JP3274566B2 - Image coding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an encoding apparatus for encoding a binary image such as a character image or a halftone image, and to divide an image into bit planes in order to encode a multi-valued image without loss. The present invention relates to an image encoding device that encodes each bit plane image by binary encoding.

[0002]

2. Description of the Related Art In a storage device such as an image filing system or a multi-value facsimile device, when a multi-value image is losslessly coded, a method of directly coding the multi-value image (this method includes a predictive division code There is a conversion method)
And a method of converting to a binary image and encoding.

[0003] Conventionally, MH, M
R, coding methods such as arithmetic coding are known.
In H and MR, it is difficult to compress an image having systematic noise such as a halftone image, and in arithmetic coding, it is generally difficult to increase the speed because a template of surrounding pixels is used. Therefore, the present applicant has previously proposed an edge conversion fill conversion method capable of reducing entropy (code amount) while being a simple process (Japanese Patent Laid-Open No. 4-354).
472).

[0004] When a multi-valued image is reversibly encoded,
A method of dividing a multi-level image into a plurality of bit planes and coding each bit plane image by a binary coding method (referred to as bit plane coding) has been adopted. However, this method alone does not significantly increase the compression ratio. Absent.

[0005]

When encoding a binary image, high compression can be achieved by combining the edge conversion fill conversion method and Huffman coding. However, since the above-described encoding is performed on a line-by-line basis, it is difficult to adapt to an encoding method that refers to pixels across lines, such as the JBIG method that is currently the standard for binary image encoding. There's a problem.

A first object of the present invention is to support a binary coding system in which pixels are referred to across lines.
It is an object of the present invention to provide an image encoding device using edge conversion and fill conversion.

A second object of the present invention is to reduce the entropy of an image by performing multi-stage edge conversion or multi-stage fill conversion in a gradation direction, a horizontal direction, and a vertical direction in encoding a multi-valued image. It is to provide an image encoding device in which is improved.

A third object of the present invention is to divide a multi-valued image into bit planes, and to perform lossless coding by binary-coding each bit plane image. Another object of the present invention is to provide an image coding apparatus using edge conversion and fill conversion, which can support a binary coding method in which pixels are referred to.

[0009]

According to a first aspect of the present invention, in encoding a binary image, multi-stage edge conversion or multi-stage conversion is performed for each line data in the horizontal or vertical direction. Image conversion means for performing a fill conversion, counting means for counting the number of conversions in the conversion means, and the number of conversions at which the code amount is minimized when a predetermined image is subjected to a predetermined encoding process on the converted image obtained from the image conversion means A detection means for detecting the optimum processing times, and performs a predetermined encoding process on the converted image output from the image conversion means when the optimum processing times are detected by the detection means, together with data indicating the optimum processing times. In the image encoding apparatus having encoding means for transmitting, the number of conversions of the multi-stage edge conversion or the multi-stage fill conversion is the same for all lines. To have.

According to the second aspect of the present invention, when encoding a binary image, vertical image conversion means for performing multi-stage edge conversion or multi-stage fill conversion for each vertical line data, and the vertical image conversion means Counting means for counting the number of conversions, and detecting the number of conversions that minimizes the code amount when a predetermined encoding process is performed on the converted image obtained from the vertical image conversion means as a first optimal processing number. Detecting means, and a horizontal image that performs multi-stage edge conversion or multi-stage fill conversion for each horizontal line data on a converted image output from the vertical image converting unit when the detecting unit detects the optimum number of times of processing. Conversion means, counting means for counting the number of conversions in the horizontal image conversion means, and a code when a predetermined encoding process is performed on the converted image obtained from the horizontal image conversion means. Detecting the number of conversions that minimizes the number of times as a second optimal processing number, and performing a predetermined encoding process on the converted image output from the horizontal image converting means when the detecting means detects the optimal number of processing times. And an encoding means for sending the data together with the data indicating the first and second optimal processing times, wherein the number of conversions in the vertical and horizontal multi-stage edge conversions or multi-stage fill conversions is , And the number of times is the same for all lines.

