JPH09121286A - Method and device for compressing picture data - Google Patents

Abstract

PROBLEM TO BE SOLVED: To efficiently compress a half-tone picture without problems of cost and scale. SOLUTION: Picture data read by an image sensor 1 is supplied to subtractors 7 to 9 through an A/D converter 2 and a correction circuit 3. Picture element data in the just preceding position stored in a memory 5 is supplied as a predictive value to substractors 7 to 9 by a predictive value extraction circuit 6. The subtractor 7 performs 8-bit subtraction between a noticed picture element and the predictive value, and the subtractor 8 performs subtraction between respective upper 5 bits, and the subtractor 9 performs subtraction between respective lower 3 bits, and respective differences are operated. Encoding in a 5-bit Huffman encoder 10 ane a 3-bit Huffman encoder 11 is performed in accordance with the condition of these difference values, and one or both outputs are generated as encoded data.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for compressing image data.

[0002]

2. Description of the Related Art Generally, an image scanner device not only performs a process of reading a document image, but also a process of transferring the read image data to a host computer (for example, a personal computer) which is a higher-level device through a predetermined interface. To do.

Here, this transfer takes the form of a serial interface or a parallel interface, but in any case, the transfer rate on these interfaces is extremely slow compared to that on the internal bus of the host computer.

Therefore, within the image scanner, it is desired to compress and encode the read original image, reduce the amount of transfer information, and shorten the transfer time.

[0005]

Various compression algorithms have been proposed and used so far. For example, the reversible compression method includes a method based on the known JBIG method, an LHA method, and the like.

However, in order to make the processing time associated with the compression negligible with respect to the time required to transfer the image data read by the image scanner, for example, it must be done by hardware. Usually, in such a case, it is possible to perform the operation at high speed by using a gate array, but in the case of the above algorithm, tens of thousands to hundreds of thousands of gates are required, resulting in an increase in size of the device. It is disadvantageous in terms of cost.

On the other hand, as a very simple processing algorithm, there is one in which a difference value is encoded by a predictive encoding method and is further encoded by using a Huffman code. To realize this with a gate array, several thousand gates are required, which is advantageous in terms of cost.

A brief description of this algorithm is as follows. That is, as shown in FIG. 1, when attention is paid to a pixel y (8 bits = 256 gradations) in image data (two lines) captured from an image sensor (hereinafter referred to as a pixel of interest), When encoding a pixel, the density of the pixel of interest at that position is predicted from surrounding pixels a, b, and c (hereinafter, referred to as a prediction value).

As the predicted value, for example, the value of the pixel a may be used, the pixel b or the pixel c may be used, and the predicted value may be predicted by, for example, "a + bc".

In any case, the predicted value and the actual pixel of interest y
And a difference value (possible values are −255 to +255).

Based on the result of this calculation, the Huffman coding table shown in FIG. 2 is used for coding.

For example, when the difference value is "-15", the group number is "4", so the code word is "10".
1 ”. Also, the additional bits become 4 bits, and the 4
The most significant bit of the bits represents the sign (+: 1,-: 0), and the lower 3 bits represent the position of "15". Therefore,
(15) → 111, (14) → 110, (13) → 10
1, (12) → 100, (11) → 011, (10) →
010, (9) → 001, (8) → 000.

Therefore, in the case of "-15", the additional bit is 0111 and the encoded data is "1010111".
Will be represented by 7 bits.

When the difference value is extremely large, such as "255", a larger number of bits is required for the input 8 bits of pixel data, but such a situation rarely occurs. Therefore, the influence of the amount of information on the entire image is small. In particular, when the image read by the image sensor is a halftone image such as a photograph, most of it is compressed to less than 8 bits.

[0015]

The invention of the present application is based on the above-mentioned predictive coding in principle, but the method is further developed so that halftone images can be efficiently and costly. An object of the present invention is to provide a method and apparatus for compressing image data, which makes it possible to obtain a sufficient effect in terms of scale.

