JPH0118606B2 - - Google Patents

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a method for compressing binary image data.
In particular, the present invention relates to a data compression method suitable for storage in an inspection system that stores a printed wiring board pattern and compares it with the stored contents, for example, in a printed wiring board pattern defect inspection system.

[Technical background]

Regarding the compression method of binary image data, various methods related to facsimile etc. have been proposed in the past, but in the inspection system of printed wiring board (hereinafter referred to as PWB), especially the inner layer inspection of multilayer PWB, the edge information of the pattern is This is important, and it is difficult to apply a data compression method that uses a mixture of graphics and characters, such as in facsimile, as is.

For example, as shown in the case of PWB inspection in Figure 2, the image obtained from one imaging device 2 by scanning the PWB 1 is decomposed into ²ⁿ pixels and data processing is performed. To obtain data, first, the analog signal of 2 ⁿ pixels from the image pickup device 2 is converted into an m-bit digital signal by the A/D converter 3. Next, in the spatial filter processing binarization circuit 4, the digital signal is subjected to, for example, 5×5 spatial filter processing, thereby converting the digital signal into approximately determined black and white binary data.
Obtain valued data. Then, in the next second spatial filter processing circuit 5, 1-bit unevenness is corrected by removing 3×3 quantization noise, for example, to obtain definite black and white binary data. This binarized data is compressed by the run length method in the next data compression circuit 6 and stored in the memory 7.

Then, the output signal (binary image signal I) from the second spatial filter processing circuit 5 which scanned the second image of the PWB 1 and processed as described above, and the output signal from the memory 7.
The first data signal of PWB1 is compared with the pixel data restored by the data restoration circuit 16.

The same number of A/D converters 3, spatial filter processing binarization circuits 4, second spatial filter processing circuits 5, data compression circuits 6, memories 7, and data restoration circuits 16 as there are imaging devices 2 are required.

By the way, in front of the data compression circuit 6, 5x5 two-dimensional spatial filter processing is performed (here, the surrounding two pixels become uncertain), and 3x3 two-dimensional spatial filter processing is performed (here, the surrounding two pixels are uncertain). As a result, as shown in FIG. 3, for the pixel data area 8 where black and white are determined, black and white data areas of 3 pixels each are created around the periphery in the main scanning direction and the sub-scanning direction. 9, 10, 11, 12 occur. Assuming that the data areas 9 and 10 in the sub-scanning direction are not subjected to data processing, and focusing on the uncertain data areas 11 and 12 in the main scanning direction, these data 11 and 12 are inherently uncertain as data. It can be seen that it is preferable to exclude the data from data compression and not to store it in the memory 7 from the viewpoint of data processing and memory saving. However, since the imaging device 2 uses a power of ²ⁿ , that is, a 2-bit image sensor, for example, a one-dimensional CCD image sensor with 2 ¹¹ = 2048 bits, the It is conceivable that control related to timing etc. becomes complicated and affects processing speed. Therefore, in the above example of the uncertain data 11 and 12, one of the problems is how to handle the 3 bits of the starting point pixel group and the 3 bits of the ending point pixel group in the main scanning direction. On the other hand, the original purpose of data compression is to reduce the capacity required for the memory 7 as much as possible, in other words, how to devise a data compression method in the data compression block 6. It becomes a problem.

[Purpose of the invention]

The present invention has been made against the above background, and its basic object is to provide a novel binary image data compression method that can improve the data compression rate and at the same time improve the data processing speed. In addition,
It should be added that the compression method used in facsimile does not suit the nature of the image that is the subject of this invention.

[Structure of the invention]

The main purpose of the present invention is to divert and utilize the run-length encoded data required for the starting pixel group as uncertain data generated by two-dimensional filter processing, and to change the starting pixel group to black or black. By forcing either white to replace the necessary data indicating whether to start with "1" or "0" as cue data, the end point pixel group is set immediately before this end point pixel group. By treating these uncertain data as continuation data of the definite data and making them run-length encoded data, run-length encoded data that is required only for the end point pixel group is unnecessary. The basic feature is that

In this specification, "black" corresponds to the L level of the binary level signal, that is, the logic "0",
On the other hand, "white" corresponds to the H level of the binary level signal, that is, logic "1".

