JPH0340557B2 - - Google Patents

Description

[Detailed description of the invention]

[Technical Field] The present invention relates to an image transmission method for transmitting still image data using a public telephone line or the like for grayscale images. [Background technology] Character code transmission paths are generally composed of 8-bit units, but prior value prediction (DPCM) is usually performed.
Since the prediction residuals non-linearly quantized in the above method require 4 bits or more in order to obtain a practical restored image, there is a problem in that the compression effect is low. In addition, a method in which the previous prediction residual is further compressed in the time axis direction using a variable sampling density encoding method is extremely effective in increasing the compression effect, but on the other hand, There was a problem in that the quality of the image deteriorated due to errors caused by the process. Therefore, the present inventors proposed a method for transmitting the line-to-line difference between the sample of the current line and the predicted value from the restored image of the previous line using variable sample density encoding in a two-dimensional predictive quantization method within a frame (specially Gansho 59
-Refer to No. 130583). In this method, the distribution of the two-dimensional prediction residuals obtained concentrates on zero, so when variable sampling density encoding is performed using the quantization characteristics as shown in Table 2, the number of pixels is 256 × 256, and each pixel is processed in units of 8 bits. When image information was compressed to about 3 bits/pixel, a restored image comparable to that of 4-bit DPCM could be obtained. In addition, in variable sampling density encoding, the compression ratio can be increased by increasing the time difference value, but for edges that change rapidly, the edge is This sometimes caused position detection errors. On the other hand, in order to eliminate this edge position error, Table 1
If all the time difference values are 1 as shown in the figure, 4 bits/
Only pixels could be compressed.

