JP2009211959A - Image pretreatment device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method in which conventional technological problems are solved and a flexible response and a high speed toward algorithmic alterations of images by software processing are made possible. <P>SOLUTION: In order to solve the problems, a device is proposed in which a prescribed treatment is carried out by a line as a unit (scanning line as a unit) when an image is processed in a scanning electron microscope. As one of that embodiment, an image processing device is proposed in which a line unit processing part is installed, and treatments necessary for image formation such as integration, image improvement, and image reduction or the like are carried out in the line unit processing part, and a two-dimensional image is formed based on the line unit data processed with such treatments. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an image processing apparatus that processes image data and an electron beam apparatus.

  The image preprocessing device used in the electron beam device generates one image from the detection value of the detector and transfers it to the image processing device. In order to generate one image, a plurality of image processings are performed. Since the image processing algorithm largely depends on the acquired image, software processing is mainly used. This software processing and subsequent image transfer depends on the scan speed, which is the input speed from the detector. After the acquisition of one image, software processing and image transfer are performed before the input of the next image is completed.

  For example, when the scan speed is 30 images / sec, the scan time per image is about 33.3 ms, and after the acquisition of one image, image processing is performed during the next 33.3 ms. After one image is generated, transfer is performed to the image processing apparatus.

  The image size generated by the image preprocessing device varies because the size differs for each purpose. There are more than 10 types of vertical sizes alone, and if there are a similar number of horizontal sizes, there are more than 100 types of image sizes. Since the image size may be changed for each scan, the set time affects the processing time of the electron beam apparatus.

  The increase in the speed of the electron beam apparatus also leads to an increase in the throughput of the apparatus. Therefore, it is essential to increase the speed including the scan speed and the set time.

JP 2001-314881 A

  If the scanning speed is improved, the electron beam apparatus itself can be increased in speed. However, the increase in scan speed needs to be performed in a time 1 / n of the TV rate for image generation and transfer, for example, at a scan speed n times the TV rate.

  In order to increase the software processing speed, parallel processing is generally used, but parallel processing division and the overhead due to the division have a great influence.

  On the other hand, in the case of dedicated hardware that can be reliably increased in speed, it becomes difficult to cope with changes in the image processing algorithm.

  In addition, there is an array processor described in Patent Document 1, but basically, processing contents are parallelized, and the influence of overhead due to parallel processing is large.

  SUMMARY OF THE INVENTION An object of the present invention is to solve the above-described problems of the prior art and to provide a method that enables flexible response to image algorithm changes by software processing and high speed.

  In order to solve the above problems, an apparatus is proposed that performs predetermined processing in units of lines (units of scanning lines) when performing image processing with a scanning electron microscope. As one aspect thereof, a line unit processing unit is provided, and the line unit processing unit performs processing such as integration, image improvement, and image reduction necessary for image formation, and based on the line unit data subjected to the processing. An image processing apparatus for forming a two-dimensional image is proposed.

  According to the above configuration, since it is possible to perform processing in units of lines constituting one screen instead of one screen, it is necessary when a predetermined line scan is completed without waiting for one screen scan. As a result, it is possible to increase the speed of image formation.

  Hereinafter, an image processing apparatus that performs predetermined processing in units of lines will be described in detail. As an example of a specific configuration for that purpose, an image processing apparatus that generates an image composed of a plurality of lines from detection values of a detector of an electron microscope and transmits the image to an image processing unit that processes the image will be described below. An image input unit that generates each line from the detection value; a line unit processing unit that calculates an output of the image input unit; an output unit that transfers an output of the line unit processing unit to the image processing unit; It is composed of a switched fabric that connects the image input unit, the line unit processing unit, and the output unit, and a packet is passed through the switched fabric between the image input unit, the line unit processing unit, and the output unit. A description will be given of an image processing apparatus in which communication is performed and the packet structure includes line processing content, line structure, and line data.

  As a further example, the switched fabric has a shared memory accessed from the image input unit, the line unit processing unit, and the image output unit.

  As yet another example, a plurality of line processing units that perform operations on a plurality of lines are connected to the switched fabric.

  The image processing apparatus is divided into an input unit, an output unit, and a line unit processing unit for each element in the image preprocessing. At this time, there may be a plurality of line unit processing units. Pipeline processing is used between the line unit processing units to achieve parallelization in the time axis direction. At this time, each process is not performed for each screen but for each line constituting the screen. As a result, it is possible to have a margin in processing time because processing can be started when one line is completed without waiting for scanning of one image.

