JPS6053981B2 - Facsimile signal quantization method - Google Patents

Description

[Detailed description of the invention] The present invention relates to a method for quantizing facsimile signals.

Generally, in a digital facsimile, a pixel signal is quantized into ``white'' and ``black'' using a single threshold value, and compression processing is performed on the quantized signal.

In the case of such a quantization method, processing near the boundaries of characters and figures becomes uniform, and as a result, the quantized binary ``white'' and ``black'' patterns are not suitable for compression processing. This results in a pattern with low compression effect. One method to improve this is "signal conversion," a method that improves the compression rate by changing a part of the quantized pattern (e.g. Journal of the Communication Society Vol. 160-A) No. 21977.
) has been proposed, but further improvement is needed in terms of image quality. The present invention develops this idea of signal conversion and provides a quantization method that can obtain high image quality and high compression rate.

The method of the present invention performs multi-tone quantization on each pixel of a facsimile signal, and then, the multi-tone quantized signal indicates that pixels having a value above a certain value are ``white'' and pixels below a certain value are ``white''.
Binary quantization is performed to "black", and for other intermediate gradation pixels, binary quantization is performed to "white" or "black" to match the binary predicted value. It is characterized by the fact that the value of the next pixel is ``white'' or ``black'' after the intermediate gradation pixel that should be subjected to binary quantization.
When the predicted value of the pixel of the intermediate gradation differs from the predicted value of the pixel of the intermediate gradation, the binary quantization of the pixel of the intermediate gradation is performed to a value closer to the predicted value.

The present invention will be described in detail below with reference to the drawings.

FIG. 1 shows one embodiment of the method of the present invention. Although two-dimensional prediction will be considered here as an example of a compression method, the present invention is not limited in application to such a method.

Two-dimensional prediction is to predict the input signal using the information of the previous scanning line and the current scanning line, and if the prediction matches, ``o''
If there is no match, a "1" output is output, and this output sequence is run-length coded and sent.

In FIG. 1, a read picture signal is input to one quantizer.

The quantizer 1 performs multi-gradation quantization on an input signal, and for the sake of simplicity, it is assumed that the input signal is quantized into four gradations. Therefore, if we associate the white level with ゜゛1r1 and the black level with ゜゛00゛, the signs ゜゜1σ゛ and ゜゛ correspond to the signal levels of intermediate gradations, but these
This is the signal level in an area where it is uncertain whether it should be white or black.

The memory circuits 3 and 4 are circuits that each store a 2-bit signal, and are composed of, for example, two stages of flip-flops. When the 4-gradation quantized signal for a pixel to be binary quantized is stored in the storage circuit 4 in 2-bit form, the storage circuit 3 also stores a similar quantized signal for the next pixel. Become. 5 is a circuit for detecting the quantized code of the code pattern ゜゜11'', ゛0σ゛.

When the code pattern stored in the memory circuit 4 is "゜1r゛゜゜00゛," it means that the signal has a value closest to "゜゜white" and "゜black."

Therefore, at this time, a signal is output from the detection circuit 5 to open the AND gate 7 via the 0R gate 6, and the signal of the most significant bit (MSB) stored in the memory circuit 4 is output to the 0R gate 8. The signal is input to the 1-line memory 9 via the 1-line memory 9.

That is, when the MSB is ``1'', it corresponds to a binary quantized signal of ``white'', and when it is ``0'', it corresponds to a binary quantized signal of ``black''. It consists of
When a signal is input from the left side, it is sequentially shifted to the right and the pixel values for one scanning line are stored.

FIG. 2 is a diagram explaining two-dimensional prediction, and L1 is the previous scanning line.

is the scanning line in which the target pixel 5 to be subjected to binary quantization currently exists. Pixel 4 is the immediately preceding pixel on the same scanning line. Therefore, the state of the 1-line memory 9 in FIG. 1 is as shown in the numbers shown in the 1-line memory 9 in FIG. 1 at the timing when the pixel 5 is output from the 0R gate 8.

10 is a predictor, which predicts pixels 1, 2, and 3 (second
The predicted value for pixel 5 is created by inputting the previous pixel 4 (see figure) and the previous pixel 4. From the binary quantized value (1 or 0) input to the 1-line memory 9, the predicted value for pixel 5 is Or it is predicted to be 0.

