JPH04312089A - Data compressor and its expander device - Google Patents

Abstract

PURPOSE:To perform the high quality compressing and expanding processing in which the character, etc., superimposed to it are not deteriorated even when the compression ratio of the image data for the background image is large. CONSTITUTION:A background image signal is compressed and processed by an image compressor 10, a superimposing image signal including a character and a mark is separately and independently compressed and processed in a superimposing image compressor 12. At this time, for the superimposing image signal, they are expressed by plural gradations, and concerning the maximum value and minimum value, the Huffman encipherment by the run-length and concerning the intermediate value, the encipherment by the code provided at the Huffman code system are performed. Thus, even when the compression ratio of the background image is large, the superimposing image signal is not completely influenced and the deterioration is reduced. The expansion is performed by the processing reverse to the compression.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data compression device and a decompression device suitable for image compression when, for example, characters or symbols are superimposed on a color moving image.

0002

2. Description of the Related Art Conventionally, when data compression processing is performed on an image signal of an image created by mixing a superimposed image signal containing characters and symbols with a background image, as shown in FIG. Before compression processing, characters and symbols are superimposed on a background image and mixed, and then compression processing is performed. In the figure, the background image signal and the superimposed image signal are input to a video mixing device 90, where mixing of the two is performed. As a result, characters and symbols are superimposed on the background image. The superimposed image signal is then supplied to an image compression device 92, where data compression processing is performed and a compressed image signal is output.

0003

However, in such prior art, superimposed image signals including characters and symbols are processed by the image compression device 92 in the same way as background image signals.
This has the disadvantage of causing deterioration. For example, when the data compression rate in the image compression device 92 is high, the image quality deteriorates accordingly when high frequency components are included. Superimposed image signals often contain high frequency components in edge portions of characters, etc., and therefore the deterioration is more pronounced than in background images.

The present invention has been made with attention to the above points, and even when the compression rate for the background image is high,
It is an object of the present invention to provide a good data compression device and its decompression device capable of compressing superimposed characters and the like with high quality.

[0005]

[Means for Solving the Problems] One aspect of the present invention is to provide a data compression device for compressing data of a background image signal and a superimposed image signal including characters and symbols. and a second data compression means that compresses the superimposed image signal reversibly or at a visually undetectable level independently of the background image signal. .

Another invention provides a data decompression device for decompressing image data compressed by the data compression device according to claim 1, further comprising a first data decompression means for decompressing a compressed background image signal; and second data decompression means for decompressing the superimposed image signal independently of the background image signal.

Still another aspect of the present invention provides a data compression device for compressing image data in consideration of run length, including data conversion means for expressing the image in a plurality of gradations; The maximum and minimum values of the gradation are encoded by run-length Huffman encoding, and the intermediate values of the gradation are encoded by assigning a code provided in the Huffman coding system to each level. It is characterized by having the following.

Still another aspect of the present invention provides a data compression device for compressing image data in consideration of run length, including data conversion means for expressing the image in a plurality of gradations, and image data obtained thereby. The maximum and minimum values of the gradation are subjected to Huffman encoding using run length, and the intermediate values of the gradation are assigned codes in the Huffman coding system, and the run length of the intermediate value and each level are and encoding means for performing encoding by adding a code shown.

[0009]

According to the present invention, the background image signal and the superimposed image signal containing characters and symbols are compressed or coded separately. At this time, the superimposed image signal is compressed using a reversible method or at a visually undetectable level. Therefore, even if the compression rate of the background image is high, the superimposed image signal is not affected at all, and its deterioration is effectively reduced.

On the other hand, image data such as characters is expressed in multiple gradations by data converting means and treated as a multivalued image. Then, the maximum and minimum values of the gradation are subjected to Huffman encoding using run length, and the intermediate values are encoded by being assigned a code provided in the Huffman coding system. Therefore, even when compared to data compression according to the G3 standard, the amount of code is approximately 1.5 to 2 times as much, and despite the high compression ratio, good quality is maintained due to the multi-level expression.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of a data compression device and its decompression device according to the present invention will be described with reference to the accompanying drawings. <Embodiment 1> First, Embodiment 1 regarding a compression/expansion apparatus system will be described. FIG. 1 shows the configuration of the data compression device. In the figure, a background image signal is input to an image compression device 10,
The superimposed image signal is input to a superimposed image compression device 12. The output sides of the image compression device 10 and the superimposed image compression device 12 are respectively connected to the input side of a multiplexer 14, and the image data after compression processing is outputted from the multiplexer 14 as a multiplexed bit stream. ing.

