JP2000244922A - Method for compressing picture data - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a method for compressing picture data by which acquired picture data with high definition can be compressed at a high compression rate. SOLUTION: Picture data (source data) acquired by a CCD 1 a 12-bit data are divided into high-order bit data in high-order 8 bits and low-order bit data in low-order 4 bits. A JPEG lossless coding is applied to the high-order bit data and a valid bit number (4 bits or below) of the low-order bit data is decided on the basis of a size (digit number) of the source data. The low-order bit data within a bit width of the valid bit number are extracted and stored in a storage medium 2 together with the compressed high-order bit data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for compressing image data, and more particularly to a method for compressing image data useful for compressing / decompressing multi-level halftone image data having a large number of gradations.

[0002]

2. Description of the Related Art With the development of image input devices (for example, CCDs used in digital cameras), image data has been taken in with high definition. The original data (RAW data) thus obtained with high definition is subjected to image processing by a user using a personal computer or the like, and thereafter, a desired image is obtained on an output device (for example, a CRT or a printer). It has become.

Image data represented by a signal (signal charge) from a CCD is converted into a digital signal of 10 to 12 bits per pixel by, for example, an A / D converter. When converted into a 12-bit digital signal, the maximum gradation is "4096". Thus 10
The image data captured in high definition with bits to 12 bits is changed into image data of about 8 bits per pixel,
The image is reproduced on the output device.

As described above, the reason why the number of bits of image data per pixel is represented by 10 bits to 12 bits at the time of capturing is as follows. This is because it is necessary to obtain image data evenly in the range. On the other hand, at the time of reproduction, since the brightness difference actually represented in one frame (screen) is smaller than that in the shooting environment, the brightness of the image in one frame can be sufficiently expressed by image data of about 8 bits.

The image data captured by the CCD is stored in a storage medium, and the image data is read out from the storage medium as appropriate. Conventionally, the storage medium is changed to 8 bits per pixel. Image data is stored.

Here, if the image data of 8 bits per pixel is stored as it is, the amount of data per frame (one screen) becomes enormous. Therefore, after compressing the image data, the image data is stored in the storage medium. Like that. General image data compression processing includes DPCM processing on original data (RAW data), Huffman coding, arithmetic coding, JPEG lossless coding using these as appropriate, Ziv-
Universal coding and the like represented by the Lempel method are known.

[0007]

However, when the image data of 8 bits per pixel is stored as the original data in the storage medium, the user can process / modify the original data of 8 bits per pixel as desired. When adding
The corrected image data becomes 8 bits or less per pixel, and the image quality deteriorates.

Thus, the high-definition image data (10 to 12 bits) obtained by the CCD is stored as it is in the storage medium as the original data, and the original data is read and processed / corrected. Is performed, the image quality of the processed / corrected image data can be improved. When the high-definition original data is stored in a storage medium, it is necessary to perform the above-described compression processing (preferably, lossless encoding processing).
Even if the same compression processing is performed on 8-bit image data, a high compression ratio cannot be achieved. This is because when image data is compressed (particularly in the case of DPCM encoding), the higher the correlation between pixel data between adjacent pixels, the higher the compression ratio can be obtained. In the case of data, the correlation of image data between pixels is remarkable in the higher 6 bits to 8 bits, but lower 3 to 4 bits.
This is because the bits have little correlation. Therefore, for 10-bit to 12-bit image data,
Even if compression is performed as it is, a high compression ratio cannot be obtained, and even after compression, a storage medium with large data and large capacity is still required.

The present invention has been made in view of such circumstances, and has as its object to provide an image data compression method capable of compressing image data obtained with high definition at a high compression ratio.

[0010]

In order to achieve the above object, the invention according to claim 1 compresses image data obtained by the image input means (1) with a fixed number of bits (12 bits) per pixel. In doing so, the original data for each pixel having the predetermined number of bits is converted into upper bit data represented by a predetermined upper bit number (8 bits) and lower bit data represented by a lower predetermined bit number (4 bits). And reversible encoding (J
PEG lossless coding); determining the number of effective bits of the lower-order bit data (4 bits) based on the original data of each pixel having the fixed number of bits (12 bits); Extracting higher-order data from predetermined bits (4 bits) of the lower-order bit data with a width of
Generating management data for individually managing upper bit data subjected to EG lossless encoding) and lower bit data having a width of the extracted number of effective bits, and the reversible code. Storing the higher-order bit data subjected to encoding (JPEG lossless encoding), the lower-order bit data having a width of the extracted effective number of bits, and the generated management data in a storage medium (2). Perform the compression processing.

According to a second aspect of the present invention, in the image data compression method according to the first aspect, the number of effective bits is determined according to an effective bit width of original data for each pixel. Things. According to a third aspect of the present invention, in the image data compression method according to the first or second aspect, the effective bit width of each pixel is determined by a value corresponding to the detection accuracy of the image input unit. The upper predetermined number of bits is set to an upper bit number that has a strong tendency to cause a correlation with a neighboring pixel, and the value is obtained by empirical rules.

(Operation) According to the first aspect of the present invention, variable-length encoding with high compression of high-order bit data having high correlation with neighboring pixels is possible for original data obtained with high definition. For lower-order bit data with low correlation, data with a width outside the range of the effective number of bits is truncated to obtain a higher compression ratio without deteriorating the image quality of the restored pixels and perform high-speed processing. Can do it.

According to the second aspect of the present invention, when the effective bit width (for example, 9 bits) of the original data is included in all or a part of the lower-order bit data, the whole or a part of the lower-order bit data is included. Since only image data is recorded, efficient storage of image data is performed. According to the third aspect of the present invention, the number of bits of the upper bit data to be subjected to lossless encoding is determined by a rule of thumb to a value having a high correlation, so that DPCM encoding or the like achieves a high compression ratio. Image data can be compressed.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS (First Embodiment) A first embodiment of the present invention will be described below with reference to the accompanying drawings. Note that the first embodiment corresponds to claims 1 to 3. FIG. 1 is a block diagram showing a configuration of an encoding processing device 10 to which the image data compression method of the present invention is applied.

