

The subtractor 34 derives the difference between the image data G for each pixel and the decoded image data g of the pixel for the previous line corresponding to that pixel, and generates difference data D between the lines. Then, the difference data D is sent to the blocking unit 11.

As described above, the blocking unit 11
The difference data D for the line is divided into one-dimensional blocks for every 16 continuous pixels, and the difference data Dk (that is, D0 to D15) corresponding to the 16 pixels in each block are sequentially transmitted to the DCT unit 12. The DCT unit 12 performs the DCT operation of Equation 1 based on the difference data Dk, and derives 16 DCT coefficients Fj (that is, F0 to F15). The DCT coefficient Fj is sent to the quantization unit 13.

In the quantization unit 13, the DCT coefficient Fj
Are quantized, and the quantization level value Qj
Generate The quantization performed here is also the same as in the first embodiment. Therefore, a quantization error occurs in the quantization unit 13. And the variable length coding unit 1
In step 4, variable length coding is performed on the quantization level value Qj, and coded data S is output.

In the error compensating section 20, the inverse quantizing section 2
1, the IDCT unit 22, the subtractor 35, and the error distribution unit 23 compensate for the quantization error generated by the quantization.

The inverse quantization unit 21 performs inverse quantization based on the quantization level value Qj to generate an inverse quantization level value qj. Then, the obtained inverse quantization level value qj is ID
It is led to the CT unit 22.

In the IDCT unit 22, the inverse quantization level value q
When j is input, an IDCT operation is performed based on the above equation 2 to derive 16 pieces of decoded difference data dk (that is, d0 to d15) corresponding to the difference data Dk for each of the 16 pixels included in the block. . The decoded difference data dk becomes difference data including a quantization error.

Then, the subtracter 35 calculates the difference between the difference data Dk corresponding to the 16 pixels of the block obtained from the blocking unit 11 and the decoded difference data dk decoded by the IDCT unit 22, and calculates the quantization error ek. Lead. That is, from the difference data D0 to the decoded difference data d0
Is the quantization error e0, and the difference data D1
Is obtained by subtracting the decoded difference data d1 from the quantization error e.
Set to 1. Hereinafter, the same operation is performed for each of the 16 corresponding pixels, and the quantization error e generated at each pixel is calculated.
k (that is, e0 to e15) is obtained. The quantization error ek for each pixel obtained in this manner is calculated by the error distribution unit 23.
It is led to.

In the error distribution section 23, as shown in the first embodiment, the quantization error ek for each pixel is distributed to pixels near the unprocessed line. Then, the quantization error ek distributed to the neighboring pixels of the unprocessed line is sent to the image memory 10 and is cumulatively added to the image data G of the distributed pixels. As a result, the image data G stored in the image memory 10 is sequentially updated to a value reflecting the quantization error ek as the encoding for each line proceeds.

As a result, similarly to the first embodiment, when the image data G of the second and subsequent lines is read from the image memory 10 and subjected to the encoding process, the quantization generated by the previous line is performed. The value reflects the error ek. In other words, since the quantization error ek caused by the previous line is included when the current line is encoded, the quantization error has been eliminated.

In this embodiment, the IDC
The decoded difference data dk obtained from the T unit 22 is added to an adder 36.
Is led to. Then, the adder 36 adds the decoded image data g for the previous line sent to the subtractor 34 and the decoded difference data dk to derive decoded image data gk corresponding to each pixel of the current line. The decoded image data gk of the current line thus obtained is sent to the decoded image memory 41, and the decoded image data g of the previous line is changed to the decoded image data gk of the current line.
Will be overwritten.

As described above, in this embodiment, when performing one-dimensional DCT on a line to be processed and performing encoding, the decoded image data g for the previous line is read from the decoded image memory 41. The correlation between the lines (that is, the correlation in the sub-scanning direction Y) can be obtained by calculating the difference between the previous line and the current line for each image. Therefore, the encoding efficiency can be improved as compared with the first and second embodiments.

