JPH04358475A - Picture encoder - Google Patents

Abstract

PURPOSE:To reproduce a picture with excellent quality by keeping resolution of a character/a line drawing from a natural picture data in which the character/line drawing are in existence in mixture so as to implement effective compression. CONSTITUTION:A multilevel picture raster data 101 in which a character/a line drawing are in existence in mixture is given to an extraction circuit 102, in which a character/line drawing binary data 103 and a character/line drawing color data 104 are extracted. On the other hand, a multilevel picture data 105 subject to block processing at the character/line drawing extract circuit 102 is sent to a DCT circuit 106 and a quantization circuit 107, where the data are compressed and quantized and the result is given to an inverse quantization circuit 108 and an inverse DCT circuit 109, where the data is expanded to an original data and the expanded data is synthesized with the character/line drawing binary data 103 and a character/line drawing color data 104 without deterioration in the resolution and decoded into the multilevel picture data.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding device for encoding natural images containing line drawings.

0002

[Prior Art] Conventionally, color multi-level image information has a huge amount, which poses difficulties in system portability and handling, and in the case of devices with a storage device, memory costs increase significantly. There was a problem.

[0003] As a means of dealing with these problems, D
An encoding method has been proposed in which image information is orthogonally transformed by CT or the like, and then its coefficients are quantized and compressed. According to this method, even ordinary natural images can be compressed to a fraction of one to several tenths.

0004

[Problems to be Solved by the Invention] However, in the conventional example described above, when an image is divided into blocks, subjected to DCT processing, quantized, and compression encoded, it is difficult to compress the image generally well compared to a natural image. If you tune a quantization table and use that quantization table to compress characters, line drawings, etc., the image will be significantly degraded. In other words, a quantization table that is tuned to provide good compression for natural images has a characteristic that emphasizes low frequency components, and cuts high frequency components of characters and line drawings that occur at edges. Put it away. For this reason, when attempting to compress multivalued image data containing a mixture of text/line drawings and natural drawings, the quality of the edge portions of the text/line drawings will be significantly degraded.

[0005] This image compression device is used in a printer,
When compressing and storing multivalued image data that includes a mixture of text/line drawings and natural images, natural images are usually captured using a scanner, etc., and due to optical conditions such as aperture loss, high frequency components may be present to some extent when the data is generated. It has been suppressed. On the other hand, image data in which character/line drawing codes are developed into bitmaps is processed on a host computer or interpreter, so it has a lot of optical loss and has large high-frequency components that reach the limits of pixel density and amplitude. There is a high possibility that there are. Therefore, it is difficult to effectively compress both using the same quantization table, and it is an extremely serious problem that by deliberately increasing the compression rate, the character quality will deteriorate in printer applications. It has become.

[0006] Furthermore, even if one attempts to obtain a compression rate equivalent to a natural image while maintaining the quality of characters and line drawings by detecting the presence or absence of characters and line drawings in each block that undergoes DCT processing and switching the quantization table, , as long as DCT processing is performed, it is inevitable that the quality of the edge portion will deteriorate.

[0007] In order to solve these problems, it is necessary to separate and extract character/line drawing data from the natural image and encode it while maintaining the resolution and signal level. For example, separated multivalued natural images are quantized and compressed after DCT processing, and characters and line drawings are processed at 1 bit/pi, for example.
Save as xel resolution save data. When expanding image data, a natural image that is inversely converted after DCT compression is merged with character/line drawing data whose resolution is preserved, so that no deterioration in quality occurs at the edge portions of characters/line drawings.

However, in order to effectively realize this, it is necessary to reliably identify and extract characters and line drawings in multivalued image data. In other words, assuming that characters and line drawings have a uniform color, it is necessary to reliably identify the extraction color and extract the data even if the characters and line drawings have complex colored shapes. becomes.

The present invention has been made to solve the above-mentioned problems, and it performs effective compression by preserving the resolution of characters and line drawings from natural image data containing a mixture of characters and line drawings.
It is an object of the present invention to provide an image encoding device that can reproduce images of good quality.

0010

Means and Effects for Solving the Problems In order to achieve the above object, an image encoding apparatus of the present invention has the following configuration. That is, a blocking device blocks input image data, an edge detecting device detects edges based on gradation information of the image data blocked by the blocking device, and detects edges detected by the edge detecting device. The image forming apparatus includes an extraction means for extracting line drawings mixed in the image data according to the gradation information, and an accumulation means for accumulating the line drawings extracted by the extraction means as gradation data equal to or lower than the input data.