The encoding processing device 10 includes an image input device (C
After the image signal input from the CD 1 is digitized, the image data is compressed and stored in the storage medium 2 by JPEG lossless encoding using both DPCM encoding and Huffman encoding. . Here, the encoding processing device 10 is provided integrally with the CCD 1 (for example, in a digital camera). Also, the compressed image data stored in the storage medium 2 is decoded by a decoding processing device 20 (FIG. 6) such as a personal computer, as described later, and then processed / corrected by the user, The image is restored.

The CCD 1 connected to the input side of the encoding processor 10 has pixels (cells) arranged in a matrix of m rows.times.n columns. As shown in FIG.
Color filters (in the illustrated example, “R”, “G”, and “B”) are arranged. The encoding processing device 10 includes a CCD 1
A / D converter 11 for converting a signal (signal charge) from the A / D into a digital signal (12 bits) of 4096 gradations, an input data buffer 12 for temporarily storing a digitized signal (original data), and an input data CPU 1 that encodes a signal (original data) temporarily stored in buffer 12
3. A main memory 14 for storing a program executed by the CPU 13, etc., for temporarily storing a signal (compressed image data) encoded by the CPU 13, outputting the signal to the storage medium 2 at a predetermined timing, and storing the signal. And the like.

The input data buffer 12 includes a CC
Pixel data for each pixel obtained by D1 (m × n)
Is stored for each row (line), and is shifted from the target line buffer 12A and the target line buffer 12A which collectively store data (n pixel data) for one row to be encoded. The buffer for one line before 12B that stores the data of one line before that is stored and the buffer 12C for two lines before that stores the data of two lines before shifted from the buffer for one line 12B.

The output data buffer 15 includes an address buffer 15A, a high buffer 15B, and a low buffer 15C, of which "compressed data block position information" described later in the address buffer 15A and "compressed data block position information" described later in the high buffer 15B. “Upper compressed data” and “lower data block position information” and “lower data” are stored in the row buffer 15C, respectively. By assigning 2 bytes or 4 bytes to the address buffer 15A in advance as the position information of one line of image data, position information of an arbitrary line can be obtained more quickly.

Next, an encoding process of image data for one frame (screen) executed by the CPU 13 of the encoding processing device 10 will be described with reference to a program flowchart of FIG. When the encoding process is started (start), first, in step S1, n pieces of pixel data (12-bit original data) stored in the target line buffer 12A of the input data buffer 12 for each row are n. ,
The pixel data of the target pixel is read.

In step S2, the read pixel data (12 bits) of the target pixel is separated into upper 8 bits (upper bit data) and lower 4 bits (lower bit data). Separating the 12-bit pixel data (original data) into the upper 8 bits and the lower 4 bits in this way, in the image data obtained by 12 bits, generally, the data from the upper 6 bits to the upper 8 bits is used. Has a high correlation with other neighboring pixel data, and can be encoded at a high compression rate. On the other hand, when the lower 4 bits become, the lower 4 bits of other neighboring pixels Is significantly reduced.

In step S3, the upper 8 bits of data are reversibly encoded by JPEG lossless encoding, and the encoded data is stored in the high buffer 15B as "upper compressed data". In the first embodiment, the JPEG lossless encoding in step S3 is generally performed according to the following procedure using both DPCM encoding and Huffman encoding.

First, the predicted value of the pixel of interest is based on the pixel value (value of higher-order bit data) of a neighboring pixel of the same color filter (same color pixel) or the pixel value of an adjacent pixel (value of higher-order bit data). Is calculated according to a predetermined prediction formula. Here, as the pixel value of the neighboring pixel, the same color pixel (R42 before two pixels) or the same color pixel (R24, R2 two lines before) on the same line with respect to the target pixel (for example, R44 in FIG. 2).
2) or the pixel value of the adjacent pixel (G43, B33, G34) with the smallest prediction error is used.

Which pixel value is used for calculating the actual predicted value, in other words, which prediction value is used as a variable (optimal prediction formula), specifically, two pixels before And the pixel value of the same color pixel in the vicinity thereof or the neighboring pixel two pixels before (all the values of the upper bit data). That is, a plurality of prediction formulas using the pixel value two pixels before the pixel of interest and the pixel value of the same color pixel or the neighboring pixel value in the vicinity of the pixel value two pixels before the target pixel are prepared. Is calculated, and the pixel value two pixels before is compared with each of the plurality of temporary predicted values, and the prediction formula that minimizes the prediction error is stored as the optimal prediction formula.
Then, the predicted value of the target pixel in the current loop is calculated using the stored optimal prediction formula.

The predicted value of the upper bit data calculated in this way is compared with the pixel value of the upper bit data (8 bits) separated in step S2 to obtain a prediction error Δ (DPCM Coding). The prediction error Δ is subjected to Huffman encoding according to the occurrence distribution, and the upper bit data is encoded (lossless encoding, variable length encoding).

Here, in calculating the prediction value, a prediction formula using the pixel value of the same color pixel in the same row and the pixel value of the same color pixel two rows before is used as the same color pixel. 2
As shown in the figure, in the case of a three-primary color CCD, for “R” and “B”, the pixel in which the color filter of the same color component is arranged does not exist one line before (one line before), and This is because there is a neighboring pixel of the same color before the row.

After the upper bit data is encoded in step S3, the effective bit number (1 to 4 bits) of the lower bit data (4 bits) is changed to the original data in the next step S4. It is determined according to the size (number of digits). Here, the number of effective bits is the number of bits of lower-order bit data (four bits) that are to be stored (to be used for restoring image data) and counted from the upper side (in FIG. 4, (Sloping part to the right).