Among these, the background image compression device 10 is
Similar to the prior art described above, data compression is performed on the image signal of the background image at an appropriate ratio. As a compression method in the background image compression device 10, for example, JPE is used.
A compression method based on the international standardization of G.G. and MPEG is used, and a specific example is shown in FIG.

[0013] Also, a superimposed image compression device 1
In the background image compression apparatus 2, compression processing for the superimposed image signal is performed using a method different from that of the background image compression apparatus 10. For example, characters and symbols can be simply binarized and CCITT Recommendation T. A compression method using run-length encoding is used as described in "Standardization of Group 3 Facsimile Devices for Document Transmission" in Section 4. This method is a reversible method that does not cause deterioration due to compression, unlike the background image compression device 10 described above. Further, the multiplexer 14 multiplexes input compressed image data, and multiplexing as shown in FIG. 3 is performed, for example.

Next, the operation of the data compression apparatus configured as above will be explained. First, the background image signal is
The image is input to the image compression device 10, where compression processing is performed using the method shown in FIG. First, it is determined whether the input image is a still image (step SA in the figure). As a result, when it is determined that the image is a still image, the image data is divided into blocks each composed of, for example, 8×8 pixels (step SB). The image data of each block is subjected to, for example, discrete cosine orthogonal transformation and quantized (step SC). The quantized image data is further subjected to zero run length encoding and Huffman encoding (step SD).
.

On the other hand, when the input image is a moving image, multiple frames are treated as one segment, one of which is used as an intra-frame for intra-frame compression, and the other frames are used as intra-frames. The frame is made up of differential data from the motion-compensated frame (step SE). Thereafter, the image data of each frame is subjected to coding processing (steps SB, SC, SD) similar to that for still images.

Next, the superimposed image signal containing characters and symbols is supplied to a superimposed image compression device 12. Superimposed image compression device 1
In No. 2, for example, characters and symbols are simply binarized and compressed using the run-length encoding described above.

The output data of the background image compression device 10 and the superimposed image compression device 12 are each supplied to a multiplexer 14, where multiplexing processing is performed. Multiplexing is performed, for example, as shown in FIG. One sequence consists of N GOPs, as shown in A of the same figure.
(group of pictures). Note that "H" in the figure represents a header. Each G.O.P.
As shown in FIG. 1B, the image data is composed of M PACKs (data packets in units of one picture). Each PACK is composed of video data, audio data, superimposed character and symbol data, other control signals, and respective headers, as shown in FIG. As described above, the multiplexer 14 multiplexes compressed data and outputs a multiplexed bitstream.

Next, the decompression device side of the first embodiment will be explained with reference to FIG. In the figure, the compressed multiple bit stream obtained by the data compression apparatus of FIG. 1 is input to a demultiplexer 20. The output side of the background image signal of this demultiplexer 20 is connected to the input side of the image expansion device 22,
The output side of the superimposed image signal is connected to the input side of the superimposed image expansion device 24. The output sides of the image decompression device 22 and the superimposed image decompression device 24 are connected to the input side of a video mixing device 26.
An image signal for display is output from.

Next, the operation of the data decompression device configured as described above will be explained. The multiplexed bit stream data is demultiplexed by the demultiplexer 20. Specifically, P shown in FIG. 3C
By reading the header H in the ACK, the subsequent data is decomposed into video data, audio data, superimposed character data, and other control signal data.

Among the decomposed data, a bit stream of video data is input to the background image decompression device 22,
Here, it is expanded into a background image signal by a process that is the inverse of compression. Further, the bit stream of character data to be superimposed is input to a superimposed image decompression device 24, where it is decompressed into a superimposed image signal by a process inverse to compression. These background image signals and superimposed image signals are each input to a video mixing device 26, where mixing by superimposition is performed. That is, a mixing process is performed in which the image signal of the background image is output where no characters or symbols exist, and the image data of the characters or symbols is output where there are characters or symbols. As a result, a display image signal is obtained.

As described above, according to this embodiment, the superimposed image signal including characters and symbols is coded or compressed separately from the background image. At this time, the superimposed image signal is reversible,
Alternatively, compression processing is performed at a level that is not visually detectable. Therefore, even if the background image deteriorates due to compression, the superimposed image signal does not deteriorate. Therefore, even if the compression rate for the background image is high, the superimposed image signal is not affected at all, and characters and symbols can be superimposed satisfactorily.