According to the third aspect of the present invention, when encoding a multi-valued image, a dividing means for dividing the multi-valued image into a plurality of bit planes; Image conversion means for performing multi-stage edge conversion or multi-stage fill conversion for each line data in the vertical direction, counting means for counting the number of conversions in the image conversion means, and a predetermined encoding process for the converted image obtained from the image conversion means Detecting means for detecting the number of conversions at which the code amount becomes the minimum when the processing is performed as a first optimal processing number; means for synthesizing a plurality of bit plane images sequentially output from the image converting means; A gradation direction image conversion unit that performs multi-stage edge conversion or multi-stage fill conversion in the gradation direction on the image of the image, a counting unit that counts the number of conversions in the gradation direction image conversion unit, Detecting means for detecting, as a second optimum number of conversions, the number of conversions at which the code amount becomes minimum when a predetermined encoding process is performed on the converted image obtained from the gradation direction image converting means; A dividing unit for dividing the converted image output from the converting unit into a plurality of bit planes, performing a predetermined encoding process on each of the divided images, and sending the encoded image together with data indicating the first and second optimal processing times The multi-stage edge conversion or the multi-stage fill conversion for each line data in the horizontal or vertical direction is the same number of times for all lines. .

[0014] The invention described in claim 4 is characterized in that the number of conversions of the multistage edge conversion or the multistage fill conversion in the gradation direction is the same for all pixels.

[0013]

In one embodiment of the present invention, binary image data is input to a horizontal or vertical edge converter on a line basis, and edge-converted in the horizontal or vertical direction. The edge conversion frequency measuring device measures the number of times of edge conversion and sends the measured value to the determination circuit. The entropy measuring device measures the entropy and sends it to the judgment circuit. The determination circuit detects the number of edge conversions that minimizes the code amount based on the result obtained by the entropy measuring device. When the coder performs coding by the JBIG method, the same number of edge conversions are performed for all lines. Thereby, the correlation of each line is maintained and the compression ratio is improved. At the time of decoding, the reverse conversion is performed as compared with the encoding. Since the edge conversion and the fill conversion are conversions that are opposite to each other, by receiving the number of conversions at the time of encoding,
The original image is restored by performing the fill conversion by the number of times.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be specifically described below with reference to the drawings.

First, the multi-stage edge conversion and the multi-stage fill conversion used in the present invention will be described. FIG. 6 is a diagram for explaining edge conversion and fill conversion. In the edge conversion, the binary pixels to be compression-encoded are sequentially compared with the pixels from the beginning, and only a transition point from a white pixel (0) to a black pixel (1) or a black pixel to a white pixel is determined as a black pixel. , And the rest as white pixels.

This conversion is also edge detection, and corresponds to the absolute value of the difference between pixels of the binary image. That is, the first pixel is output as it is, and becomes an output sequence of pixel values obtained by performing exclusive OR processing between adjacent pixels sequentially from the first pixel. FIG.
In the example of, the pixel row on the scanning line for the binary image A is “00”.
111000 ", and when this data string is converted so that only the change point of the pixel is 1, the data string" 0010 "is obtained.
0100 "is obtained, and when edge conversion is performed on all lines, an image B is obtained as a result.

The fill conversion is a conversion in which black pixels (1) appearing first to one pixel before the next appearing black pixel (1) are filled with black pixels. Thereafter, when a further black pixel appears, the conversion is to fill the area with a black pixel up to the next black pixel. In other words, this is a process of converting a white pixel into a black pixel or a black pixel into a white pixel every time a black pixel appears.

In the example shown in FIG. 6, the pixel row on the scanning line for the binary image B is "00100100". When the above-described fill conversion is performed on this data row, the data row "001" is obtained.
11000 "is obtained, and when the fill conversion is performed on all the lines, an image A is obtained as a result. In this manner, the edge conversion and the fill conversion are operations opposite to each other, and are opposite in the encoding and the decoding. Further, the edge converter and the fill converter are configured by exclusive OR with one pixel shifter as shown in Fig. 7 and Fig. 8, respectively, and the hardware is simple.