In the first embodiment, the effective bit number of the lower-order bit data is such that the effective bit width of the 12-bit original data is, in principle, 9 bits (FIGS. 4A to 4D). )), Its value is required.
In this embodiment, the effective bit width is CC
It is determined according to the D detection accuracy. That is, when the size of the original 12-bit data is 12 digits, one bit that cannot be represented by the higher-order bit data (8 bits) is represented by the lower-order bit data, so that the number of effective bits is 1 bit. (A hatched portion falling rightward in FIG. 4A). When the size of the original data is 11 digits, the upper bit data (8
Since two bits that cannot be represented by the lower 7 bits of the bits are represented by the lower bit data, the number of effective bits is 2 bits (the lower right hatched portion in FIG. 4B). Similarly, when the size of the original data is 10 digits, the number of effective bits cannot be represented because 3 bits cannot be represented, and when the size of the original data is 9 digits, the number of valid bits cannot be represented. 4 bits (FIG. 4
(C), (d), a hatched portion falling to the right.

When the size of the original data is between 4 bits and 8 bits, 9 bits of effective bit width cannot be ensured. Therefore, the effective bit width of the original data depends on the size (number of digits) of the original data. At this time, the number of effective bits of the lower bit data is 4 bits (FIG. 4).
(E), (f), the hatched portion falling to the right. Further, when the size of the original data is smaller than 4 bits, the number of effective bits is fixed to 4 bits (a lower right hatched portion in FIG. 4G).

According to the effective bit width of the original data, the number of bits (valid bits) of the lower 4 bits of the data to be stored and the number of invalid bits (see FIG. 4, the lower left diagonal line) is determined. The data having the width of the effective bit number of the lower bit data is stored in the lower data row buffer 15C.

In step S5, step S1 described above is performed.
It is determined whether or not the processing of Step S4 has been performed for one row, that is, all of the n pixels included in the current target row, and when the processing for one row has been completed ( The determination result is “Yes”, and the process proceeds to the next step S6.
In step S6, the upper compressed data obtained for one row (n pixels) is blocked as an upper compressed data block, and the lower data is blocked as a lower data block and combined with each other.

In step S7, the position information of the lower data block in the same row is added to the head position of the upper compressed data block. In step S8, position information indicating the head position of the compressed data block of each row is added to the address buffer 15A at the earliest position of one frame of image data. In this manner, the “upper side compressed data” and “lower side data” for one row to which “compressed data block position information” and “lower side data block position information” stored in the address buffer 15A are: It is stored in the storage medium 2 at a predetermined timing.

When the encoding of the image data for one row which is the subject of this time is completed, in the next step S9, the processing of the above-described steps S1 to S8 is performed in the m rows constituting one frame (one screen). It is determined whether or not all the operations have been performed. If the processing for one frame (processing for m rows) has not been completed yet, the determination result of step S9 is “No”.
Then, the above-described steps S1 to S8 are repeatedly executed.

On the other hand, if the result of the determination in step S9 is "Ye
In the case of s ", that is, when the processing for all rows (m rows) of one frame is completed, the present program is terminated as it is. As described above, the image data obtained with 12 bits per pixel is used. The reason why the effective bit width of the original data is set to 9 bits in principle is that the output signal of the CCD 1 represented by 12 bits generally includes a noise inherent in the CCD 1 and a signal from the CCD 1 to the encoding processing device 10. This is because it is known from empirical rules that a noise component occurs about 3 bits in a signal having a saturation level represented by 12 bits.

Table 1 shows the relationship among the original data of 12 bits, the effective bit width of the original data, the bit position (effective bit position) of the signal to be restored, and the number of effective bits of the lower bit data. Things. In Table 1, "1" indicates the head position of the effective component, "x" indicates a valid bit, and "y" indicates an invalid bit. The effective bit position is Bi from the upper side.
It is represented by t1 to Bit12.

[Table 1] For example, when the original data has 12 digits, the effective bit width is 9 bits, and the effective bit positions are Bit1 to Bit9. Therefore, only the upper 1 bit (Bit 9) of the lower bit data becomes a valid bit “x” and the remaining lower 3 bits (Bit 9)
10 to Bit 12) are invalid bits “y”. Next, the data format (FIG. 5) of the compressed image data (for one frame) encoded by the above-described image encoding process (FIG. 3) will be described.

As shown in this figure, the image data for one frame has "compressed data block position information" indicating the position of the compressed data block for each line of one frame (m lines) at the earliest position. After that, the image data block for each row is stored in a repeating pattern from the first row to the m-th row. In the image data block of each row, the “lower side data block position information” of each row is provided at each head position.
Following this, one row of “upper side compressed data block” and one row of “lower side data block” are sequentially stored.

Here, the "compressed data block position information" provided at the earliest position of one frame of image data is the position information of the image data block of each row of one frame (m lines) of image data. 4 in FIG. 4). The “lower data block position information” provided at the head position of the image data block of each row indicates the position information (↑ in FIG. 4) of the “lower data block” within the image data block of each row. It is.

More specifically, as shown in a partially enlarged manner in FIG. 5 (the image data block in the second row is enlarged in FIG. 5), the “lower side data block position information” in the second row is In the subsequent “upper side compressed data block”, “upper side compressed data” of all the pixels (for n pixels) included in the second row are divided into blocks. In the “lower side compressed data block” following the “upper side compressed data block”, the “lower side data” of all the pixels (for n pixels) included in the second row
Is blocked.

As described above, the "compressed data block position information" for each line is stored at the earliest position of the image data for one frame, so that the "higher side""Compresseddata" can be read at random. Here, the “compressed data block position” is
This indicates the position of the image data block in each row at the same time as indicating the head position of the “upper side compressed data block”. Therefore, it is also possible to randomly read out both “upper side compressed data” and “lower side data” of a specific row.

As described above, the "lower data block position information" is added to the head position of the image data block of each row because the length of the "higher compressed data block" is as described above. This is because the data length is variable-length coded by JPEG lossless coding and the data length is different for each row. FIG. 6 shows another example of a data format when the “upper side compressed data block” and the “lower side data block” are stored in the storage medium 2.