【table】

【table】

[Table] [Object of the Invention] The present invention has been made in view of the above-mentioned problems. In image compression, the lower bits of the digital values of the restored image are allowed to have an amplitude error of about a certain standard value and suddenly change. It is an object of the present invention to provide an image transmission method that can reduce edge position errors and increase the compression rate. [Disclosure of the Invention] The method of the present invention provides a line buffer in a feedback loop inside a prediction circuit to perform primary prediction between lines, and also performs primary prediction between lines in a feedback loop outside the prediction circuit. It includes a compression quantization process and a decompression reproduction process that performs quadratic encoding or variable sampling density encoding of the residual in the line direction, and includes the above 1.
Compression coding is performed to the secondary prediction residual in the direction perpendicular to the next prediction residual, and the zero code, which is the code of the residual zero of this secondary prediction residual code, is compressed by the number of codes or the number of lines containing only zero codes. It is characterized in that the code is compressed and transmitted by run-length encoding, and the minimum step value of the amplitude difference value of the minimum compression quantization is made larger than the minimum step value of the digital value of the sample, and the The error distribution of the two-dimensional prediction within is concentrated at zero, and the range of the time difference value 1 of the quantization characteristic of variable sample density encoding is widened by compressing information at the encoding level by run-length encoding. It is possible to reduce the sudden edge position error caused by variable sampling density encoding, and also allow amplitude errors in the lower bits of the digital values of the restored image to increase the proportion of zero codes and increase the compression rate. The time difference value of the quantization characteristic can be made small so as to reduce the edge position error that is sharper than the edge position error. FIG. 1 shows an example of an apparatus implementing the method of the present invention. In the figure, a composite video signal output from a monochrome camera 1 is separated into synchronization signals by a synchronization separation circuit 2, and then sent to an A/D.
The data is digitized by the converter 3 and stored in the frame memory 5 by the data selector 4. The address stored in this frame memory 5 is controlled by the address control selector 13 controlled by the timing circuit 12.
To be determined. The data stored in the frame memory 5 is automatically compressed and encoded by the microcomputer 6, and sent to the public telephone line l via the microcomputer bus 19, modem interface 16, modem 7, and NCU 20. This device is used, for example, in a security system that mutually monitors abnormalities between a store and a residence.On the receiving side, a device similar to the one shown in the figure is used to generate an image using a D/A converter 8 and a synthesis circuit 9. Restoration of
Display on monitor TV10. Note that 14 is a program ROM of the microcomputer 6, 15 is a RAM for data processing, and 11 is a mode switching circuit. The frame memory 5 is for one screen, i.e.
In the case of 256 x 256 pixels and 8-bit gradation, it consists of 64 Kbytes, and reading and writing are performed at 200 nSec per pixel, including A/D and D/A conversion. FIGS. 2a and 2b are block diagrams showing the main parts of the transmitting side and the receiving side, respectively, of an embodiment of the system of the present invention. In the figure a, the digital value of the sample is multiplied by M1 by a predetermined value M1 by the multiplication circuit 21, and the prediction residual between lines is calculated by the difference circuit 22.
Furthermore, the variable sample density compression circuit 23 provided in the feedback loop outside the prediction circuit 25 quantizes the amplitude and time axis in the secondary prediction in the line in the direction perpendicular to the primary prediction. The intra-line prediction error code of the intra-line prediction residual of the inter-line prediction residual is compressed and transmitted by the zero-code compression circuit 24 by length-length encoding of the zero code. Here, the prediction circuit 25 multiplies a prediction coefficient, but if the prediction coefficient is set to 1.0, no multiplication is required and it is suitable for processing by a microcomputer. A line buffer 26 is provided in the feedback loop inside the prediction circuit 25 to perform primary prediction between lines. Prediction circuit 2
The variable sampling density compression circuit 23 and the variable sampling density expansion circuit 27 are provided in the feedback loop outside of the circuit 5. The variable sample density compression circuit 23 includes a line buffer 23a and a variable sample density encoding circuit 23, and the variable sample density expansion circuit 27 includes a line buffer 27a and a variable sample density decoding circuit 27b.
It is composed of. The variable sampling density expansion circuit 27 is for expanding the output code of the variable sampling density compression circuit 23, and the expanded output is further combined with the value of the line buffer 26 inside the prediction circuit 25 by the interpolation and synthesis circuit 28. It is designed to be input to a prediction circuit 25 for the next line. Here, when the sample interval is thinned out at a constant interval (for example, when the sample interval is thinned out every 4 pixels to make 64 x 64 pixels), the interpolation synthesis circuit 28 interpolates the interval to 256 x
This is to restore it as a 256 pixel image.
For example, if the sample is a 256 x 256 pixel area that has changed between frames and is thinned out every 4 pixels, the changed area is transmitted as a coarse image, and the background 256
When displaying a composite image on top of a detailed 256-pixel image, the changing parts are interpolated. Figure 3 illustrates an example of a variable sampling density encoding method, in which the relationship between the sampling period and the sample value is defined by a triangle, and by moving this triangle as shown in the figure, the difference is shown by the arrow. By transmitting this, on the receiving side, the sampling interval is determined from this difference value and the triangle, and the original waveform is restored. The smaller the variation in sample values, the longer the sampling interval becomes and the more data is compressed. On the receiving side in FIG. 2b, there are a zero code expansion circuit 29 that expands the received encoded image signal to zero codes, and a variable sample density decoding circuit 30a that decodes the prediction residual between lines. and a variable sample density expansion circuit 30 consisting of a line buffer 30b.
, an interpolation synthesis circuit 31 that performs interpolation synthesis similarly to the transmission side, a prediction circuit 32, and a line buffer 33.
comprising a decoding loop and a division circuit 34,
The image for one line is decoded, and the digital value of the decoded image signal is divided by a predetermined value M1, which is the same as the scale factor on the transmission side, in the division circuit 34 to restore the sample. If the scale factor is set to a power of 2, a shift operation in binary registers is sufficient, so the sample of the image with 6 bits per pixel is multiplied by 4, and 1
By performing calculations with the same precision as an 8-bit pixel sample, the calculation precision is effectively less than a decimal point, so
Image quality can be improved. However, using the standard quantization characteristics of variable sample density encoding shown in Table 1 above, it is possible to compress a 256 x 256 pixel sample of 8 bits/pixel to approximately 4 bits/pixel, and furthermore, Table 5 For example, if 12 pixels in one line are zero pixels (zero code), the code of 4 bits x 12 = 48 bits is converted to F3 in hexadecimal by using the zero code compression conversion table of the zero code compression circuit 24 shown in
It can be compressed into an 8-bit code. However, with the combination of these tables, the compression effect using the zero code compression circuit 24 cannot be expected; however, in the present invention, the quantization characteristics shown in Table 3 are used. In other words, Table 3 shows a characteristic in which the minimum step value of the amplitude value of the quantization characteristic is four times the minimum step value of the digital value of the sample, and using this characteristic, compression by the variable sample density method and Table 5 As a result of zero-code compression using Therefore, there was no edge position error, and a higher compression ratio than the combination of Tables 1 and 5 was obtained. Furthermore, when the quantization characteristics in Table 4 with a large time difference value are combined with the zero code compression in Table 5, the image becomes approximately 2
It can be compressed to bit/pixel, the amplitude error can be made smaller than 4, and the edge position error can be made smaller than 2, resulting in a higher compression ratio than the combination of Tables 2 and 5. Also, set the digital value of the specimen to a predetermined value.
Scale up by M1 to perform predictive compression calculations with finer precision than the minimum step value of the digital value of the sample, and multiply the amplitude error by M1 times the step value of the minimum amplitude difference value of the quantization characteristic. By increasing the step value of the minimum bit of the operation below,
Along with improving the compression ratio and reducing edge position errors, calculation accuracy and image quality have improved. FIGS. 4a and 4b show another embodiment. In this example, prior value prediction (DPCM) is performed when all the amplitude difference values of the quantization characteristics of the example in Fig. 1 are set to 1.
Figure a shows the transmitting side, and a compression circuit 2 for prior value prediction (DPCM) is installed in the feedback loop outside the prediction circuit 25.
3' and an expansion circuit 27' are provided.
The next prediction is performed, and the line buffer 26 in the feedback loop inside the prediction circuit 25 is provided at the same position as the prediction circuit 25, and the second prediction within the line is performed in a direction perpendicular to the first prediction. The receiving side in FIG. 4b differs from the circuit in FIG. 1b in that a DPCM expansion circuit 30' is provided in place of the variable sample density expansion circuit 27. However, in this embodiment, by encoding the inter-line prediction residual with intra-line DPCM and zero-code compression by the zero-code compression circuit 24, if the quantization characteristics shown in Table 3 are used, the result is 3 bits/pixel equivalent. I was able to compress it. Furthermore, the use of DPCM has the advantage of reducing the line buffer used, making it easy to implement hardware, and facilitating high-speed compression.