Although rules for creating such predicted values can be freely given, a predicted pattern as shown in FIG. 3, for example, can be considered from the general characteristics of calligraphy and strokes. Each of FIGS. 3a to 3p corresponds to FIG. 2, respectively. For example, as shown in FIG. “゜1
Output ゛. Also, as shown in FIG. 3f, pixel 1,
If 3 is "white", 2 is "black", and the previous pixel is "black", pixel 5 is predicted to be "black" and logic '408 is output.
Similarly, the predictor 10 outputs the logic ``1'' to the output line 11 when the predicted value is ``white'' for all pixels, and ``0'' when the predicted value is ``black''. 12 is an exclusive OR circuit which inputs "1" when the two logics are different.

Now, if there are detection outputs of ゜゜11゛ and "゜00゛" from the detection circuit 5, the logic "1" or "0" is output from the 0R gate 8, so these are compared with the predicted value in the circuit 12, and the predicted value is When the value does not match, the logic ゜゜1゛ is outputted from the circuit 12.

On the other hand, when the pixel value is other than ゜11゛ or ゜00゛, the MSB of the memory circuit 4 of the 2-bit flip-flop is used as a general rule.
Instead, the output of the child measuring device 10 is 1 as the final binary quantization.
It is input to the line memory 9.

That is, when the value is other than ゜"1 block゛, ゜゜00゛, there is no output from the 0R gate 6, and the output of the predictor 10 is output from the AND gate 13,
It is input to the 1-line memory 9 via the 0R gate 8.
Furthermore, since the final binary quantization and the predicted value from the predictor 10 match, the circuit 1
A logic "0" is output from 2 to indicate a match. Therefore, at times other than "11", . . . "00", the predictions always match, and the compression ratio is improved. however,
An exception is made for a specific pattern, and when this pattern appears, the exception detection circuit 14 operates to input the MSB of the storage circuit 4 to the one-line memory 9 as the final binary quantized value j.

This results in a pixel improvement. A memory circuit 3 consisting of two stages of flip-flops is provided for this purpose.

That is, as shown in FIG. 4a, since the multi-gradation quantization value of pixel 5 is ゜"0r゛"(101-001<111-011), it is a value closer to ゜"black".

), ゜゛white゛ matches the child measurement value, but
If the next pixel 6 is ゜゜00゛, that is, ゜゜black゛゛, a prediction mismatch will eventually occur in the next signal.Therefore, in this case, binary quantization is applied to the closer black signal ゜1゛. do. That is, for example, when a specific pattern as shown in the table below is detected, it is binary quantized as shown in the table below. By doing this, when ゜゜bleeding゛A occurs, as shown in Fig. 4b, it is quantized to white, obtaining sharp corners, and at the same time increasing the compression ratio, as shown in Fig. 4c. The “black” shaded area B
In this case, unlike the case of single signal conversion in the conventional method described above, data can be saved reliably. Such processing has been impossible with conventional methods. In this way, multi-gradation quantization is performed on 2 bits, and for pixels that can be clearly determined as black or white, such as "00゛," signal.

Then, for uncertain pixels such as ゛゛゛,゜゜゜゛10゛, this is detected by the ``11゛,゜゜00゛ detection circuit, and the predicted value determined by the predictor 10 is converted into a binary quantum In addition, if quantization to either black or white has no effect on the compression rate (for example, when the specific pattern in the table above is detected), "゜10゛ is white,""11゛,"O D is binary quantized to a more probable value such as black "゜00゛."

As explained in detail above, since intermediate gradation values are treated as uncertain areas and binary quantized into values that match the predicted values, the coincidence obtained from the exclusive OR circuit 12 in FIG.
The number of matches of non-matching logic signal sequences increases, and by compressing and transmitting it using run-remps encoding, etc.
According to the present invention, it is possible to transmit a facsimile quantized signal with high efficiency and high quality.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an embodiment of the method of the present invention. Figures 2 and 3 show prediction patterns for two-dimensional prediction. FIG. 4 shows a diagram for explaining the effects of the present invention. In the figure, 2 is a multi-gradation quantizer, 10 is a predictor, 12 is an exclusive OR, 3 and 4 are storage circuits, and 9. is a 1-line memory, and 5 is a ゜゜1r, ゜゜00゛ pattern detection circuit.

Claims (1)

[Claims] 1 Multi-gradation quantization is performed on each pixel of the facsimile signal, and then pixels whose multi-gradation quantization signal is above a certain value and pixels below a certain value are binary quantized into "white" or "black". , for other intermediate gradation pixels, binary quantization is performed to "white" or "black" to match the binary predicted value, and the intermediate gradation to which binary quantization should be performed is When the value of the next pixel following the pixel is "white" or "black" and is different from the predicted value of the pixel of the intermediate gradation, perform binary quantization of the pixel of the intermediate gradation. A facsimile signal quantization method characterized in that the predicted value and the value of the next pixel are set to a value that is closer to each other.