In the data format shown in FIG.
In compressing and storing image data for one frame, “compressed data block position information” and “lower-order data block position information” are alternately stored for each row at the earliest position. Is different from the data format shown in FIG.

As described above, the "compressed data block position information" and the "lower data block position information" are alternately stored in the foremost part of the image data for one screen, so that only the image data of a specific row is stored. It is not necessary to read the image data of another row when extracting the image data, and the processing speed is increased. Next, decoding of image data encoded according to the above-described procedure will be described.

FIG. 7 is a block diagram showing a decoding processing device 20 for decoding the compressed image data stored in the storage medium 2. The decoding processing device 20 decodes the compressed image data from the storage medium 2 and outputs the decoded image data (12-bit original data) to the image processing / correction device 30. Input data buffer 21 for reading compressed image data, storage medium 2
CPU for restoring the compressed image data from the CPU, a restored data buffer 23 for temporarily storing and outputting the restored image data (original data) for each row, and a main memory 24 for storing programs executed by the CPU 22, etc. And is constituted by.

An image processing / correction device 30 is connected to the decoded data buffer 23 of the decoding processing device 20. The image processing / correction device 30 restores image data (1) based on processing / correction information (information input by a user's keyboard operation or the like) input from outside.
2 bits (original data) is processed / corrected.
The processed / modified image data (for example, 8-bit image data) is output to an image output device (for example, a CRT or a printer) 40 so that a desired image can be obtained. When the above-described decoding device 10 is used in a digital camera, the decoding device 2
0, the image processing / correcting device 30 is mainly configured by a personal computer.

Here, the input data buffer 21 of the decoding processor 20 comprises an address buffer 21A, a high buffer 21B, and a low buffer 21C. The “compressed data block position information” from the storage medium 2 is temporarily stored in the address buffer 21A, and the “upper side compressed data” and “lower side data block position information” from the storage medium 2 are stored in the high buffer 21B. Stored temporarily. The “lower data” from the storage medium 2 is temporarily stored in the row buffer 21C.

The restored data buffer 23 of the decoding processor 20 constitutes a buffer 23A for the target line, a buffer 23B for the previous line, and a buffer 23C for the previous line. Then, the restored image data (12-bit original data) of the current line to be restored (restoration target line) is temporarily stored in the target line buffer 23C. The one-line preceding buffer 23B temporarily stores the one-line preceding image data (12-bit original data) shifted from the target line buffer 23A, and the one-line preceding buffer 23C stores the one-line preceding buffer 23C. 23
The image data (the original data of 12 bits) two rows before shifted from B is temporarily stored.

FIG. 8 is a diagram showing the C of the above-described decoding processing device 20.
It is a program flowchart which shows the image decoding process performed by PU22. When the image decoding program is started, first, in step S21,
The position information (compressed data block position information, lower data block position information) taken into the input data buffer 21 from the storage medium 2 is read by the CPU 22.

In step S22, based on the read “compressed data block position information (information indicating the position of ↓ in FIG. 4)”, the image data block of the line (one line) targeted in the current loop The reading of all data in is performed. In step S23, step S2
Step S2 based on the “lower-side data block position information (information indicating the position of ↑ in FIG. 4)” read in step S1.
A process is performed to separate the data read in 2 into an upper compressed data block and a lower data block.

In step S24, well-known JPEG lossless decoding is performed on the upper compressed data stored in the separated upper compressed data block. The JPEG lossless decoding is performed in a procedure reverse to the coding procedure performed in step S3 of the above-described image coding processing (FIG. 3). That is, when decoding the upper bit data (8 bits) of the target pixel, first, the pixel value of the pixel two pixels before (same color pixel) decoded two loops before is obtained. Next, a plurality of predicted values (temporary predicted values) are obtained for the pixel two pixels before, based on a plurality of prediction formulas using pixel values of neighboring pixels of the same color or neighboring pixels. Further, the pixel value two pixels before and the tentative predicted values are respectively compared to obtain a prediction error Δ,
A prediction formula (optimal prediction formula) that minimizes this is selected in advance. Then, using the optimal prediction formula stored two loops ago, the prediction value of the target pixel in the current loop is obtained, and the upper bit data of the target pixel (from the stored prediction error Δ and this prediction value) (8 bits).

In step S25, the process of restoring the lower-order bit data stored with the width of the effective number of bits in step S4 of the above-described image encoding process (FIG. 3) to 4-bit data is performed. At this time, the lower 4-bit data to be restored using the lower bit data stored at the time of encoding is filled in from the upper side. The remaining portion truncated as invalid bits during encoding may be randomly filled with “0” and “1” in order to restore the noise portion, or may be a number that can be represented by the number of bits of the remaining portion. May be filled with a number that is half of the number. In this way, the lower bit data can be restored.

In step S26, the above-mentioned step S24
The 12-bit image data (original data) is obtained from the high-order bit data (8 bits) restored in step S25 and the low-order bit data (4 bits) restored in step S25. In step S27, the above-described step S
It is determined whether or not the processing from step S24 to step S26 has been performed for one row, that is, all the n pixels included in the same row. When the decoding processing for one row is completed, the process proceeds to step S28. move on.

In step S28, the above-described step S
It is determined whether or not the processing from step 22 to step S27 has been performed for all rows (m rows) constituting one frame (one screen). If the processing for one frame has not been completed yet, Returning to step S22, the process is repeated.
When the processing has been completed for all the rows (m rows) of one frame (the determination result in step S28 is "Yes"), the program is terminated as it is (END). In this image decoding process, in order to perform the process efficiently, as long as the determination result in step S27 is “No”, step S27 is performed.
Returning to 24, while the determination result of step S28 is "No", the process returns to step S22 to repeat the process. However, as shown in FIG. 9, the determination result of step S27 is "No". If there is, or step S28
When the determination result is "No", the processing may be repeated from the reading of the position information in step S21.