Preferably, the same color detection means inputs a color image and detects that the line drawing portions extracted by the extraction means from one color data of the color system are the same color; The apparatus is characterized in that it further comprises an output means for determining that the line drawing data stored in the storage means is valid according to the result in the storage means and outputting the determined data.

0012

DESCRIPTION OF THE PREFERRED EMBODIMENTS A preferred embodiment of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a schematic block diagram showing the configuration of an image encoding apparatus in this embodiment. In the figure, multivalued image raster data 1 containing a mixture of input characters and line drawings
01, character/line drawing binary data 103 and character/line drawing color 104 are extracted by the extraction circuit 102 and encoded. Here, the character/line drawing binary data 103 has the same resolution as the input multivalued image, and the character/line drawing part is set to "1".
, otherwise is bitmap data expressed as "0". Further, the text/line drawing color 104 indicates the color of the "1" portion of the text/line drawing binary data 103, and holds the same gradation information as the input multi-value image data. Then, the character line drawing binary data 103 and the character line drawing color 104 are processed by a synthesis circuit 111.
transmitted to.

On the other hand, multivalued image data 105 that has been divided into blocks by the character/line drawing extraction circuit 102 is transmitted to a DCT circuit 106 and a quantization circuit (Q) 107, where it is compressed and quantized. Then, the data is decompressed into the original data by the inverse quantization circuit 108 and the inverse DCT circuit 109, and the synthesis circuit 11
1, it is combined with character/line drawing data without resolution degradation and restored to multivalued image data.

Next, the detailed configuration of the above-mentioned character/line drawing extraction circuit 102 will be explained below with reference to the block diagram shown in FIG.

Multivalued image data 101 is first input to a blocking circuit 201 . Then, the blocking circuit 20
Step 1 is to divide the raster image data into blocks of X×Y pixels as shown in FIG. 3(a) for each color plane of the color system, and convert the raster image data into raster scan data within the block as shown in FIG. 3(b). . Generally, X and Y are equal (
X=Y), and if the input image data is color image data, it is output as blocks of data 202A to 202C for each color plane of the color system. 203 shown in FIG. 2 is the block data that is one block ahead of the block data 202A to 202C described above, and the color of this data is one of the color planes of the color system, and is the representative color for edge extraction. Become. Further, 204 is a block synchronization signal indicating the scan period for each block.

An outline of edge detection in this embodiment will now be explained with reference to FIG. 4.

First, as shown in FIG. 4(a), three pixels (x, y), (x+1, y), (x,
y+1), detect the signal level differences between pixels (x, y) and (x+1, y) and between (x, y) and (x, y+1), respectively, and if there is a difference above a certain level, Counters corresponding to the signal levels of both pixels where the difference exists are counted up. There are as many counters as there are signal levels. For example, if the signal level is expressed in 8 bits, there are 28 = 256 counters. In other words, pixels (x, y) and (x+1, y) and (x, y) and (
If there is a difference of more than a certain value between both pixel (x, y), the signal level counter of the pixel (x, y) is counted up twice. Then, the above-mentioned detection process is performed using the intra-block coordinates (
0,0) to (X-1,Y-1). In addition, in the case of x=X-1, y=Y-1, (x+1, y
)(x, y+1) is outside the block, but the signal level is treated as being equal to (x, y). In other words, the difference is "0" and no counting is performed.

In the case (b) shown in FIG. 4, the signal level counter of the pixel indicated by the ◯ symbol is counted up. Here, it is assumed that parts 401 and 402 have a uniform color,
The rest of the image is a mixed-color background. In this example, the signal level counter corresponding to the portion 401 indicates the maximum count value, and is identified as the color having the most edges. That is, this color portion is identified as a character/line drawing, and its signal level is output from the edge detection circuit 205 as a character/line drawing signal level 206. This output is
When the processing of the block of interest is completed, the block period signal 20
4, it is held in the flip-flop 207. As described above, the edge detection circuit 205 focuses on three pixels at the same time, but the three pixels 203 may be parallel, or a register may be provided in the edge detection circuit 205 to input three pixels serially.

Next, when the scanning in the edge detection circuit 205 is completed, the process moves on to the next block, but the block data 203 outputted immediately before from the blocking circuit 201 described above is converted into blocked data 202A. ~
Output as C. In addition, the block data 203 is
In contrast to the representative color used to detect edges, the blocked data 202 uses the entire color plane (202A) of the color system coordinates.
~C). From among these, select the color that is the same as the edge detection representative color,
That is, the blocked data 202A, which is the same data as the previous block data 203, is transferred to the flip-flop 2.
It is compared with the character/line drawing signal level 208A held at 07, and the same signal level is output from the equivalency detection circuit 209. That is, the equivalence detection circuit 209 outputs "1" if the two match, and otherwise outputs "0". Then, the binary data is held in the bitmap memory 211 as binary image data. Shown in Figure 4 (b)
In the case of , the signal level part 401 is extracted and a bit map of "1" or "0" is saved on the bitmap memory 211 as shown in (c).