Therefore, for example, the image data compression method of the present invention operates as follows. That is, a method of encoding the target pixel data based on the difference between the input multi-valued pixel data and the predicted value obtained based on the multi-valued pixel data in the vicinity of the input multi-valued pixel data,
The difference between the predicted value and the pixel of interest is detected, the input multi-valued pixel data and the predicted value are divided into upper m bits and lower n bits, respectively, and the respective difference data are encoded independently, and based on the detected difference. Then, the higher m bits of differential encoded data or / and the lower n bits of differential encoded data are selected and output.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An example of an embodiment according to the present invention will be described in detail below with reference to the accompanying drawings.

First, the outline of the present invention will be described below.

When the image read by the image sensor is a halftone image such as a photograph, the higher the reading resolution, the smaller the density difference (difference value) between adjacent pixels in the read image. Therefore, although the predictive coding method described above has an advantage in this respect, it also has a problem.

That is, the fact that the difference value (density difference) is small means that the variation of the difference is likely to appear in the lower several bits. In other words, in most cases, the upper bits do not change between pixels, or if any, the change is very small.

Therefore, in the present embodiment, the image data input from the image sensor (of course, the data after the shading correction processing) is divided into upper n bits and lower m bits, and the prediction code is applied to each of them. By performing the coding process, the predictive coding process is realized more efficiently.

In order to simplify the explanation, an example in which the input image data is 8 bits and n = 5 and m = 3 will be described below. Note that, as the predicted value of the pixel of interest in the present embodiment, the value of the immediately preceding pixel (pixel a in FIG. 1) will be described as being used.

In the embodiment, the following three equations (i)-
Calculate (iii). That is, (i) Di (8) = Y (8) -X (8) (ii) Di (5) = Y (5) -X (5) (iii) Di (3) = Y (3) -X (3) Here, Di is the difference value, Y is the actual value of the target pixel, and the value of the pixel at the position immediately before the target pixel used as the predicted value (corresponding to pixel a in FIG. 1). The number in parentheses indicates the number of bits, and “5” in the parentheses indicates the upper 5 bits and “3” indicates the lower 3 bits. Note that, depending on the above calculation, a positive or negative sign is generated, so the difference value D is exactly one bit larger,
It is presented above for clarity of the following description.

Regarding Di (3), as shown in FIG.
It is assumed that the calculation is performed in the relationship shown in.

For example, when X (3) = 3 and Y (3) = 0, Y (0) is at the position of the value "0" shown in the figure, and X (3)
“3” of “3” is “3” in the + direction shown in the figure. Therefore,
D (3) = 3. Also, for example, X (3) = 7, Y
When (3) = 1, X (3) is viewed from Y (3)-
D (3) =-2 because it is at a position advanced by "2" in the direction.
Becomes The above is shown for all cases in FIG.
Table.

In the present embodiment, the following cases are classified based on the results of the above-mentioned arithmetic expressions (i) to (iii), and the respective cases are encoded.

Case (A) −7 ≦ D (8) ≦ 7 and D (5) ≠ 0 Case (B) D (5) = 0 and −7 ≦ D (8) ≦ 7 Case (C) | D ( 8) |> 7 and D (5) ≠ 0 (| x | indicates the absolute value of x).

In the cases (A) and (B), the Huffman coding table shown in FIG. 5 is used, and in the case (C), the Huffman table shown in FIG. 6 is used.

That is, in cases (A) and (B),
The possible value of the difference value D (3) is -3 to + after all.
Since a value of 4 (range of −3 to +3 in case (A)) can be taken, the code code is generated as shown in these cases. In the figure, there are two types of code words (a) and (b) because the states at the pixel position next to the pixel of interest are cases (A) and (B) (code word (a) Case) or case (C) (in the case of codeword (b)). For example, when the difference value is “4”, the code word is either one, but the number of additional bits is 1 bit, and that bit is “1” (if this bit is “0”, the difference value is It means "0").

After all, the case (A) encodes the lower 3 bits of the difference value in the pixel of interest, and the case (B) encodes the upper 5 bits thereof.