The above-described examples of edge conversion and fill conversion are examples performed in the main scanning direction (which is referred to as the horizontal direction), but the same applies to the case in which the conversion is performed in the sub-scanning direction (vertical direction). .

FIG. 9 shows an example in which edge conversion and fill conversion are connected in multiple stages. For example, an 8-bit pixel column “001”
When one edge conversion is performed on 10101 ”(Org), a pixel row of the difference pattern“ 00101111 ”(Op1) is obtained. Further, when the edge conversion is performed on this difference pattern (that is, the edge conversion is performed twice) Repeat)
It becomes "00111000" (Op2), and multi-stage edge conversion is realized by repeating such edge conversion. Further, it can be seen that when edge conversion is repeated eight times, the original pixel row (Org) is obtained, and the pixel row changes cyclically. The fill transform is an inverse transform of the edge transform, and similarly changes cyclically.

At the time of encoding, it is sufficient to select and encode a pixel row in which edge conversion is repeated as many times as the entropy is minimized. For example, in the case of Huffman encoding in FIG. 9, a place where the number of runs is minimized, that is, What is necessary is just to select a pixel column (Op2) in which edge conversion is repeated twice. At the time of decoding, an original pixel row can be obtained by the same number of inverse transforms, that is, in this case, two fill transforms.

Next, edge conversion and fill conversion in the gradation direction will be described. For example, as shown in image C of FIG.
If there is a multi-valued image represented by 8 bits per pixel, a binary data string can be obtained by scanning a certain pixel g from MSB (most significant bit) to LSB (least significant bit). This is the same as the binary representation of the pixel. Edge conversion and fill conversion in the gradation direction are performed on this data string.

When a data string obtained by scanning a certain pixel g of the image C from the MSB to the LSB is “11010111”, by performing edge conversion on this data string, a data string “10111100” can be obtained. This is a converted pixel after the edge conversion. When the same processing is performed for all pixels, an edge-converted image D in the gradation direction can be obtained.

<Embodiment 1> FIG. 1 shows the configuration of an image coding apparatus according to Embodiment 1 of the present invention. The binary image data to be compressed is input to the horizontal or vertical edge converter 1 in line units. The horizontal or vertical edge converter 1 performs a horizontal or vertical edge operation on input data, and sends the processed data to the entropy measuring device 2 and the coder 4. The edge operation is repeated, and the edge conversion number measuring device 5 measures the number of edge operations and sends the measured value to the determination circuit 3. The entropy measuring device 2 measures the entropy of the processing data and sends the result to the determination circuit 3. The determination circuit 3 detects the number of edge operations (optimal processing times) that minimizes the code amount based on the result obtained by the entropy measuring device 2.

The judging circuit 3 selects the image data of the optimum number of times of processing from the data edge-converted by the horizontal or vertical edge converter 1 and sends it to the coder 4. The image data after the edge conversion is subjected to a predetermined encoding process in the coder 4 and then output as compressed data together with data indicating the number of times of processing obtained by the edge conversion frequency measuring device 5.

By the way, when coding a binary image by the JBIG method, the coder 4 uses Q, which is a kind of arithmetic coding method.
An M coder is used. When encoding is performed, reference pixels are used, and the arrangement of 10 pixels to be referred to (this is referred to as a template) is determined by the above method.

As described above, in the case where the coder 4 is a QM coder with a template that refers to 10 pixels and is used in the JBIG system, since the coder 4 refers to pixels on other lines, edge conversion with other lines is performed. If the number of times is different, the correlation between the lines is lost and the coding efficiency is rather lowered.

Therefore, in the present invention, when a QM coder that refers to another line is used, the same number of edge conversions is performed for all lines. Thereby, the correlation of each line is maintained, and an improvement in the compression ratio can be expected.