<Example 2> Next, Example 2 of the present invention will be described. This example is an effective compression method for image data such as superimposed characters.
It gives one. There are methods for compressing image data that have been put to practical use in facsimiles and the like. According to this, binary image data is compressed by Huffman encoding the run length of white data and the run length of black data. As is well known, such technology has been clarified as "CCITT Recommendation T.4, Standardization of Group 3 facsimile equipment for electronic document transmission" and is generally referred to as the G3 standard.

By the way, if the superimposed image data only distinguishes between the background image and parts such as characters, 2
It can be expressed as value data. However, in such binarized data, lines that are slightly inclined from the horizontal (0° < α < 45°) shown in Figure 10 (A) and circles PA, PB, and PC shown in Figure 10 (B) , a step-like noise is generated in each part of the PD as shown in the figure. Superimposed image data that includes characters and symbols includes many parts that meet such conditions. For this reason, step-like noise in edge portions becomes considerably noticeable due to binarization. Therefore, in order to obtain a high-quality image, it is necessary to treat superimposed image data including characters and symbols as multivalued data.

In fact, if the superimposed image data is treated as multivalued data in advance, the above-mentioned step-like noise becomes visually inconspicuous. If it is 8-10 value data, 25 of the luminance signal of the video signal of the background image
It is so great that it can no longer be distinguished from 6 values (8 bits).

Therefore, the method of G3 is used independently for each bit plane of the superimposed image data,
This may be a possible way to perform compression. However, in this method, the amount of information increases by more than eight times compared to compression by simple binarization. On the other hand, there are limitations on the media when it comes to recording image data, and there is also a demand that superimposed image data must be recorded within these limitations.

The present embodiment takes this point into consideration, and is a data compression method that is suitable for compressing the superimposed image data in the first embodiment and has a high compression ratio and is capable of high quality compression. FIG. 5 shows the configuration of this embodiment. In the figure, superimposed image data of characters and symbols is input to a pre-processing section 110. The output side of this pre-processing section 110 is connected to the input side of a run-length counter 112, and the output side of the run-length counter 112 is connected to the input side of a run-length encoding section 114. The output side of this run-length encoding section 114 is connected to the input side of a Huffman encoding section 116, and a bit stream of compressed image data is output from this Huffman encoding section 116. It has become.

Among the above units, the pre-processing unit 110 is
For example, if the input superimposed image data is
It performs pre-processing to quantize into bits and 8 gradations. For example, in an image scanner or the like, an image is scanned and read in multiple gradations, and a similar operation is performed in the pre-processing section 110. Note that if the input superimposed image data is completely binarized, appropriate filtering is performed to artificially increase the gradation to 6 to 8 and quantization is performed. As gradation.

FIG. 6 shows 8 levels from the black level to the white level.
Six intermediate gradations included in the gradation are shown, and their levels are ESC1, ESC2, ESC3, ..., E
Each is represented by SC6.

Next, returning to FIG. 5, the run length counter 112 calculates the run length in the input data, that is, the extent to which white and black, which are the minimum and maximum values of the above-mentioned eight gradations, are continuous. It is for counting. Furthermore, the run-length encoding unit 114 performs encoding corresponding to the input run-length count value. Next, the Huffman encoding unit 116 performs Huffman encoding on white and black of the input data, and applies a predetermined code to the other halftone data represented by ESC. It is for assignment.

Next, the operation of the second embodiment configured as above will be explained. The superimposed image data is input to the pre-processing section 110, where the above-described pre-processing is performed. As a result, the superimposed image data is expressed in 3 bits and 8 gradations and output from the pre-processing section 110.

Here, suppose that the form of the character represented by the superimposed image data is the alphabet "A" as shown in FIG.
It is assumed that data scanned in the direction of is input to the pre-processing section 110. Note that the background is black and the characters are white. When scanning in the direction of the arrow FA, the sequence goes from the background part to the border part to the character part to the border part to the background part. If we look at this in pixel units, the image data will be arranged like this. Note that the transition of the intermediate level displayed by ESC at the boundary portion differs depending on the image. This image data includes the “black” part, “E
It is composed of three parts: the "SC" part and the "white" part.