On the other hand, in case (C), when the value of the pixel of interest greatly differs from the predicted value (the absolute value of the difference is 7
Greater than). In this case, the upper 5 bits,
The difference value of each of the lower 3 bits is encoded. Since a maximum of 5 bit operation is performed, the range that can be taken as the difference value is −31 to +31 as shown in FIG.

Codewords (1 to 5 bits) and the additional bits shown in the drawing are added according to the range of each difference value. For the additional bits, the most significant bit is used for the code, and the distance is assigned to the remaining bits.

For example, as the difference value, -7 to -4 and 4 to
In the case of 7, since the number of additional bits is 3 bits, 000B → -4 001B → -5 010B → -6 011B → -7 100B → + 4 101B → + 5 110B → + 6 111B → + 7 (B is Indicates binary). Other allocations can be easily guessed.

In case (C), the upper 5 bits and the lower 3 bits are encoded, and one pixel is represented by two code codes.

FIG. 7 shows a coded data format in the embodiment. The same figure (A) is case (A),
The figure (B) is the case (B), and the figure (C) is the case (C).
3 shows the data format of each of the encoded data.

In the code word in FIG. 5, for example, the code word “110” in (a) (next pixel is either case (A) or (B)) and FIG. 6 in case (C).
Although the codewords “110” of are in agreement with each other, the problem of consistency does not occur. The reason is given below.

Now, as density data, 255 (0th), 240 (1st), 239 (2nd), 235
(3rd), 232 (4th), 180 (5th), 1
It is assumed that pixel data (8 bits each) is input in the order of 86 (sixth), 187 (seventh), ....

Here, the values of the upper 5 bits, the lower 3 bits, and the calculated values of D (8), (5), and D (3) shown above are summarized and shown in FIG. .

Here, for example, the fourth pixel (the density value is 2
32) and the fifth pixel (having a density value of 180), it can be seen that the density change during this time is relatively large. Therefore, in the coded data of the fourth pixel, the code word is “0000” (meaning that the next fifth pixel is the case (C)), and the additional bit is “0”,
It can be identified that at least the fifth pixel represents the encoded code of one pixel by the upper 5 bits and the lower 3 bits.

Focusing on the fifth pixel, the code word in the high-order 5-bit code is "110", the additional bit is 3 bits and becomes "011", and the code word of the following low-order 3 bits is "01" (next The sixth pixel of the case (A),
(Any of (B)), and the additional bit becomes “0”.

As described above, since the consistency can be obtained while reflecting the case of the next pixel in the code word, no problem occurs.

As described above, FIG. 9 shows a concrete configuration example of an apparatus for realizing the above embodiment.

In the figure, 1 is an image sensor, 2 is an A / D converter for converting an analog signal from the image sensor into 8-bit digital data, and 3 is a correction circuit for performing shading correction processing and enlargement / reduction processing. .

Reference numeral 4 is a delay circuit for delaying one pixel, and is composed of, for example, a D type flip-flop. The reason for providing this delay circuit is that when the pixel data from this circuit is the target pixel data, the pixel data obtained from the correction circuit 3 is the next pixel data, and depending on the situation of the next pixel as described above. This is to generate a codeword. A memory 5 stores the input 8-bit image data.
In the embodiment, since the pixel data a immediately before the pixel of interest in FIG. 1 is used as the prediction value, it may be configured by a D type flip-flop, for example. However, if the predicted value is determined based on either the pixel c or d or the calculation result of ac as the predicted value, at least one line of memory capacity is required.

Reference numeral 6 is a prediction value extraction circuit for extracting and outputting a prediction value based on the pixel data stored in the memory 5.
Reference numerals 7 to 9 are subtractors for calculating the expressions (i) to (iii) described above. The subtractor 7 uses the input pixel data as the target pixel data (8 bits), and then outputs it from the prediction value extraction circuit 6. The predicted value (8 bits) is subtracted and the result is output. Further, the subtractor 8 subtracts the upper 5 bits of the predicted value from the predicted value extraction circuit 6 from the upper 5 bits of the pixel of interest and outputs the result. Then, the subtractor 9 subtracts the lower 3 bits of the predicted value from the lower 3 bits of the pixel of interest and outputs the result.