〔Example〕

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be specifically described below with reference to embodiments shown in the accompanying drawings.

FIG. 1 is an explanatory diagram of the principle of an embodiment of a PWB inspection system similar to FIG. 2. In FIG. 1A, S indicates data of the starting point pixel group, E indicates data of the end point pixel group, and V indicates a continuous pattern of "black" and "white" pixel data in the final data area. These continuous patterns include a series of only black (a-1 in the same figure), a series of only white (a-2 in the same figure), a series of black, black, white, white... (a-3 in the same figure), a series of black and white, black and white...
There are four consecutive ways (a-4 in the same figure). For any of these four patterns, the starting pixel group S is all black. In other words, a continuous pattern is defined as one that always starts with black. In this way, when black finishes, white will always follow, and there is no need to specify white. Therefore, white/black (1/0) flags are not necessary and can be expressed only by run-length encoding.

On the other hand, all of the end point pixel group E is equalized to the pixel data immediately before the end point pixel group E. If the immediately preceding pixel data is black, it is treated as black, and if it is white, it is treated as white. Of course, in the case of a-1, the color is black, in the case of a-2, it is white, in the case of a-3, it is black, and in the case of a-4, it is white. In the case of a-3 and a-4, the immediately preceding pixel data may be opposite black and white, but of course all of the end point pixel group E are made the same black/white.
With this definition, the run-length encoded data of the data of the end point pixel group E is included in the description of the run-length encoded data of the immediately preceding data, and the number of words can be reduced.

In addition, in the cases of a-1 and a-3, the run-length encoded data of the data of the starting pixel group S is included in the subsequent black run-length encoded data, which also reduces the number of words. can be done.

FIG. 1B is a schematic diagram of the word structure for the run-length encoded description used in this embodiment. One word consists of 5 bits, the first 4 bits (2 ²
The bits) are for run-length encoding description, and the LSB is the terminator T. Terminator T is "0"
If it is "1", it means that the run length description ends, and if it is "1", it means that the description continues.

Since the embodiment uses a one-dimensional image sensor of 2 ¹¹ =2048 bits, it can be expressed with a maximum of three words, and in the case of a three-word configuration, it has the structure b-3 in FIG. 1B. This is a structure used, for example, when describing a-1 in FIG. 1A. With 3 words, 4096 bits of pixel data can be run-length encoded.

For 2 words, as shown in Figure b-2, the 4 bits of the first word and the 4 bits of the next word total 8.
Bits can represent up to ²⁸ = 256 pieces of identical data.
Similarly, for one word, as shown in b-1 of the same figure,
It can represent up to 4 bits, ie, up to 2 ⁴ =16 identical data.

1 terminator T separated by "0"
A word or two-word or three-word run-length encoding description becomes a black or white run-length description. The distinction between black and white can be determined by whether "0" of the terminator T is an odd number or an even number from the start of one line. That is, odd numbers are black and even numbers are white.

Now, if the run length is expressed as a variable length of 4, 8, or 12 up to 3 words using a 5-bit fixed word, then the pattern in Figure 1 Aa-1 becomes B in the same figure.
becomes b-3 ("11111 11111 01110"), a-2
The patterns are b-1 (“00100”) and b-3
(“11001 11111 01110”) combination a-3 and a
The pattern -4 is a combination of b-1, b-2, and b-3.

Typical compressed data is shown in FIG. 1C. Figure c
-1 is when black continues in the starting pixel group and the black is for 14 pixels or more (when the starting pixel group and the ending pixel group are each 3 pixels), and the first
This corresponds to cases a-1 and a-3 in Figure A. Also,
When the number of pixels is 13 or less, it becomes c-2 in the figure. Figure c-
3 is a case where white continues after 3 pixels in the starting pixel group, and corresponds to cases a-2 and a-4 in FIG. 1A,
The first word indicates black for three pixels.

Note that length 3 is expressed as "0010" here. If a method is used in which the starting pixel group is forced to be black as in this embodiment, there is no need to define the length of 0, which is usually done in facsimile, and the numerical expression is the value of n-1. I can express it. Specifically, length 1 is "0000", length 2 is "0001", length 3 is "0010", etc.
Therefore, if the length 16 is expressed, it will be "10000", but according to this method, it will be "1111", and 4 bits will be sufficient.