That is, the horizontal or vertical edge converter 1
Then, for example, edge conversion is performed on the first few lines. The data after the edge conversion is input to the entropy measuring device 2, and the number of edge conversions is measured by the edge conversion number measuring device 5. The entropy measuring device 2 measures the entropy of the several lines after the edge conversion, and determines the result by a determination circuit 3.
Send to

The determination circuit 3 determines the number of edge conversions at which the code amount is minimum based on the result obtained by the entropy measuring device 2. The number of times of edge conversion (optimal number of times of processing) at this time is stored in the determination circuit 3, and the number of times of edge conversion is also transmitted at the time of encoding. Then, the number of edge conversions from the determination circuit 3 is instructed to the horizontal or vertical edge converter 1,
Edge conversion is performed for all the lines at the designated number of edge conversions. The data after the edge conversion is input to the coder 4 and subjected to a predetermined encoding process, and is output as compressed data together with data indicating the number of times of edge conversion.

As described above, in the first embodiment, the edge conversion is performed the same number of times for all the lines in consideration of the JBIG method, so that the correlation of each line is maintained and the compression ratio can be improved.

In the above-described embodiment, the number of edge conversions that minimizes the code amount is determined from the edge conversion of the first few lines in consideration of the processing time. However, the present embodiment is not limited to this. Instead, the number of edge conversions may be determined from a predetermined number of lines in the input image, or the number of edge conversions that minimizes the code amount after performing edge conversion on all lines may be determined. Good.

<Embodiment 2> FIG. 2 shows the configuration of an image coding apparatus according to Embodiment 2 of the present invention. The description of the same components as those in FIG. 1 is omitted. In the second embodiment, as in the first embodiment, it is assumed that the coder 18 is a QM coder with a template that refers to 10 pixels and is used in the JBIG system.

The binary image data to be compressed is input to the vertical edge converter 11 in units of vertical lines, and the vertical edge converter 11 performs a vertical edge operation on the input data, and The data is sent to an entropy meter 12 and a lateral edge converter 14. The entropy measuring devices 12 and 16, the determination circuits 13 and 17, and the edge conversion frequency measuring devices 15 and 19 have the same functions as those in FIG.

As in the first embodiment, the edge operation of the first few lines in the vertical direction is repeated, and the edge conversion frequency measuring device 15 measures the number of edge operations and sends the measured value to the judgment circuit 13. The entropy measuring device 12 measures the entropy of the processing data, and sends the result to the judgment circuit 13. Based on the result obtained by the entropy measuring device 12, the judgment circuit 13 executes the edge operation of the edge operation that minimizes the code amount. Detect the number of times.

The vertical edge converter 11 includes a decision circuit 13
The edge conversion is performed for all the lines according to the number of edge conversions detected in step (1), and the data after the edge conversion is input to the horizontal edge converter 14.

The horizontal edge converter 14 performs a similar edge operation on the transmitted data after the edge conversion on a horizontal line basis, and sends the processed data to the entropy measuring unit 16 and the coder 18. The horizontal edge converter 14 repeats the edge operation of the first few lines in the horizontal direction, and the edge conversion frequency measuring device 19 counts the number of edge operations, sends the measured value to the determination circuit 17, and outputs the measured value to the entropy measuring device. At 16, the entropy is measured and sent to the judgment circuit 17. The determination circuit 17 detects the number of edge operations that minimizes the code amount based on the result obtained by the entropy measuring device 16.
The same number of edge conversions is performed for all the lines with the number of edge conversions detected in.

The data after the horizontal edge conversion is applied to the coder 18
, And is subjected to a predetermined encoding process. The encoded data is added with the number of edge conversions obtained by the determination circuits 13 and 17 and transmitted as compressed data.

<Embodiment 2> FIG. 2 shows the configuration of an image coding apparatus according to Embodiment 2 of the present invention. The description of the same components as those in FIG. 1 is omitted. In the second embodiment, as in the first embodiment, it is assumed that the coder 18 is a QM coder with a template that refers to 10 pixels and is used in the JBIG system.