Therefore, according to the image data encoding apparatus 100b shown in this embodiment, only the quantization error ek is extracted by the subtractor 35 of the error compensating unit 20, and this quantization error ek is converted to the unprocessed line. Since it is configured to be distributed to neighboring pixels, encoding can be performed without losing information originally included in image data, so that a code with high quality image quality is maintained. Can be performed. In addition, since encoding is performed using the correlation in the sub-scanning direction Y, encoding efficiency can be improved.

In the first to third embodiments described above, by providing a buffer memory between the quantization unit 13 and the variable length coding unit 14, the data amount can be further reduced. FIG. 9 is a diagram showing an outline of such an encoding process. FIG. 9 shows, as an example, a case where a buffer memory provided between the quantization unit 13 and the variable length coding unit 14 stores the quantization level values Qj for eight lines in the main scanning direction X. I have. First, as shown in FIG. 9, quantization level values Q0 to Q15 corresponding to the same block position for eight lines are extracted. Then, as shown by the arrows in FIG. 9, the quantization level values Q0 of the first to eighth lines are sequentially output to the variable length coding unit 14,
Next, the quantization level values Q1 for the first to eighth lines are sequentially output to the variable length coding unit 14. Hereinafter, similarly, the quantization level values Qj for the first to eighth lines are sequentially output to the variable length encoding unit 14, and finally, the quantization level values Q15 for the first to eighth lines are variable length encoded. Output to the conversion unit 14. By sequentially outputting the quantization level values Qj for eight lines to the variable length encoding unit 14 in the form along the sub-scanning direction Y, the value of “0” becomes smaller as the value of j approaches 15. Are continuously variable length coding units 1
4 more often. Therefore, in the variable-length encoding unit 14, run-length encoding can be applied to such a portion where the value of “0” is continuously distributed, so that the data amount can be further reduced. It becomes possible.

When outputting the quantization level value Qj from the buffer memory as described above, if all the values output after the arbitrary quantization level value Qj are "0",
If an EOB (End Of Block) signal indicating the end of data is output, the data amount can be further reduced.

In the first to third embodiments, the encoding processing for performing DCT, which is one of orthogonal transforms, has been described. However, it is needless to say that the present invention can be applied to other orthogonal transforms.

<4. Fourth Embodiment> Next, a fourth embodiment will be described. FIG. 10 is a block diagram showing an image data encoding device 100c according to the fourth embodiment of the present invention. In this embodiment, encoding of image data is performed using well-known vector quantization. Blocks having functions similar to those of the above-described embodiments are denoted by the same reference numerals.

The image data encoding apparatus 100c is also provided with an error compensator 20 for compensating for a quantization error generated by quantization. When two-dimensional image data to be encoded is input, the two-dimensional image data is stored in the image memory 10, and stored in the image memory 10 by the blocking unit 11 and the vector quantization unit 17 functioning as an encoding unit. The encoded image data is encoded.

The error compensating unit 20 includes a decoding unit 25
, A subtracter 37 and an error distribution unit 23.

The image data G stored in the image memory 10
Are read out line by line along the main scanning direction X as in the first embodiment. The read image data G is sent to the blocking unit 11. As described above, the blocking unit 11 divides the image data for one line into one-dimensional blocks for every 16 continuous pixels, and stores the image data Gk (that is, G0 to G15) corresponding to the 16 pixels in each block.
Are sequentially transmitted to the vector quantization unit 17.

In the vector quantization section 17, 1
A six-dimensional space is defined. This 16-dimensional space is divided into, for example, 65536 subspaces, and one quantized representative vector is determined for each of the divided spaces. An index value Q of 0 to 65536 is individually set for each quantized representative vector. Then, in a memory (not shown) provided in the vector quantization unit 17,
The correspondence between each quantized representative vector and the corresponding index value Q is stored.

Then, the vector quantizing unit 17 converts the one-dimensional continuous 16 bits contained in the block to be encoded.
When the image data Gk is input, the image data G0 to G
15 is defined as a position vector in the 16-dimensional space, and the position vector is specified in which of the 65536 subspaces in the 16-dimensional space is included.
Then, a quantized representative vector of one specified subspace is obtained, and an index value Q of the specified quantized representative vector is obtained by referring to the memory. Then, the vector quantization unit 17 encodes and outputs the index value Q.