Furthermore, the above embodiment has a configuration particularly suited to a system for inspecting high-density PWB. Assuming that a 2048-bit one-dimensional CCD image sensor images an area of 32 mm in the main scanning direction and separates it into 16 x 16 μm pixels, the line width of the above board is
When the width is 100 to 200 .mu.m and it crosses a line, the number of pixels of the line is 16 .mu.m.times.16=256 .mu.m, so it can be expressed with less than 16 pixels, that is, with 4 bits. Therefore, in the worst case, even if a large number of lines are crossed, the number of words will not increase.

Furthermore, since the substrate base portion is overwhelmingly wider than the PWB wiring pattern, the starting pixel group is forced to be black, that is, the data “0” in the substrate base portion. Due to the nature of the deterministic data (A in Figure 1) having a low probability of starting from white, if we do the above, the probability that the run-length encoding of the starting pixel group will be absorbed into the subsequent run-length encoding of black. becomes higher, and the efficiency of data compression increases accordingly. In this way, in the above embodiment, the data compression rate is improved by skillfully utilizing the basic properties of PWB (properties related to standard basic grids, macroscopic properties, etc.). In addition to forcing the starting pixel group to be "black", it can also be set to "white" depending on the imaging target.
In some cases, it may be forced to be ``black'', and is not limited to ``black''.

Next, FIG. 4 shows an outline of a circuit specifically implementing the above data compression method.

FIG. 4 is a diagram showing details of the data compression circuit 6 shown in FIG. 2, and FIG. 5 is a timing chart thereof.

In FIG. 4, 21, 23, 24, 25 are AND gates, 22 is a D flip-flop (D-
FF), 26 is an OR gate, 27 is a run length conversion circuit, and 28 is a switch.

As shown in FIG. 5, the starting pixel group area signal S is 3
The final data area signal V is for 2042 pieces, and the end point pixel group area signal E is for 3 pieces, for a total of 2048 (=
2 ¹¹ ) pixel signals.

In FIG. 4, "0" is set via switch 28.
When a signal is given to the AND gate 23, even if the other input starting point pixel group area signal S of the AND gate 23 is "1", the output of the AND gate 23 becomes "0", and the final data area signal V and the end point Since the pixel group area signal E is also "0", the outputs of the AND gates 24 and 25 are also "0", and the OR gate 2
Since all three inputs of 6 are "0", the output of the OR gate 26 is "0", and when the synchronization signal CK rises and the scanning pixel effective area signal R is "1", the run length encoding circuit At 27, data to be run-length encoded is captured.

Next, when the confirmed data area signal V becomes "1",
The AND gate 24 is opened and the binary image data I enters the run length encoding circuit 27 via the AND gate 24 and the OR gate 26.

The synchronization signal is sent to the D-FF 22 via the AND gate 21 while the final data area signal V is "1".
CK enters. At the rise of the synchronization signal CK to the D-FF22, the 2 input to the D input terminal of the D-FF22
A value image signal I ('1' or '0') is held at the Q output. Every time the synchronization signal CK enters the D-FF 22, the Q output is replaced with a new binary image signal I ("1" or "0").

Then, when the final data area signal V becomes "0", the final signal ("1" or "0") of the binary image signal I when the final data area signal V was "1"
is held in the D-FF22. Then, the end point pixel group area signal E becomes "1", and the AND gate 2
5 is opened, and the Q output of the D-FF 22 is input to the run length circuit 27 via the OR gate 26.

The run-length circuit 27 takes in the output signal of the OR gate 26 at the rising edge of the synchronization signal CK when the scanning pixel effective area signal R is "1" and performs run-length encoding. The run-length encoding shown in FIG. 1B and the restoration of the run-length encoded signal (the data restoration circuit 16 in FIG. 2) can be performed using known techniques, and therefore will not be described in detail.

In FIG. 4, when a "1" signal is applied to the AND gate via the switch 28, when the starting pixel group area signal S is "1", the "1" signal is sent to the run length circuit 27 via the AND gate 23 and the OR gate 26. ” will be given three times.