Next, the storage medium 2 for the compressed image data
A modified example of the data format at the time of storage in the storage device will be described with reference to FIGS. In this modification, the upper compressed data block and the lower data block are stored in the storage medium 2 and, for one frame of image data, the upper compressed data blocks of all rows (m rows)
The lower-order data blocks of all rows (m rows) are stored separately.

FIG. 10 shows a data format in this case. In this modification, as shown in the figure, both "compressed data block position information" and "lower data block position information" are stored at the earliest position of one frame of image data.
After that, “upper data blocks” are stored collectively for m rows, and then “lower data blocks” are stored collectively for m rows.

As described above, m rows of “upper data blocks” are collected, and m rows of “lower data blocks” are collected and stored in the storage medium 2 in a separated state.
At the time of reproducing the image data, for example, when decoding and outputting only the high-order 8-bit data (when reproducing only the coarse 8-bit image), the “compressed data block position information (FIG. Information on the position indicated by the middle ↓) "
, Only the “higher-side compressed data” needs to be decoded, and the processing speed when reproducing an image can be increased.

A specific procedure for decoding only the upper compressed data and reproducing a coarse image (upper 8 bits) at a high speed will be described with reference to a program flowchart shown in FIG. This program is executed by the CPU 22 of the decryption processing device 20 (FIG. 6). First, in step S31, "compressed data block position information" is read.

In step S32, the data (upper-side compressed data) in the upper-side compressed data block for one row is read using “compressed data block position information (position information indicated by ↓ in FIG. 10)”. Will be In step S33, decoding of the read high-order side compressed data for one pixel is performed by JPEG lossless decoding in a procedure reverse to the JPEG lossless coding executed in the encoding process. The decoding in step S33 is performed in the same procedure as in step S24 in FIG.

In step S34, the above-described step S
It is determined whether or not the processing of 33 is performed for one row, that is, n pixels included in the same row. When the processing of one row is completed, the process proceeds to the next step S35. In step S35, steps S32 to S34 described above are performed.
All lines (m lines) that make up one frame (one screen)
Is determined, and if the processing for one frame has not yet been completed, the above-described step S3
2 and the process is repeated.

On the other hand, when it is determined in step S35 that the decoding process has been completed for all the lines of one frame, the present program is terminated as it is (END). According to this modified example, when restoring the image data from the storage medium 2 in which the high-definition original data (RAW data) represented by 12 bits per pixel is compressed and stored, the higher-order highly correlated data is obtained. By restoring only 8 bits, a coarse image can be reproduced at high speed. In this image decoding process, in order to perform the process efficiently, the process returns to step S33 while the determination result in step S34 is “No”, and the determination result in step S35 is “No”.
Is returned to step S32 and the process is repeated. However, as shown in FIG.
No. 4 is “No” or in step S
When the determination result of 35 is “No”, the processing may be repeated from step S31.

According to the image data encoding and decoding methods of the first embodiment described above, the high-order 8 bits of the original data of 12 bits, which have a high correlation with the neighboring pixels, are obtained. JPEG lossless encoding with high compression for bits (high-order bit data) becomes possible,
For lower-order bit data having low correlation, high-speed processing can be performed by truncating data out of the range of the number of effective bits. As a result, the processing speed of the entire 12-bit original data can be increased while securing a high compression rate.

Further, since the effective bit width for the entire 12-bit original data is set to 9 bits, the data of the 9-bit effective bit width can be obtained by using all or a part of the lower bit data without fail. Since encoding / decoding is performed, efficient storage of image data is performed.

(Second Embodiment) Next, a second embodiment of the present invention will be described.
Will be described with reference to Tables 2 to 4.
In the second embodiment, the number of effective bits of the lower bit data (lower 4 bits) is determined to a value corresponding to the effective signal level of the saturation level of the CCD 1. The image data compression in the second embodiment is as follows.
It is executed by the same encoding processing device and decoding processing device as in the first embodiment. Table 2 shows that the saturation level is 10
It shows the relationship among the number of signal charges of the 10,000 (1.0 × 10 ⁵ ) electrons of the CCD 1, the signal value from the CCD 1, the effective signal level, and the noise component.

[Table 2] Now suppose that the number of signal charges at the saturation level is 1.0 × 10
In ⁵ CCD 1 of (electrons), seeking 1/2 square of its brightness (electrons), consider the case of this a valid signal level tone. The half of the number of signal charges 1.0 × 10 ⁵ (electrons) at the saturation level of the CCD 1 is about 316.2, and the effective signal level at the saturation level is represented by 316.2 gray levels. be able to.

Therefore, even if the output of the CCD 1 is at the saturation level, its brightness can be sufficiently represented by 9 bits (2 ⁹ = 512). In calculation, the number of signal charges at the saturation level is represented by 316.2 gradations, so that the lower gradation 12.95 can be processed as a noise component. Since this lower gradation 12.95 is larger than 3 bits (2 ³ = 8), at the saturation level, at least the lower 3 bits of the data represented by 12 bits may be discarded (ignored). it can.

Here, at the saturation level (316.2 gradations), the number of lower bits that can be removed as a noise component is 1
Focusing on 2.95 gradations, as a noise component,
The magnitude (signal value S ₁ ) of a signal that can make just three bits a noise component is calculated. When the noise component has just 3 bits, the saturation level (1.0 × 10 ⁵
If it is considered that a signal charge number (X ₁ × 10 ⁵ electrons) having a ratio of X ₁ (0 <X ₁ <1) is generated with respect to the electrons, the effective signal level is (X ₁ × 10 ⁵ ). ^{1 /}
It is represented by ^two tones.

As described above, the noise component at the saturation level represented by 12 bits (at the time of 4096 gradations) is “1”.
2.95 (gradation) ". Here, the noise component is “8
(Gradation) ", the signal value S ₁ represented by 12 bits is set to“ 4096 × X ₁ ”. Also, the number of signal charges (X ₁ × 10
The gradation of the effective signal level of ( ⁵ electrons) is (X ₁ × 10
⁵ ) Since it is ^1/2 , “X ₁ × 4096” matches a value obtained by multiplying “2 ³ ” by (X ₁ × 10 ⁵ ) ^1/2 from a linear relationship.