【table】

【table】

【table】

【table】

[Table] [Effects of the Invention] As described above, in the present invention, a line buffer is provided in the feedback loop inside the prediction circuit to perform primary prediction between lines, and a line buffer is provided in the feedback loop outside the prediction circuit. It includes a compression quantization process and an expansion reproduction process that performs secondary encoding or variable sample density encoding of the primary prediction residual between lines in the line direction, and secondary prediction in the direction perpendicular to the primary prediction residual. At the same time, the residual is compressed and encoded, and the zero code, which is the code for the zero residual of this second-order predictive residual code, is compressed by run-length encoding using the number of codes or the number of lines containing only zero codes, and then transmitted. , and since the minimum step value of the amplitude difference value of the minimum compression quantization is made larger than the minimum step value of the digital value of the sample,
A high compression ratio can be achieved, and if the time difference value of the quantization characteristic is made small and the amplitude difference value is made large, there is an effect that the compression effect and the edge position error can be reduced at the same time.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing one embodiment of the present invention;
2A and 2B are block diagrams of the main parts of the same as above, FIG. 3 is an explanatory diagram showing the operation of the same as the above, and FIGS. 4A and 4B are block diagrams of the main parts of another embodiment. 1 is a monochrome camera, 2 is a synchronous separation circuit, 3 is an A/D converter, 4 is a data selector, 5 is a frame memory, 6 is a microprocessor, 7 is a modem, 8 is a D/A converter, 9 is a synthesis circuit, 1
0 is a monitor TV, 11 is a mode switching circuit, 21 is a multiplication circuit, 25 is a prediction circuit, 22 is a difference circuit, 2
3 is a variable sample density compression circuit, 23' is a DPCM compression circuit, 24 is a zero code compression circuit, 26 is a line buffer, 27 is a variable sample density expansion circuit, and 27' is a
DPCM expansion circuit, 29 is zero code expansion circuit, 3
0 is a variable sampling density expansion circuit on the reception side, 30' is a DPCM expansion circuit on the reception side, 32 is a prediction circuit on the reception side, 33 is a line buffer on the reception side, 34 is a division circuit, and M1 is a predetermined value.

Claims (1)

[Claims] 1. A line buffer is provided in a feedback loop inside the prediction circuit to perform primary prediction between lines, and a primary prediction residual between the lines is provided in a feedback loop outside the prediction circuit. It includes a compression quantization process and a decompression playback process that performs secondary encoding or variable sample density encoding in the line direction, and compresses and encodes this into a secondary prediction residual in the direction perpendicular to the primary prediction residual. The zero code, which is the code with zero residual of the second-order predictive residual code, is compressed by run-length encoding using the number of codes or the number of lines consisting of only zero codes, and is transmitted, and the amplitude difference of compression and quantization is the minimum. An image transmission method characterized in that the minimum step value of the value is larger than the minimum step value of the digital value of the sample. 2. Multiply the digital value of the sample by a predetermined value at the input section of the prediction circuit to multiply the minimum step value of the sample by the predetermined value, and also multiply the minimum step value of the amplitude difference value of compression quantization by the original value of the sample. 2. The image transmission method according to claim 1, wherein the step value is set to be larger than the original minimum step value of the sample by a predetermined value times an allowable amplitude error of the sample.