Here, when the vector data G0 to G15 are used as position vectors in a 16-dimensional space, the vector quantizer 17 approximates one quantized representative vector, so that a quantization error occurs. Become.

The decoding section 25 has a memory in which the correspondence between the index value Q and the corresponding quantized representative vector is stored, similarly to the memory provided inside the vector quantization section 17. , When an index value Q is input, a corresponding quantized representative vector is specified,
From the quantized representative vector, the decoded image data g0 to g15
Get. The decoded image data gk thus obtained
Is sent to the subtractor 37.

Then, in the subtracter 37, the image data Gk for the corner pixels in the block obtained from the blocking unit 11 and the decoded image data g corresponding to those pixels
k and a quantization error e generated for each pixel.
k is sent to the error distribution unit 23.

In the error distribution section 23, the quantization error ek for each pixel is distributed to pixels near the unprocessed line, as in the first embodiment. Then, the quantization error ek distributed to the neighboring pixels of the unprocessed line is sent to the image memory 10 and is cumulatively added to the image data G of the distributed pixels. The image data G stored in the image memory 10 is sequentially updated to a value reflecting the quantization error ek as the encoding for each line progresses.

As a result, the image data G for the second and subsequent lines
Is read from the image memory 10 and the encoding process is performed, the quantization error ek caused by the previous line is
Has been reflected. In other words, since the quantization error ek caused by the previous line is included when the current line is encoded, the quantization error has been eliminated.

Therefore, according to the image data encoding apparatus 100c shown in this embodiment, only the quantization error ek is extracted by the subtractor 37 of the error compensating section 20, and this quantization error ek is converted to the unprocessed line. Since it is configured to be distributed to neighboring pixels, encoding can be performed without losing information originally included in image data, so that a code with high quality image quality is maintained. Can be performed.

The decoded image memory 41 shown in the third embodiment can be provided also in the image data encoding device 100c shown in this embodiment. That is, in order to use the correlation between adjacent lines,
The difference between the image data G and the decoded image data can be vector-quantized. In this case, the coding efficiency can be further improved.

<5. Fifth Embodiment> Next, a fifth embodiment will be described. FIG. 11 is a block diagram showing an image data encoding device 100d according to the fifth embodiment of the present invention. In this embodiment, image data is encoded using well-known subband encoding. Blocks having functions similar to those of the above-described embodiments are denoted by the same reference numerals.

The image data encoding apparatus 100d of this embodiment includes an image memory 10 for storing two-dimensional image data to be encoded, and 16 band-pass filters BPF0 to BPF15 having different band frequencies. Analysis filter bank 18 and band pass filters BPF0 to BPF0
A plurality of quantization processing units 19a to 19 that perform quantization on each of the analysis values f0 to f15 output from the BPF 15
The quantization unit 19 in which p is provided, and the quantization level values Q0 to Q15 obtained from the respective quantization processing units 19a to 19p.
Variable-length coding unit 1 for sequentially inputting and performing variable-length coding
4 is provided.

When two-dimensional image data to be encoded is input, it is stored in the image memory 10, and the analysis filter bank 18, the quantization unit 19, and the variable length encoding unit 14 function as encoding means. , Image memory 1
The encoding is performed on the image data stored in 0.

In this embodiment, an error compensating section for compensating for a quantization error generated by quantization is provided inside each of the quantization processing sections 19a to 19p as described later.

The image data G stored in the image memory 10
Are read out line by line along the main scanning direction X in the same manner as in the first embodiment, and are sequentially analyzed by the analysis filter bank 18.
Sent to

In the analysis filter bank 18, each band-pass filter BPF0 to BPF15
Are sequentially input to generate analysis values f0 to f15 of the image data G corresponding to each frequency range. Here, at the time of analysis for each frequency range in the analysis filter bank 18, since it is configured to be down-sampled to 1/16, one set of analysis values f0 to f15 is provided for every 16 pixels.
Is generated.