X ₁ × 4096 = 8 × (X ₁ × 10 ⁵ ) ^1/2 ... [1] From the above formula [1], X ₁ ≒ 0.38147 is obtained.
Therefore, the signal value S _{1 at which} the noise component is exactly “8 (gradation)” is “4096 × X ₁ ６３1563”.

Similarly, the noise component is just “4 (gradation)”.
When a signal value S ₂ corresponding to the number of signal charges is obtained,
Since “X ₂ × 4096” matches the value obtained by multiplying “4” by (X ₂ × 10 ⁵ ) ^1/2 , Equation [2] is established from the linear relationship. X ₂ × 4096 = 4 × (X ₂ × 10 ⁵ ) ^1/2 ... [2] From the above formula [2], X ₂上 記 0.09537 is obtained.

Therefore, the signal value S _{2 at} which the noise component becomes exactly 4 (gradation) is “4096 × X ₂ ≒ 391”. Similarly, when the signal value S ₃ corresponding to the number of signal charges at which the noise component is 2 (gradation) is obtained, “X ₃ × 4096” becomes “2”.
And (X ₃ × 10 ⁵ ) ^1/2 , which should be equal to the value.

X ₃ × 4096 = 2 × (X ₃ × 10 ⁵ ) ^1/2 [3] According to the above formula [3], X ₃ ≒ 0.02384.
The signal value S _{3 at} which the noise component is exactly 2 (gradation) is “40
96 × X ₃ = 98 ”. Table 3 shows an example in which the thus obtained signal values S ₁ , S ₂ , and S ₃ are used as threshold values.

[Table 3] In Table 3, "S ₁ (= 1563)" and "S ₂ =
(391) ”instead of“ S ₃ (= 98) ”
2 ”,“ 384 ”and“ 96 ”are used as thresholds,
A fourth range is set (“4095 to 1552”).
(1st range) "" 1551-384 (2nd range) "
"383-96 (third range)""95-0 (fourth range)".

Here, “S ₁ (= 1563)” and “S ₂ =
(391) ”instead of“ S ₃ (= 98) ”
The thresholds of “2”, “384”, and “96” are used for the following reasons. That is, in the second embodiment, 12-bit original data is separated into upper 8 bits (upper bit data) and lower 4 bits (lower bit data),
The upper 8 bits are JPEG lossless-encoded / decoded in the same procedure as in the embodiment. In order to recognize the number of effective bits of the lower bit data, only the upper bit data is used.

Here, “S ₁ (= 1563)” is “1,1,0,0,0,0,1,1,0,1,1” and “S ₂ (= 39) in the binary system.
“1)” is “1,1,0,0,0,0,1,1,1”, and “S ₃ (= 98)” is “1,1,0,0,0,1,0” . Therefore, focusing on only the upper 8 bits of these, and regarding the lower 4 bits, assuming that the value is “0,0,0,0”, “1,1,0,0,0,0, 1,0,0,
“0,0” = 1552, “1,1,0,0,0,0,0,0,0” = 384,
“1,1,0,0,0,0,0” = 96 is used as the threshold value.

By the way, when the signal value (12-bit data) indicating the number of signal charges acquired by the CCD 1 is within the above-mentioned first range, when Bit 1 is “1”, when Bit 1 is “0”, There are times. Therefore, when Bit1 is "1", 9 bits, when Bit1 is "0", 8 bits,
Each becomes the effective bit width of the original data. At this time, the number of effective bits of the lower bit data is 1 bit, and 3 bits of the lower 4 bits are discarded as noise components.

Similarly, when the signal value indicating the number of signal charges acquired by the CCD 1 is within the above-mentioned second range, Bit 1 is “0”, Bit 2 is “1”, and Bit 1 and Bit 2
Are both “0”. Therefore, the effective bit width of the original data is 8 or 7, the number of effective bits of the lower bit data is 2 bits (upper 2 bits), and the remaining 2
Bits are truncated as noise components.

When the signal value indicating the number of signal charges is within the third range, the effective bit width of the original data is 7 or 6, and the effective bit width of the lower bit data (4 bits) is The numbers are both 3 bits, and the upper 3 bits are stored as valid.

When the signal value indicating the number of signal charges is within the above-described fourth range, the effective bit width of the original data is a value of 6-1. The number of bits is 4 bits, that is, the lower bit data is all treated as valid. Table 4 shows the method of determining the number of effective bits of the lower-order bit data from the higher-order bit data (8 bits) of the 12-bit original data, which is processed in hexadecimal.

[Table 4] In Table 4, "2748""812""156""76"
In hexadecimal notation (“2748 = 0, x, A,
B, C "" 812 = 0, x, 3,2, C "" 156 = 0, x, 9, C "" 7
6 = 0, x, 4, C ”), where“ 0, x ”in the fourth and fifth digits indicates hexadecimal notation. As described above, it is expressed in binary notation. In the original data (12 bits), the higher-order bit data (8 bits) determines how many bits of the lower-order bit data are to be handled as valid data. Digit,
4 bits) (“0, x, A, B” “0, x, 3,2” “0, x,
9 ”“ 0, x, 4 ”), how to handle the first digit is determined.

The above example "2748 = 0, x, A, B, C""8
12 = 0, x, 3,2, C "" 156 = 0, x, 9, C "" 76 = 0, x, 4,
In “C”, the first digit value when expressed in hexadecimal is “C” = 0, b, 1,1,0,0 ”(“ 0, b ”indicates that the value is expressed in binary. ). Therefore, how many significant bits of the binary number “1,1,0,0” are valid data is determined by the second and third digits or the second digit of the hexadecimal value. ("A, B""3,2""9""4").