Regarding the output of the pre-processing section 110, the black portion and the white portion are processed by the run-length counter 112, the run-length encoding section 114, and the Huffman encoding section 116 in the same manner as in the G3 standard described above. Run-length encoding and Huffman encoding are performed. The run lengths of the black and white parts are both "4", so
The image data string is: Black 4 → ESC1 → ESC3 → ESC6 → White 4 →
It can be displayed as ESC6 → ESC4 → ESC2 → Black 4. For the codes specifically assigned to black 4 and white 4, for example, the code tables shown in Table 1 and Table 2 are referred to.

[0033]

[Table 1]

[0034]

[Table 2]

The codes in these tables are Huffman codes that have been created in advance to be statistically efficient for various images, and short codes are assigned in order of run length with a high probability of occurrence. For example, the black 4 is encoded as "1011" with reference to "4" in the background run length from Table 1. Also, white 4
is encoded as "011" with reference to "4" in the character part run length from Table 1. The same applies to image data of other run lengths.

Next, for ESC1 to ESC6, which are halftones, codes are defined as shown in Table 3 as part of the Huffman code system shown in Tables 1 and 2, and encoding is performed using these codes. will be held.

[0037]

[Table 3]

For example, ESC1 is “0000000
1000”. The same applies to others. Such encoding is performed in the Huffman encoding unit 116,
A bitstream of code data will be output.

In the above example, the code data will be output in the following order. Note that the compression data may be expanded by performing the reverse of the compression process described above.

As described above, according to this embodiment, characters etc. to be superimposed are expressed with gradations of two or more values,
Since the boundary between the background and characters is expressed using a special code representing halftones, the step-like noise shown in FIG. 10 disappears and good quality can be obtained.In particular,
If superimposed characters, etc. are expressed with 8 or more gradations, there will be almost no visual difference from the background image, which has 256 gradations, and even with data compression, Good image quality can be obtained.

<Example 3> Next, Example 3 of the present invention will be described. The basic configuration is the same as that of the second embodiment, but the compression method for halftones is different. In this embodiment, the halftones of image data such as 3-bit, 8-gradation superimposed characters that have been pre-processed in the pre-processing unit 110 are expressed as L1, L2, . . . as shown in FIG.
, L6.

Similar to the second embodiment, a case will be explained using an example in which the letter "A" is scanned in the direction of the arrow FA in FIG. Similarly, the image data will be arranged as follows. This image data consists of three parts: a "black" part, a halftone "L" part, and a "white" part.

With respect to the output of the pre-processing section 110, the black portion and the white portion are subjected to run-length encoding similar to the above-mentioned G3 standard by a run-length counter 112 and a run-length encoding section 114. It is done. The run lengths of both the black and white parts are "4".
Therefore, the image data string is as follows: black 4→L1→L3→L6→white 4→L6→L4→L2→black 4.

In this case, in this embodiment, only one code is provided in the Huffman code to identify L indicating the intermediate tone L. Specifically, as part of the Huffman code system shown in Tables 4 and 5, a code "11" corresponding to L shown in Table 6 is provided.

[0045]

[Table 4]

[0046]

[Table 5]

[0047]

[Table 6]

Furthermore, as to how many consecutive L codes there are, the Huffman codes of character part run lengths shown in Tables 4 and 5 are used. Then, a specific 3-bit fixed length code for each level of halftone is assigned. Specifically, as shown in FIG. 8, for each halftone level L1, L2, ..., L6, "001", "01"
0”, “011”, “100”, “101”, “110”
” code is assigned to each. In addition, the black level is
000'', and the white level is ``111''.

Therefore, in the above example, black 4→
L → White 3 → 001 → 011 → 110 → White 4 → L → White 3 →
110 → 100 → 010 → Black 4.

The data thus run-length encoded is supplied to the Huffman encoding unit 116, and is expressed in Table 4.
Huffman coding is performed with reference to the code tables shown in Tables 5 and 6. The codes in these tables are Huffman codes that have been created in advance to be statistically efficient using various images, and short codes are assigned in order of run length with a high probability of occurrence. For example, the black 4
is encoded as "1011" with reference to "4" in the background run length from Table 4. Furthermore, from Table 4, white 4 is encoded as "011" with reference to "4" in the character part run length. The same applies to image data of other run lengths.

In the above example, Code data will be output in this order.

This embodiment also provides the same effects as the second embodiment, except that the compression method is different. Furthermore, in both embodiments 2 and 3, the amount of code after compression is 1.0% compared to the case where superimposed characters, etc. are treated as binary images and encoded using the G3 standard method.
This is about 5 to 2 times as large, making it possible to provide superimposed image data containing high-quality characters and symbols.