Reference numerals 10 and 11 are Huffman encoding circuits,
As described above, the codeword + additional bits are generated and output according to the cases (A) to (C). The encoding circuits 10 and 11 incorporate a logic for encoding. However, a table for performing the encoding process may be stored. The encoding circuit 10 performs the encoding process for the upper 5 bits, and the encoding circuit 11 performs the encoding process for the lower 3 bits. It also performs a process of generating a codeword based on the values of the pixel of interest and the next pixel.

The coded data synthesizing circuit 12 includes subtractors 7 to 7.
According to the subtraction result of 9, it is determined which of the cases (A) to (C) the target pixel corresponds to. For example, in case (A), the Huffman encoding circuit 11 outputs 3 The bit encoded data is input in the case (B), the 5-bit encoded data from the Huffman encoding circuit 10 is input, and in the case (C), the encoded circuits 10 and 11 are input in this order. , Combine those data.

Reference numeral 13 is a data alignment circuit, which makes encoded data into, for example, 8-bit units (aligns) and sequentially stores them in the output buffer memory 13. If it is less than 8 bits, it is connected to the next encoded data and waits until it becomes 8 bits or more.

The encoded data is stored in the output buffer memory 14 in this manner. This data will be transferred to the host computer, which is a higher-level device, via the interface 15.

In FIG. 9, the portion surrounded by the broken line (reference numeral 100) is actually composed of a gate array. Moreover, since there is actually a control system for moving the scanner or the document, the configuration shown in the drawings is not all.

Now, the processing procedure in the above configuration can be summarized and shown as shown in FIG. Here, the output buffer memory 13 for encoded data of 1 byte (8 bits)
Shows the procedure of the storing process to.

First, in step S1, the image data is input, and in steps S2-4, the difference value from the predicted value (the value of the immediately preceding pixel in the embodiment) is 8 bits, upper 5 bits, lower 3
Calculate each bit. In step S5, the state of the pixel of interest is identified according to the calculation result, and the process branches to the corresponding process.

For example, in case (A), the process proceeds to step S6, Huffman coding of the lower 3 bits is performed, and the code data (code word + additional bit) is processed in step S7.
Generate In case (B), Huffman coding of the upper 5 bits is performed in step S8, and code data is manufactured in step S9. Then, in case (C), code data of upper 5 bits and lower 3 bits are generated in steps S10 and S11.

Thus, in step S12, it is determined whether or not the generated code data has 8 bits, and if it is less than 8 bits, the process returns to the process of inputting the next pixel data (step S1), The above process is repeated. Then, when a block of code data of 8 bits or more is formed as code data, the data is aligned in units of 8 bits and stored in the output buffer memory 13.

As described above, according to the present embodiment,
Since the input 8-bit multi-valued image data is divided into upper 5 bits and lower 3 bits and encoded for each, for example, in the case of an image with a lot of gentle density,
Most of them are substantially represented by only a 3-bit code or a 5-bit code, and the compression rate can be increased as compared with the conventional simple Huffman coding. Therefore, for example, by applying the present invention to an image scanner device,
It becomes possible to transfer the image data to the host computer, which is a higher-level device, in a shorter time.

In the embodiment, the Huffman encoder 1
Although 0 and 11 have been described as being divided into two, the encoding logic or table stored therein is not for 8 bits each, and is simply a double of the table used for normal Huffman encoding. It also has no capacity. Therefore,
Although the device configuration requires a few more gates, it is a size that can be ignored.

Furthermore, in the embodiment, the division ratio is ranked in the top five.
Although the bit and the lower 3 bits are used, the present invention is not limited to this, and other combinations of the number of bits (for example, upper 6 bits, lower 2 bits, etc.) may be used.
In particular, the input image may be, for example, 16 bits instead of 8 bits, and in this case, for example, upper 10 bits and lower 6 bits may be used. However, when the inventor read various images and experimented, when the input image was 8 bits, the upper 5 bits and the lower 3 bits described in the above embodiment were used.
The division at the ratio of bits was the most stable, and good results could be obtained.