The binary image data to be compressed is input to the vertical edge converter 11 in units of vertical lines, and the vertical edge converter 11 performs a vertical edge operation on the input data. The data is sent to an entropy meter 12 and a lateral edge converter 14. The entropy measuring devices 12 and 16, the determination circuits 13 and 17, and the edge conversion frequency measuring devices 15 and 19 have the same functions as those in FIG.

As in the first embodiment, the edge operation of the first few lines in the vertical direction is repeated, and the edge conversion number measuring device 15 measures the number of edge operations and sends the measured value to the judgment circuit 13. The entropy measuring device 12 measures the entropy of the processing data, and sends the result to the judgment circuit 13. Based on the result obtained by the entropy measuring device 12, the judgment circuit 13 executes the edge operation of the edge operation that minimizes the code amount. Detect the number of times.

The vertical edge converter 11 includes a decision circuit 13
The edge conversion is performed for all the lines according to the number of edge conversions detected in step (1), and the data after the edge conversion is input to the horizontal edge converter 14.

The horizontal edge converter 14 performs a similar edge operation on the transmitted data after the edge conversion in units of horizontal lines, and sends the processed data to the entropy measuring unit 16 and the coder 18. The horizontal edge converter 14 repeats the edge operation of the first few lines in the horizontal direction, and the edge conversion frequency measuring device 19 counts the number of edge operations, sends the measured value to the determination circuit 17, and outputs the measured value to the entropy measuring device. At 16, the entropy is measured and sent to the judgment circuit 17. The determination circuit 17 detects the number of edge operations that minimizes the code amount based on the result obtained by the entropy measuring device 16.
The same number of edge conversions is performed for all the lines with the number of edge conversions detected in.

The data after the horizontal edge conversion is applied to the coder 18.
, And is subjected to a predetermined encoding process. The encoded data is added with the number of edge conversions obtained by the determination circuits 13 and 17 and transmitted as compressed data.

<Embodiment 3> Embodiment 3 shown in FIG. 3 is an image coding apparatus when an input image is a multi-valued image.
As the coder, a QM coder is used as in the first and second embodiments. The multi-valued image is divided into a plurality of bit planes by a bit plane divider 20. For example, a multi-valued image having 8 bits per pixel is divided into eight bit planes as shown in an image C in FIG. Horizontal or vertical edge converter 2
1. Entropy measurement device 22, Edge conversion frequency measurement device 2.
5. The determination circuit 23 functions in the same manner as in the first embodiment.

However, the horizontal or vertical edge converter 21
Performs the same number of edge conversions on all the horizontal or vertical lines for each divided bit plane as in the first embodiment. Thereby, the correlation of each line is maintained, and an improvement in the compression ratio can be expected.

The data subjected to the edge conversion for each bit plane is converted by the bit plane synthesizer 24 into the image C of FIG.
And input to the gradation direction edge converter 26. The gradation direction edge converter 26 performs the gradation direction edge operation as described with reference to FIG. 10 and sends the processed data to the entropy measuring device 27 and the bit plane divider 30.

The bit plane divider 30 divides the edge-converted image sent from the gradation-direction edge converter 26 into bit-plane images and sends the bit-plane image to the binary coder 31.
For example, if the input image is an 8-bit multi-valued image,
Divided into two bit plane images. The divided binary bit plane data is separately encoded in the binary coder 31. In the binary coder 31, since encoding is performed for each bit plane, if the number of times of edge conversion in the gradation direction differs for each pixel, there is no correlation with adjacent pixels in the same bit plane, and the coding efficiency is reduced. Will be invited. Therefore, when encoding by dividing into bit planes, the number of times of edge conversion in the gradation direction edge converter 26 is set to be the same for all pixels. As a result, the correlation between adjacent pixels is maintained, and an improvement in the compression ratio can be expected.

To make the number of edge conversions in the gradation direction the same for all pixels, the following processing is performed. That is,
The gradation direction edge converter 26 is configured to output some pixels (for example, neighboring pixels including the pixel g in the image C) among all the pixels.
Is subjected to edge conversion. Edge conversion frequency measuring instrument 29
Then, the number of edge operations is measured for each pixel, and the measured value is sent to the determination circuit 28. The entropy measuring device 29 measures the entropy after the edge conversion, and sends the result to the determination circuit 28.