<Other Embodiments> The present invention is not limited to the above-mentioned embodiments, and includes, for example, the following embodiments. (1) The present invention may be applied to color images as well as black and white images. In particular, as the compression and decompression methods in the first embodiment, it is not always necessary to apply the second and third embodiments, and other methods may be used. The same applies to the multiplexing method. (2) When quantizing the superimposed image data in the pre-processing unit 110, for example, as shown in FIG.
Nonlinear assignment between input and output may be performed using the γ curve GA. In this way, fewer intermediate gradations can be used effectively.

(3) As shown in Tables 3 and 6, EOL (
END OF LINE) code is defined in the Huffman code and added after each scan line data. This EOL code is a unique code that does not appear in one line of data signal, and by adding it, resynchronization after an error burst becomes possible.

(4) In addition, the codes shown in each table are not limited to the above embodiments, and may be changed as necessary. (5) Furthermore, in the above embodiment, the image data is expressed in eight gradations, but it may be expressed in a plurality of other gradations.

(6) Furthermore, data compression processing may be performed by a microcomputer or the like. (7) In addition to characters, the process can be applied to similar items such as numbers and symbols. In particular, Examples 2 and 3 are not necessarily limited to those that are superimposed.

[0057]

As explained above, according to the data compression device and its decompression device according to the present invention, the background image signal and the superimposed image signal can be compressed separately and independently.
Since we decided to decompress, we can select the necessary processing method for both, and even if the compression ratio for the background image is high, we can perform good compression and decompression processing without deteriorating the superimposed characters etc. There is an effect.

In addition, image data is expressed in multiple gradations, and the maximum and minimum values of the gradations are Huffman encoded using run length, and the intermediate values are
Since codes such as those provided in the Huffman code system are assigned, the image data can be compressed with a high compression ratio and high quality.

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram showing a data compression device according to a first embodiment of the present invention.

FIG. 2 is a configuration diagram showing a data decompression device according to a first embodiment of the present invention.

FIG. 3 is an explanatory diagram showing an example of multiplexed image data in the embodiment.

FIG. 4 is a flowchart showing an example of a background image compression method in the embodiment.

FIG. 5 is a configuration diagram showing a second embodiment of the present invention.

FIG. 6 is an explanatory diagram showing halftone levels in Example 2.

FIG. 7 is an explanatory diagram showing a specific application example of compression according to the second embodiment.

FIG. 8 is an explanatory diagram showing halftone levels in Example 3 of the present invention.

FIG. 9 is a graph showing another example of quantization in pre-processing.

FIG. 10 is an explanatory diagram showing problems caused by binarization of image data.

FIG. 11 is a configuration diagram showing a conventional example.

[Explanation of symbols]

10... Image compression device (first data compression means), 12...
Superimposed image compression device (second data compression means), 14... multiplexer, 20... demultiplexer, 22... image expansion device (first data expansion means), 24
. . . Superimposed image decompression device (second data expansion means), 26 . . . Video mixing device, 110 . ,B,
C...Multiplexed bitstream, ESC1, ESC2, E
SC3, ESC4, ESC5, ESC6, L1, L2,
L3, L4, L5, L6...Middle tone level.

Claims (4)

[Claims] 1. A data compression device for compressing a background image signal and a superimposed image signal including characters and symbols, comprising: a first data compression means for compressing the background image signal; and a first data compression means for compressing the background image signal; A data compression device comprising: second data compression means for compressing a signal reversibly or at a visually undetectable level independently of the background image signal. 2. A data decompression device for decompressing image data compressed by the data compression device according to claim 1, comprising: first data decompression means for decompressing a compressed background image signal; A second process for expanding the imposed image signal independently of the background image signal.
A data decompression device comprising: data decompression means. 3. A data compression device that compresses image data in consideration of run length, comprising: data conversion means for expressing the image in multiple gradations; For values and minimum values,
A data compression device comprising: encoding means that performs run-length Huffman encoding, and encodes the intermediate value of the gradation by assigning a code provided in a Huffman coding system to each level. . 4. A data compression device that compresses image data in consideration of run length, comprising: data conversion means for expressing the image in multiple gradations; For values and minimum values,
Huffman encoding is performed using run length, and the intermediate value of the gradation is assigned a code provided in the Huffman coding system, and encoding is performed by adding a code indicating the run length of the intermediate value and each level. A data compression device comprising: encoding means.