Although not specifically described, for example, a driver program for driving and controlling the image scanner device of the present apparatus is operating on the host computer, and the encoded data received by this program via the interface is Decoding and reading also provides the decoded image data to the original application program.

[0059]

As described above, according to the present invention, it is possible to provide a method and apparatus for compressing image data of a halftone image efficiently and at high speed without any problem in terms of cost and scale.

In particular, when the present invention is applied to the inside of the image scanner device, it is possible to reduce the amount of information when transferring the read image data to the host device, and consequently to shorten the time required to transfer the image data. become.

[0061]

[Brief description of the drawings]

FIG. 1 is a diagram showing a relationship between a pixel of interest in Huffman coding and peripheral pixels for calculating a prediction value.

FIG. 2 is a diagram showing an example of a Huffman coding table.

FIG. 3 is a diagram for explaining an outline of difference value calculation of lower 3 bits in the embodiment.

FIG. 4 is a diagram showing all combinations of difference values of lower 3 bits in the embodiment.

FIG. 5 is a diagram showing a relationship between a code word and additional bits for encoding the lower 3 bits in the embodiment.

FIG. 6 is a diagram showing a relationship between a codeword and additional bits for encoding the upper 5 bits in the embodiment.

FIG. 7 is a diagram showing a format of encoded data in the embodiment.

FIG. 8 is a diagram showing an example of image data and actual compressed code data in the embodiment.

FIG. 9 is a block configuration diagram of a hardware portion that performs an encoding process in the image scanner device according to the embodiment.

FIG. 10 is a flowchart showing a procedure of encoding processing according to the embodiment.

[Explanation of symbols]

 1 Image Sensor 2 A / D Converter 3 Correction Circuit 4 1 Pixel Delay Circuit 5 Memory 6 Prediction Value Extraction Circuit 7 8-bit Subtractor 8 5-bit Subtractor 9 3-bit Subtractor 10 5-bit Huffman Encoder 11 8-bit Huffman Code Coder 12 code data synthesis circuit 13 data alignment circuit 14 output buffer memory 15 interface

Claims (5)

[Claims] 1. A method for encoding pixel data of interest based on a difference between input multi-valued pixel data and a predicted value obtained based on multi-valued pixel data in the vicinity of the input multi-valued pixel data. Then, the difference between the prediction value and the pixel of interest is detected, the input multi-valued pixel data and the prediction value are divided into upper m bits and lower n bits, and the respective difference data are independently encoded, and the detected difference is detected. A method of compressing image data, comprising selecting and outputting the higher m bits of differential encoded data or / and the lower n bits of differential encoded data based on the above. 2. The image data compression method according to claim 1, wherein when the input multi-valued pixel data is 8 bits, the m is 5 and the n is 3. 3. An apparatus for encoding pixel data of interest based on a difference between input multi-valued pixel data and a predicted value obtained based on multi-valued pixel data in the vicinity of the input multi-valued pixel data. Detecting means for detecting the difference between the predicted value and the pixel of interest; first encoding means for encoding the difference of the upper m bits of the input multivalued pixel data and the predicted value, and the input multivalued pixel data. A second coding means for coding the difference of the lower n bits of the prediction value and the predicted value, and coded data generated by the first coding means based on the difference detected by the detecting means, or / And a selecting means for selecting the encoded data generated by the second encoding means, the image data compressing apparatus. 4. A reading unit for optically reading an original image for inputting multi-valued pixel data, and the coded data selected by the selecting unit are transferred to a higher-level device via a predetermined interface. The image data compression apparatus according to claim 3, further comprising a transfer unit. 5. The first detecting means calculates the difference value of all bits between the pixel of interest and the predicted value, and the second calculating means calculates the difference of the upper m bits of the pixel of interest and the predicted value.
4. The image data compression apparatus according to claim 3, further comprising: an arithmetic unit for detecting the difference between the predicted value and the pixel of interest based on the arithmetic results of the first and second arithmetic units. .