In this embodiment, the number of starting pixel groups and the number of ending pixel groups are explained as three, but they can be increased or decreased depending on the setting of the size of the spatial filter, and the present invention is not limited to three. Needless to say.

Note that the printed wiring board pattern data of the starting point pixel group S and ending point pixel group E in FIG. , in this example
By moving the image by 32 mm, the imaging device 2 performs sufficient overlap (in units of 1 mm) so that the portion from the starting point pixel group S to the ending point pixel group E in the forward sub-scanning enters the fixed data area in the backward sub-scanning. It's getting old. There is no printed circuit portion to be inspected in the non-overlapping starting pixel group S of the two imaging devices 2 at both ends and the non-overlapping ending pixel group E at the opposite end.

In this way, by overlapping, it is possible to scan the entire area of the printed circuit to be inspected, and by performing back-and-forth sub-scanning, the number of imaging devices 2 required is half the number required for the full width in the main scanning direction. The printed circuit board 1 returns to the starting position by reciprocating, and there is no deterioration in operating efficiency due to the reciprocating.

[Effect]

When restoring a run-length encoded signal, it is desirable that the number of pixel signals is 2 ⁿ , and as in the above embodiment, the number of pixel signals is 2 ¹¹ = 2048 by including the corresponding signals of the starting point pixel group and the ending point pixel group. This will make it easier to handle when restoring. That is, when restoring 2042 signals after converting them into run lengths without using this invention, there is a commonly used device that treats the signals corresponding to the start pixel group and the end point pixel group as 2 ⁿ pieces. 2048 signals must be inserted before and after, or before and after, 2048 signals and passed to the next stage circuit.

The 2042 signals are run-length, and a circuit is required to determine or detect where to insert the signals corresponding to the start pixel group and end point pixel group, and the timing is also required, making the restoration circuit complex. .

In the present invention, such a problem does not occur because signals in place of the starting point pixel group and the ending point pixel group are inserted during run length conversion. moreover,
The next 2048 pieces of data (2049th piece) is the starting pixel data of the next scanning line and is forced to black, so the pattern shown in Figure 3 has the advantage of being treated as continuous run-length encoded data. It also has

〔Effect of the invention〕

As described above, the present invention handles the starting point pixel group and the ending point pixel group as uncertain data generated by two-dimensional filter processing in the same way as definite data, so data processing based on powers of 2 is performed. This is easy to do, improves processing speed, and forces the starting pixel group to be black (or white), so there is no need to use a black/white flag or define a length of 0. Since there is no data, there is no increase in the number of words, and if the data after the starting pixel group is black, the number of words can be reduced and the data compression rate can be improved.Furthermore, since the ending pixel group is the same data as the previous data This has the effect of reducing one word of data with a high probability and further improving the data compression rate. Furthermore, by forcibly inserting the data of the starting point pixel group and the ending point pixel group, there is an effect that the restoration of the run-length signal can be handled easily.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram of an embodiment of the present invention, FIG. 2 is an explanatory diagram of a block according to the embodiment, and FIG. 3 is an explanatory diagram of generation of uncertain data based on two-dimensional filter processing.
FIG. 4 is a block diagram of a specific circuit applied to one embodiment, and FIG. 5 is a timing chart of FIG. 4. S, 11...Start pixel group, E, 12...End point pixel group, V...Fixed data, T...Terminator,
1... Printed wiring board (PWB), 2... Imaging device, 3... A/D converter, 4... Spatial filter processing binarization circuit, 5... Second spatial filter processing circuit, 6... Data Compression circuit, 7...Memory, 2
1, 23, 24, 25...and gate, 22...
...D flip-flop (D-FF), 26...OR gate, 27...Run length encoding circuit.

Claims (1)

[Claims] 1. A target pattern is photoelectrically scanned, an analog image input signal obtained by scanning is converted into a 1-bit binary signal, and the data of the binary signal is data compressed by run-length encoding. In this case, the analog image input signal is first A/D converted to m bits, subjected to two-dimensional filter processing, and then converted to a binary signal, and
For the starting point pixel group and the ending point pixel group as uncertain data generated by the two-dimensional filter processing,
A binary image data compression method characterized by forcing a starting point pixel group to be either black or white, and performing run-length encoding so that an end point pixel group is equalized to data immediately before the end point pixel group.