Next, the same color detection circuit 214 detects whether the "1" portion in the bitmap pattern described above is the same color on another color plane of the color system. This scans the other color planes 202B and 202C, and if the "1" portions of each bitmap pattern match, the control signal 215 of the switch 216 is set to "L" and the bitmap memory 211 side is selected. , if they do not match, it is determined that the block is not a text/line drawing part, and the text/line drawing binary data 103 is selected so as to always be "0" in that block.

The same color detection circuit 214 also outputs signal levels 208B and C of color planes B and C in the portion where the bitmap memory pattern is "1", respectively. this is,
It is synthesized with the character/line drawing signal level 208A in the synthesis circuit 217 and output as the character/line drawing color data 104 of the block of interest. This character/line drawing color data 104 has the same gradation information as the input image data 101. Here, the characters
In order to synchronize the line drawing signal level 208A with the scanning results for bitmap creation and same color detection, the line drawing signal level 208A is delayed by a two block delay 218 for the time it takes to scan two blocks. This data is processed by the same color detection circuit 214 to
If it is determined that it is not a line drawing, meaningless data is output, but since it has meaning only when the character/line drawing binary data is "1", no problem occurs.

Next, the edge detection circuit 2 in this embodiment
The detailed configuration of 05 will be explained below with reference to FIG.

In the example shown in FIG. 5, the three pixels described above are input in parallel as block data 203a to 203c, and data 203a is the pixel (x, y), data 20
3b is (x+1, y), data 203c is (x, y+1
).

First, the difference between the block data 203a to 203c is calculated by the subtracters 501a and 501b, and the difference is calculated by the absolute value circuit (
Absolute values are taken at ABS) 502a and 502b. Then, the comparators 503a and 503b compare it with a threshold value 504, and if the difference (absolute value) is greater than the threshold value, "1" is output, and if it is equal or smaller, "0" is output. As a result, the pixel (
x, y) and adjacent pixels (x+1, y) and (x, y+1)
It is determined whether an edge exists between. That is, if the above-mentioned output is "1", it is determined that an edge exists. Here, the output of the comparator 503a is “1”.
”, the pixel (x, y
) and the signal level counter of the pixel (x+1, y) are counted up, and the output of the comparator 503b becomes "1".
In this case, the signal level counters of pixels (x, y) and (x, y+1) are counted up. Therefore, both “
1”, the signal level counter of pixel (x, y) is 2
times are counted up. Here, the counter module 509 has a counter corresponding to all the signal levels that the input image data 203a to 203c can take, and has an input of 8b.
In the case of it, 28=256 counters each correspond to a signal level.

Next, the counter module 509 described above
We will explain how to count. Here, the supplied clock has a rate twice as high as the pixel scanning speed, and is divided into 1/2 by a frequency divider 506 to switch the switches 505 and 507, respectively. This switch 5
When 05 is on the comparator 503a side, the switch 507 is on the data 203b side. The output of switch 505 is input to the input gate of counter module 509. Further, the decoder 508 outputs the output of the switch 507 and the pixel (
x, y) is input, the signal level is decoded, and two output lines are asserted at the same time. These output lines are connected to the input AND gates of the corresponding signal level counters, and the signal level counters of pixels where edges are present are counted up according to the above-mentioned clock. The maximum signal level of the output from each counter is encoded by a maximum value selection circuit 510 and extracted. This is the extracted color A mentioned above as the extracted signal level 206.
It is. Further, the counter module 509 is reset by the block synchronization signal 204 every block scan.

Next, the configuration of the same color detection circuit 214 will be explained with reference to FIG. The input image data 202C has a 1 block delay 60 to synchronize with bitmap creation.
1, it is delayed by one block scan time. On the other hand, the clock supplied at twice the pixel scan rate is divided into 1/2 by the frequency divider 602 and inputted to the AND gate 613. In addition, character line drawing bitmap data 2 is extracted and binarized with the same pixel density as the original data.
13 is input to the AND gate 613, and the output data from the gate 613, that is, the “
The input image data 202C corresponding to "1" is held in the flip-flop 604. Then, when the next bitmap data 213 is "1", the input image data 202C corresponding to The data is held in each flip-flop 604. Next, these data are compared in a comparator 606, and if they are not equal, the output "1" is outputted from the flip-flop 607. There will be parts with unequal colors in the extracted text/line drawing areas.