In the judgment circuit 28, an entropy measuring device 29
The number of edge operations (optimal processing times) that minimizes the code amount is detected based on the result obtained in (1), and the number of edge conversions is instructed to the gradation direction edge converter 26.
The gradation direction edge converter 26 performs edge conversion on all pixels with the same number of edge conversion times as instructed by the determination circuit 28. The binary coder 31 encodes the data for each bit plane, and outputs compressed data together with data indicating the number of edge conversions obtained by the edge conversion number measuring devices 25 and 29.

Embodiment 4 FIG. 4 shows the structure of a decoder for the binary image coding of FIG. The encoded information is divided by the decoder 41 into compressed image data and data indicating the number of times of edge conversion, and after the compressed image data is decompressed,
The data of the number of times is sent to the horizontal direction fill converter 43, and the data of the number of times is sent to the number of times of fill conversion determination unit 42.

The horizontal-direction fill converter 43 repeats the horizontal-direction fill conversion for the image data by the number of times of the fill conversion obtained by the number-of-times-of-fill-conversion determiner 42, and converts the converted data to the vertical-direction fill converter 44 Send to The vertical direction fill converter 44 repeats the fill conversion in the vertical direction by the number of times of the fill conversion obtained by the fill conversion number determination unit 42, and outputs the original binary image data.

FIG. 5 shows the configuration of a decoder for the multi-level image coding of FIG. The compressed data is divided by a decoder 51 into compressed bit plane image data and data indicating the number of times of edge conversion.
The data of the number of times is sent to the multivalued image generator 52 after restoration, and the number-of-times-of-fill-conversion determination unit 53 is sent. The multi-level image generator 52 integrates a plurality of restored bit plane data, generates one multi-level image, and sends it to the gradation direction fill converter 54.

The gradation direction fill converter 54 repeats the fill conversion in the gradation direction by the number of times of the fill conversion obtained by the fill conversion number determiner 53 for the multi-valued image data.
The converted data is sent to the horizontal or vertical fill converter 55. The horizontal or vertical fill converter 55 repeats the fill conversion in the horizontal or vertical direction by the number of times of the fill conversion obtained by the fill conversion number determiner 53, and outputs the original multi-valued image data.

In the configuration of the above-described embodiment, the edge conversion is used for encoding, and the fill conversion, which is the inverse of the edge conversion, is used for decoding. The same applies to the case where the fill conversion is used for the conversion and the edge conversion is used for the decoding.

[0056]

As described above, according to the first and second aspects of the present invention, multi-level edge conversion or fill conversion is performed on a binary image to convert the binary image into data optimal for the encoding process to be applied. By performing a predetermined encoding process after the above, the compression efficiency can be increased. That is, for example,
The number of conversions of multi-stage edge conversion or fill conversion is set to the same number for all lines, and by maintaining the correlation between lines, coding is performed using a binary coding method that refers to pixels across lines, such as the JBIG method. Even when the coding is performed, the coding efficiency is improved.

According to the third and fourth aspects of the present invention, the multi-level image encoding apparatus performs multi-stage edge conversion or fill conversion in the gradation direction, the horizontal direction, and the vertical direction to reduce the entropy of the image. Encoding efficiency can be increased.
In addition, the multi-valued image is divided into bit planes, and each bit plane image is binary-coded to enable lossless coding, and the number of times of multi-stage edge conversion or multi-stage fill conversion is set to the same number for all lines. By doing so, the encoding efficiency is improved even when encoding is performed by a binary encoding method that refers to pixels across lines, for example, the JBIG method. Furthermore, by making the number of edge conversions in the gradation direction the same for all pixels, the correlation between adjacent pixels is maintained, and the compression ratio can be improved.

[Brief description of the drawings]

FIG. 1 shows a configuration of an image encoding device according to a first embodiment of the present invention.

FIG. 2 illustrates a configuration of an image encoding device according to a second embodiment of the present invention.