(Third Embodiment) Next, a third embodiment of the present invention will be described.
The embodiment will be described with reference to FIG. 13 and Table 5. In the third embodiment, the number of effective bits of lower-order bit data (4 bits) of the original 12-bit data is
Based on the signal level (the number of signal charges) from the CCD 1,
It is determined for each signal. Note that the image data compression in the third embodiment is also executed by the same encoding device and decoding device as in the first embodiment.

FIGS. 13A to 13H show the original data of 12 bits, the upper bit data (8 bits), the lower bit data (4 bits), the size (number of digits) of the original data,
This shows the relationship between the effective bit widths. Table 5
Shows the relationship between the original data of 12 bits, the effective bit width of the signal, and the effective bit position. In Table 5, "1"
Indicates the start position of the effective component, “x” indicates the effective bit,
“Y” indicates an invalid bit. The effective bit position is represented by Bit 1 to Bit 12 from the upper side.

[Table 5] As shown in Table 5, the size of the 12-bit original data is 1
In the case of two digits (Bit 1 to Bit 12), the size of the data that can be displayed with these 12 digits is “4095 (gradation)”, which is 1.0 × 10 ⁵ (electrons) because of the saturation level at this time. The effective signal level is (1.0 × 10 ⁵ )
^1/2 ≒ 316.2 (gradation) (FIG. 13A). Since this gradation can be displayed by 9 bits, the upper 9 bits from Bit 1 to Bit 9 of the original 12-bit data can be set as an effective bit width, and the remaining 3 bits from Bit 10 to Bit 12 can be set as invalid bits. . Therefore, for the lower bit data (Bit9 to Bit12), only the upper one bit (Bit9) becomes a valid bit (the number of valid bits = 1 bit).

Hereinafter, the size of the 12-bit original data is 1
Even in the case of one digit or less, a noise component capable of truncating data is obtained from the effective signal level, and based on this, the effective bit width of the original data and the number of effective bits of the lower bit data are obtained as follows. Become. That is, when the size of the 12-bit original data is 11 digits (Bit 2 to Bit 12), 11
The size of the data that can be displayed by the digit is “2047”, which is の of the saturation level, and is therefore 0.5 × 10 ⁵ (electrons).
⁵ ) ^1/2 ≒ 223.6 (gradation) (FIG. 13B).
Since this gray scale can be displayed in 8 bits, the effective bit width is set to 8 bits from Bit 2 to Bit 9 of 12 bits, and Bi
The remaining three bits from t10 to Bit12 can be set as invalid bits. Therefore, also in this case, with regard to the lower bit data (Bit9 to Bit12), only the upper one bit (Bit9) becomes a valid bit (the number of valid bits = 1 bit).

The size of the original 12-bit data is 10
In the case of digits (Bit3 to Bit12), the size of data that can be displayed with 10 digits is “1023”, which is 1/1 of the saturation level.
Since it is 4, 0.25 × 10 ⁵ (electrons), the effective signal level at this time is (0.25 × 10 ⁵ ) ^1/2 ≒ 15
8.1 (gradation) (FIG. 13C). This gradation also
Since it can be displayed in 8 bits, 8 bits from Bit 3 to Bit 10 of the 12-bit original data are set as the effective bit width, and
The remaining two bits from 11 to Bit 12 can be set as invalid bits. Therefore, in this case, with respect to the lower bit data (Bit9 to Bit12), the upper two bits (Bit9, Bit9)
10) becomes effective bits (the number of effective bits = 2 bits).

Similarly, the original data is 9 digits (Bit 4 to Bit 12)
In the case of (FIG. 13D), since the effective signal level can be displayed by 7 bits, Bit 4 of the 12-bit original data
7 bits from Bit 10 to Bit 10 are the effective bit width, and Bit11, Bi
Two bits of t12 can be set as invalid bits. Therefore, also in this case, the lower bit data (Bit9 to Bit1)
Regarding 2), the upper two bits (Bit9, Bit10) are effective bits (the number of effective bits = 2 bits).

When the original data has 8 digits (Bit 5 to Bit 12) (FIG. 13 (e)), the effective signal level can be displayed with 7 bits, so that Bit 5 to Bi of the 12-bit original data can be displayed.
7 bits up to t11 can be set as an effective bit width, and only Bit 12 can be set as an invalid bit. Therefore, in this case,
As for the lower bit data (Bit9 to Bit12), the upper three bits (Bit9 to Bit11) are effective bits (the number of effective bits = 3 bits).

When the original data has 7 digits (Bit 6 to Bit 12) (FIG. 13 (f)), the effective signal level can be displayed with 6 bits, so that the 6-bit Bi of the 12-bit original data can be displayed.
Six bits up to t11 can be set as an effective bit width, and only Bit 12 can be set as an invalid bit. Therefore, also in this case,
As for the lower bit data (Bit9 to Bit12), the upper three bits (Bit9 to Bit11) are effective bits (the number of effective bits = 3 bits).

Further, when the original data has 6 digits or less, the valid signal level can be displayed if it has at least 6 bits. At this time, all 12 digits are valid bits. Therefore, in this case, the lower bit data (Bit9 to Bit1)
Regarding 2), all 4 bits (Bit 9 to Bit 12) are valid bits (the number of valid bits = 4 bits). As described above, according to the third embodiment, the number of effective bits of the lower-order bit data is determined according to each effective signal level, so that an efficient image corresponding to the signal value is obtained. Data compression is performed.

In the first to third embodiments, the case where the original data (RAW data) of an image is 12 bits has been described. However, image data larger than 12 bits and image data smaller than 12 bits are described. In any case, the present invention is applicable, and in this case, the processing speed can be increased while securing a high compression ratio.
In the first to third embodiments described above, an example is described in which the 12-bit original data (RAW data) is separated into upper 8 bits and lower 4 bits. In the case of 12-bit image data, it is generally empirically known that the range is from 6 bits to 9 bits.), However, the number of bits other than 8 bits can be set. In this case, the number of upper bits to be used as upper bit data may be determined for each frame (one screen).