Note that the output of the flip-flop 607 is:
During the next block scan, it is held in the flip-flop 608 by the block synchronization signal 204. Further, the block synchronization signal 204 is transmitted through the flip-flop 604.
, 605, and 607, respectively. Then, the output of the flip-flop 605 becomes the extracted color of the character/line drawing portion, and when the block scan ends, it is held in the flip-flop 609 and becomes the extracted color 208C.

Further, 610b is the same circuit as 610a described above, and performs the same processing on the image input data 208B.

[0030] In either of the two modules, the character/
When non-uniformity in the color and signal level of the line drawing area is detected, O
The output of the R gate 611 becomes "1", and it is determined that there are no characters or line drawings in that block. In this case, as described above, extracted color B (208B), extracted color C (208B),
C) and extracted color A (208A) have no meaning.

Next, the configuration of the synthesis circuit 111 will be explained with reference to FIG.

Image data 1 restored by IDCT109
10 is replaced with the character line drawing color 104 by the switch 701 according to the "1" part of the bitmap data 103. This output is restored by the rasterization circuit 702 as the original raster image data.

As explained above, in this embodiment, the signal level that is most in contact with the edge is extracted independently for each block as the character/line drawing level.
Line drawings often span blocks, and in adjacent blocks, there is a high possibility that the extracted color of the previous block is significant. Therefore, a configuration is also possible in which the extraction signal level of the previous block is extracted preferentially. For example, as shown in FIG.
The extraction signal 208A of the previous block is sent to the edge detection circuit 205.
The counter module 509 shown in FIG. By multiplying the output of by a constant value,
This increases the probability of being selected preferentially.

As explained above, according to this embodiment,
Saves the resolution of characters and line drawings from color raster image data containing a mixture of text and line drawings, performs effective compression,
It becomes possible to reproduce images with good quality.

[0035]

[Other Embodiments] Next, another embodiment according to the present invention will be described below with reference to FIG.

FIG. 9 is a block diagram showing the configuration of an image encoding apparatus according to another embodiment, in which the transmission line section in FIG. 1 is constructed with a memory. In FIG. 9, 9
01 is a character/line drawing binary memory, and input image data 1
It is a binary data memory having the same pixel density as 01. Reference numeral 902 is a color memory for character and line drawings, and input image data 1
The same gradation as 01 is maintained for each region of the DCT processing block. An image memory 903 stores DCT processing of input image data and quantized compressed data.

The present invention may be applied to a system made up of a plurality of devices, or to an apparatus made up of one device. It goes without saying that the present invention can also be applied to cases where the present invention is achieved by supplying a program to a system or device.

[0038]

[Effects of the Invention] As explained above, according to the present invention,
It becomes possible to preserve the resolution of text and line drawings from natural image data containing text and line drawings, perform effective compression, and reproduce images of good quality.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of an image encoding device in this embodiment.

FIG. 2 is a diagram showing the configuration of a character and line drawing extraction circuit.

FIG. 3 is a diagram for explaining blocking of input image data.

FIG. 4 is a diagram for explaining edge extraction.

FIG. 5 is a diagram showing the configuration of an edge extraction circuit.

FIG. 6 is a diagram showing the configuration of a same color detection circuit.

FIG. 7 is a diagram showing the configuration of a synthesis circuit.

FIG. 8 is a diagram showing a modification of the present embodiment.

FIG. 9 is a block diagram showing the configuration of an image encoding device in another embodiment.

[Explanation of symbols]

102 Character line drawing extraction circuit 106 DCT processing circuit 107 Quantization circuit 108 Inverse quantization circuit 109 Inverse DCT processing circuit 111 Synthesis circuit 201 Blocked circuit 205 Edge detection circuit 209 Equivalence detection circuit 211 Bitmap memory 214 Same color detection circuit 217 Character line drawing extraction color synthesis circuit

Claims (2)

[Claims] 1. Blocking means for blocking input image data; edge detection means for detecting edges based on gradation information of the image data blocked by the blocking means; The present invention is characterized by comprising an extracting means for extracting line drawings mixed in the image data according to gradation information of the edge, and an accumulating means for accumulating the line drawings extracted by the extracting means as gradation data equal to or lower than the input data. An image encoding device that uses [Claim 2] Input a color image and select one of its color systems.
a same color detection means for detecting that the line drawing parts extracted by the extraction means from the color data are of the same color; and a same color detection means for determining that the line drawing data accumulated by the storage means is valid according to the result of the same color detection means. The image encoding apparatus according to claim 1, further comprising output means for outputting the image.