A set of analytical values f0 obtained in this manner
.About.f15 are sent to the quantization unit 19, and quantization is performed by quantization processing units 19a to 19p provided individually for each.

FIG. 12 shows each of the quantization processing units 19a to 19p.
FIG. 2 is a block diagram showing an internal configuration of the device. In FIG. 12, any one of the analysis values f0 to f15 is indicated as fJ. Further, each of the quantization processing units 19a to 19p
Is provided with an error compensator 20 as shown.

As shown in FIG. 12, the analysis value fJ is sent to the adder 38 when input to the quantization processing units 19a to 19p. In the adder 38, the quantization error EJ generated for the previous line stored in the error memory 24 and the analysis value fJ are added to generate a compensation value f'J of the analysis value. This compensation value f'J is sent to the quantization unit 13.

The quantizer 13 quantizes each of the analysis value compensation values f'J to generate a quantization level value QJ. The quantization performed here is also the same as in the first embodiment. Therefore, the quantization unit 13
, A quantization error occurs. The quantization level values QJ generated in this manner are used as the quantization processing units 19a to 19a.
The output of 9p is sent to the variable length coding unit 14 and sent to the inverse quantization unit 21 of the error compensation unit 20.

The inverse quantization unit 21 performs inverse quantization based on the quantization level value QJ to generate an inverse quantization level value qJ. Then, the obtained inverse quantization level value qJ is guided to the subtractor 39.

Then, the difference between the compensation value f'J of the analysis value and the inverse quantization level value qJ is obtained in the subtractor 39, and the quantization error eJ in the analysis value fJ is obtained. The quantization error eJ in the obtained analysis value fJ is stored in the error memory 24.
Sent to

As described in the second embodiment, the error memory 24 overwrites the area where the quantization error EJ has been stored with the quantization error eJ obtained from the subtractor 39. As a result, the quantization error eJ stored in the error memory 24 is sequentially updated as the encoding for each line progresses. Then, encoding of the next line is started, and when an analysis value fJ at a position corresponding to the quantization error eJ is obtained, the quantization error eJ stored in the error memory 24 is sent to the adder 38 as EJ. It will be.

In the encoding process of the first line of the image data G, since the error memory 24 has been initialized, the compensation value f'J of the analysis value obtained by the adder 38 becomes equal to the analysis value fJ. . When the image data G of the second and subsequent lines are sequentially read from the image memory 10 and the encoding process is performed, the quantization error eJ generated by the previous line is stored in the error memory 24. The compensation value f'J of the analysis value is a value reflecting the error caused by the previous line. In other words, since the quantization error ej caused by the previous line is reflected in the analysis value fJ when the current line is encoded, the quantization error can be eliminated.

The variable length coding section 14 performs variable length coding on the quantization level values Q0 to Q15 and outputs coded data S.

As described above, according to the image data encoding device 100d shown in this embodiment, the error compensating unit 2
Since only the quantization error eJ is extracted by the subtractor 39 of 0, and the quantization error eJ is configured to be distributed to the pixels near the unprocessed line, the quantization error eJ was originally included in the image data. Since encoding can be performed without losing information, encoding can be performed while maintaining high quality image quality.

[0104]

As described above, according to the first to fifth aspects of the present invention, an image memory for storing image data for each pixel of a two-dimensional image to be encoded,
Image data is sequentially read out along a predetermined line direction,
Encoding means for generating encoded data by performing one-dimensional predetermined encoding based on image data for a plurality of pixels continuous along the line direction, and an already processed line generated by the encoding means And error compensation means for compensating the error by reflecting the error of the error in the coding of the pixels of the unprocessed line, so that a high-quality image in which such an error is compensated can be obtained.

According to the sixth and seventh aspects of the present invention, a step of sequentially reading out image data for each pixel of a two-dimensional image to be encoded along a predetermined line direction;
A step of generating encoded data by performing one-dimensional predetermined encoding based on image data of a plurality of pixels that are continuous along the line direction; And a step of compensating the error by reflecting the error in the coding of the pixels of the unprocessed line. Therefore, such an error can be compensated to obtain a high-quality image.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an image data encoding device according to a first embodiment of the present invention.