In the first to third embodiments, an example has been described in which the prediction error Δ (or the quantization prediction error δ) is encoded using a Huffman encoding table or the like. Error Δ (or quantization prediction error δ) is calculated as Ziv-Lemp
It may be encoded by universal encoding such as el. In the above-described first to third embodiments, the position information (data block position information) is added to the compressed image data in consideration of partially reproducing the image of one frame. If you always want to completely restore one frame,
Location information can also be omitted.

In each of the first to third embodiments described above, an example in which the encoding processing device and the decoding processing device are separately configured (for example, the encoding processing device is a digital camera side, Although the decoding device has been described on the personal computer side, these two devices may be incorporated in one device (for example, both may be incorporated in a digital camera or the like).

In the first to third embodiments, the process of separating the image data into the upper side and the lower side, and the encoding process by the DPCM encoding of the upper bit data,
In both cases, an example has been described in which the processing is performed by the CPU 13 in the encoding processing apparatus 10. However, the present invention is not limited to this, and the encoding by the dedicated LSI integrated on an external chip is performed by performing the DPCM encoding. The CPU 13 in 10 may be configured to separate only the upper side and the lower side. In this case, the external dedicated LSI is replaced with DPCM encoding.
An LSI that performs encoding by Ziv-Lempel or the like may be used. Also, for decoding, a chip in which dedicated LSIs are similarly integrated is externally attached to the decoding processing device 20, and DPCM decoding of higher-order bit data is performed by a dedicated LSI in the chip (or by Ziv-Lempel or the like). (Decoding). In this case, the CPU 22 of the decoding processing device 20 may combine the upper bit data and the lower bit data after decoding.

In the case where the above-described first to third embodiments are applied to a digital camera, based on the gain of an amplifier that amplifies a signal output from a CCD,
The noise amount, which is an invalid bit portion of the lower bit data, may be obtained by calculation. Furthermore, the invalid bit portion of the lower-order bit data may be obtained based on the ISO sensitivity setting that is the basis for determining the gain of the amplifier in the digital camera. In digital cameras, CCD
Since the amount of noise changes according to the gain of the amplifier that amplifies the output signal, the effective bit length may be changed according to the gain of this amplifier or the ISO sensitivity. As a result, a signal having an appropriate effective bit length can be obtained regardless of the sensitivity setting and gain setting in the digital camera, and as a result, the compression efficiency can be increased.

Further, in the above-described first to third embodiments, the effective bit length is described as the bit length determined based on the characteristics of the input device. Even if a bit length smaller than a bit is used as an effective bit, there is almost no influence on image quality. Therefore, the effective bit length may be changed based on the shooting conditions, the field of use, and the like.

Although the first to third embodiments have been described on the assumption that a still image is obtained by a digital camera or the like, the present invention is also applicable to the case of obtaining a moving image. It is.

[0091]

According to the first aspect of the present invention described above,
The image data obtained with high definition is separated into upper bit data and lower bit data, and the number of effective bits of the lower bit data is determined based on the original data before separation to determine the lower bit data. To extract the top data,
It is possible to perform compression processing excluding noise components included in image data, and to perform efficient compression.

According to the second aspect of the present invention, image data is compressed at a high compression rate while reliably storing valid data that is truly significant. According to the third aspect of the present invention, the width of the higher-order bit data having a strong correlation is obtained by an experiment or the like, so that the image data is compressed at a high compression rate according to each image. .

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an encoding processing device according to a first embodiment of this invention.

FIG. 2 is a diagram showing a layout of a color filter arranged on a light receiving surface of the image input device 2.

FIG. 3 is a flowchart illustrating an image encoding process according to the first embodiment.

FIG. 4 shows the original data, the upper bit data, the lower bit data, the size of the original data,
It is a figure showing each relation of effective bit width.

FIG. 5 is a diagram showing a first example of a format of compressed image data obtained according to the first embodiment.

FIG. 6 is a diagram illustrating a second example of the format of the compressed image data obtained according to the first embodiment.

FIG. 7 is a block diagram illustrating a configuration of a decoding processing device according to the first embodiment of this invention.

FIG. 8 is a flowchart illustrating an image decoding process according to the first embodiment.

9 is a flowchart showing another mode of the image decoding process in FIG.

FIG. 10 is a diagram showing a format of compressed image data in a modification of the first embodiment.

FIG. 11 is a flowchart illustrating another image decoding process.

12 is a flowchart showing another mode of the image decoding process in FIG.

FIG. 13 is a diagram illustrating a relationship among original data, upper bit data, lower bit data, the size of the original data, and effective bit width according to the second embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image input apparatus (CCD) 2 Storage medium 10 Encoding processing unit 12 Input data buffer 13 CPU 15 Output data buffer 20 Decoding processing unit 21 Input data buffer 22 CPU 23 Restoration data buffer

Claims (3)

[Claims] 1. An image data compression method for compressing image data obtained with a fixed number of bits per pixel by an image input means, wherein the original data for each pixel having the fixed number of bits is represented by a higher order predetermined number of bits. Separating upper-order bit data into lower-order bit data represented by a predetermined lower-order bit number; applying reversible encoding to the upper-order bit data; Determining the number of effective bits of the lower-order bit data based on data; extracting the higher-order data among predetermined bits of the lower-order bit data with a width of the effective bit number; Management data for individually managing the high-order bit data subjected to the conversion and the low-order bit data having the width of the extracted effective number of bits. And storing the high-order bit data subjected to the reversible encoding, the low-order bit data having a width of the extracted number of effective bits, and the generated management data in a storage medium. A method of compressing image data. 2. The image data compression method according to claim 1, wherein the number of effective bits is determined according to an effective bit width of original data for each pixel. 3. The image data compression method according to claim 1, wherein the effective bit width of each pixel is a value corresponding to a detection accuracy of the image input unit, and the upper predetermined number of bits is , A high-order bit number that has a strong tendency to cause a correlation with neighboring pixels, the value of which is obtained by an empirical rule.