FIG. 3 illustrates a configuration of an image encoding device according to a third embodiment of the present invention.

FIG. 4 illustrates a configuration of a decoder for image encoding according to a second embodiment.

FIG. 5 illustrates a configuration of a decoder for image encoding according to a third embodiment.

FIG. 6 is a diagram for explaining edge conversion and fill conversion.

FIG. 7 shows a hardware configuration of an edge converter.

FIG. 8 shows a hardware configuration of a fill converter.

FIG. 9 shows an example in which edge transformation and fill transformation are connected in multiple stages.

FIG. 10 is a diagram for explaining edge conversion in the gradation direction.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Horizontal or vertical edge converter 2 Entropy measuring device 3 Judgment circuit 4 Coder 5 Edge conversion frequency measuring device

Claims (4)

(57) [Claims] 1. An image conversion means for performing multi-stage edge conversion or multi-stage fill conversion for each horizontal or vertical line data when encoding a binary image, and a counting means for counting the number of conversions in the conversion means. Detecting means for detecting the number of conversions that minimizes the code amount when a predetermined encoding process is performed on the converted image obtained from the image converting means as an optimum number of times of processing, and detecting the number of optimum processing times by the detecting means Performing a predetermined encoding process on the converted image output from the image conversion unit when the image processing unit performs the multi-stage edge conversion or An image coding apparatus, wherein the number of times of multi-stage fill conversion is the same for all lines. 2. A vertical image conversion means for performing multi-stage edge conversion or multi-stage fill conversion for each vertical line data when encoding a binary image, and counting the number of conversions in the vertical image conversion means. Counting means; detecting means for detecting, as a first optimal processing number, the number of conversions that minimizes the code amount when a predetermined encoding process is performed on the converted image obtained from the vertical image conversion means; Horizontal image conversion means for performing multi-stage edge conversion or multi-stage fill conversion for each horizontal line data with respect to the converted image output from the vertical image conversion means when the optimal processing number is detected by the means; Counting means for counting the number of conversions in the directional image converting means, and converting the number of conversions at which the code amount becomes minimum when a predetermined encoding process is performed on the converted image obtained from the horizontal image converting means. Detecting means for detecting as the second optimal processing number, and performing a predetermined encoding process on the converted image output from the horizontal image converting means when the detecting means detects the optimal number of processing; In the image encoding apparatus, the encoding means for transmitting the data together with the data indicating the second optimum number of times of processing, wherein the number of times of the vertical and horizontal multi-stage edge conversion or the multi-stage fill conversion is the same for all the lines. An image encoding device, wherein the number is a number. 3. A dividing means for dividing a multi-valued image into a plurality of bit planes when encoding the multi-valued image, wherein each of the divided bit-plane images is divided into horizontal or vertical line data. Image conversion means for performing multi-stage edge conversion or multi-stage fill conversion, counting means for counting the number of conversions in the image conversion means, and a code when a predetermined image is subjected to a predetermined coding process to the converted image obtained from the image conversion means. Detecting means for detecting the number of conversions that minimizes the amount as a first optimal processing number; means for synthesizing a plurality of bit-plane images sequentially output from the image converting means; Gradation direction image conversion means for performing multistage edge conversion or multistage fill conversion in the direction, counting means for counting the number of conversions in the gradation direction image conversion means, and said gradation direction image conversion means Detecting means for detecting, as a second optimal processing number, the number of conversions at which the code amount becomes minimum when a predetermined encoding process is performed on the converted image obtained from the image processing unit, and a conversion output from the gradation direction image converting unit Dividing means for dividing the image into a plurality of bit planes; and coding means for performing a predetermined coding process on each of the divided images and transmitting the data together with the data indicating the first and second optimal processing times. Wherein the number of conversions of the multi-stage edge conversion or the multi-stage fill conversion for each of the horizontal or vertical line data is the same for all lines. 4. The image encoding apparatus according to claim 3, wherein the number of conversions of the multi-stage edge conversion or the multi-stage fill conversion in the gradation direction is the same for all pixels.