FIG. 2 is a diagram showing a procedure for reading image data from an image memory.

FIG. 3 is a diagram illustrating blocking of one line of image data in a blocking unit.

FIG. 4 is a diagram illustrating a correspondence between an input and an output in a quantization unit.

FIG. 5 is a conceptual diagram for deriving a quantization error.

FIG. 6 is a diagram illustrating the concept of error distribution.

FIG. 7 is a block diagram illustrating an image data encoding device according to a second embodiment of the present invention.

FIG. 8 is a block diagram illustrating an image data encoding device according to a third embodiment of the present invention.

FIG. 9 is a diagram illustrating an outline of an encoding process in which a buffer memory is provided.

FIG. 10 is a block diagram illustrating an image data encoding device according to a fourth embodiment of the present invention.

FIG. 11 is a block diagram illustrating an image data encoding device according to a fifth embodiment of the present invention.

FIG. 12 is a block diagram illustrating an internal configuration of a quantization processing unit.

FIG. 13 is a block diagram showing a conventional image data encoding device.

[Explanation of symbols]

 Reference Signs List 10 image memory 11 blocking unit 12 DCT (discrete cosine transform) unit 13 quantizing unit 14 variable length coding unit 20 error compensating unit (error compensating means) 21 inverse quantizing unit 22 IDCT (discrete cosine inverse transform) unit 23 error Distributor 24 Error memory

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5C059 KK00 LB05 MA02 MA04 MA23 MA47 MC11 MD04 ME01 TA41 TA46 TC06 TC08 TC42 UA02 UA05 UA36 UA38 5C078 AA04 BA57 BA62 CA22 DA00 DA01 DB05 9A001 EE04 EE05

Claims (7)

[Claims] 1. An apparatus for encoding image data for performing image compression, comprising: (a) an image memory for storing image data for each pixel of a two-dimensional image to be encoded; By sequentially reading out the image data along a predetermined line direction and performing one-dimensional predetermined encoding based on the image data for a plurality of pixels continuous along the line direction, encoded data is obtained. Encoding means for generating; and (c) an error compensating means for compensating for the error by reflecting an error about the processed line generated by the encoding in encoding of pixels of the unprocessed line. An encoding device for image data, characterized in that: 2. The image data encoding device according to claim 1, wherein the encoding unit converts the image data of a line to be processed read from the image memory and the encoded data of one line before. An image data encoding device, which performs a predetermined encoding on a difference from decoded image data. 3. The image data encoding device according to claim 1, wherein the encoding unit sequentially reads the image data along a predetermined line direction and continuously reads the image data along the line direction. An image data encoding apparatus, comprising: performing block division for each of a predetermined number of pixels, performing predetermined orthogonal transformation on the image data for each block, and performing predetermined encoding. 4. The image data encoding device according to claim 1, wherein the encoding unit sequentially reads out the image data along a predetermined line direction, and reads a predetermined number of the image data successively along the line direction. An image data encoding apparatus, comprising: performing block division for each pixel; and performing predetermined vector quantization on the image data for each block. 5. The image data encoding device according to claim 1, wherein the encoding unit performs sub-band encoding by sequentially reading out the image data sequentially along a predetermined line direction. An encoding device for image data as a feature. 6. A method of encoding image data to perform image compression, comprising: (a) sequentially reading out image data for each pixel of a two-dimensional image to be encoded along a predetermined line direction; (B) performing coded one-dimensionally based on the image data for a plurality of pixels that are continuous along the line direction to generate coded data; A step of compensating the error by reflecting the error on the processed line generated in the step (b) in the coding of the pixels of the unprocessed line, and encoding the image data, Method. 7. The image data encoding method according to claim 6, wherein said step (b) comprises: said image data of a line to be processed read from said image memory; A predetermined encoding is performed on a difference from decoded image data obtained by decoding the image data.