  By connecting the input unit, the output unit, and the line unit processing unit to the switched fabric, it is possible to increase / decrease the line unit processing unit with respect to the image algorithm change with the minimum necessary man-hours. For example, when the image thinning is newly performed and this processing is put in between the processes so far, the line unit processing unit that performs this image processing thinning is connected to the switched fabric, and the destination of the line unit processing unit in the preceding stage is By using the newly added line unit processing unit, the processing can be added. That is, the only change in the system is to change the destination of the line unit processing unit that transfers data to the newly added line unit processing unit.

  Furthermore, not only line data but also data contents and processing contents are added to the data structure of data transfer between processing units (input unit, line unit processing unit, output unit) divided into a plurality of processes, so that data is first passed. The image size can be set only in the input unit. As a result, there is no need to set the line unit processing unit and the output unit. Even when the number of line unit processing units increases due to processing subdivision or complication, or when the image size further increases, the electron beam by this setting is used. Does not affect the speed of the device at all. Even in the actual image change, even when the image is pipelined in time series, the setting can be made without having to be aware of the image breaks outside the input region.

  The same data structure may be used even when there are a plurality of inputs and outputs. In this case, if the line processing unit and the output unit do not use any data from other input units, the image pre-processing devices are simply arranged in parallel. In addition, when there is a line unit processing unit or an output unit that processes data from a plurality of input units, the line unit processing unit performs image composition, and the output unit transmits data to the same output destination. In image composition, if a flag is set for image composition of processing contents, the composition is performed when a corresponding line arrives, so the data structure is the same as in the case of single input. Since there are flags corresponding to the processing contents in the same way, it is only necessary to output both, and the data structure is the same.

  Furthermore, the image preprocessing system has a shared memory on the switched fabric, and has a structure in which image data including line data is shared by the input unit, the output unit, and the line unit processing unit. In this case, only the processing contents and the data contents in the data structure are exchanged between the respective processing units, and the data is read out from the address pointer in the shared memory described in the data contents and written back. In the case of n integration processing, for example, when the m-th processing of continuously flowing images is performed, it is necessary to store n images from m−n + 1 to m-th images. . By having a shared memory instead of a memory for each line unit processing unit, the structure of the line unit processing unit can be shared without depending on the data processing content, and the increase in image size is shared. This can be done by changing only the memory.

  Furthermore, the image pre-processing system has a multi-line unit processing unit that processes a plurality of lines, not a line unit process. For example, in the N × N filter process, N / 2 lines before and after the target line are required, and therefore the data structure is not passed to the next process until the N lines are prepared. Therefore, a buffer for storing a plurality of lines is required. However, in the shared memory structure according to the second aspect, a buffer for storing the processing contents and processing data for N lines may be used. Thereby, it becomes possible to cope with all image processing.

  In the image pre-processing device, the switching speed is used to divide the input unit, output unit, and line unit image processing unit into processing units, and the data structure between them is processing content, data content, and line data. Speedup, various image size settings in a short time, and image algorithm changes can be realized with minimal changes.

  Hereinafter, a first embodiment of the present invention will be described by taking an electron beam apparatus as an example with reference to the drawings. FIG. 1 shows an image preprocessing unit 102 that performs integration processing, image quality improvement processing, and image reduction processing, and an electron beam apparatus including the image preprocessing unit 102. The image preprocessing unit 102 includes a switched fabric 104, an input unit 105 connected thereto, a line unit processing unit 106 that performs integration processing, a line unit processing unit 107 that performs image quality improvement processing, and a line unit that performs image reduction processing. The processing unit 108 and the output unit 109 are configured.

  FIG. 2 shows a data structure transmitted and received between the input unit 105, the line unit processing unit 106, the line unit processing unit 107, the line unit processing unit 108, and the output unit 109 via the switched fabric 104. The data structure is composed of a flow 201, data content 202, and line data 203. The flow 201 is composed of a processing number 204, a processing order 205, an integration presence / absence 206, an image quality improvement presence / absence 207, and a reduction processing presence / absence 208. , An image size 209, a line number 210, and a data length 211. As the configuration contents of the flow 201 and the data contents 202, a pointer from the top of the data shown in FIG. The data structure itself is not fixed.

  A flow of a series of processes will be described.

  As shown in FIG. 1, the data output from the detector 101 is input to the input unit 105, and the flow 201 and the data content 202 are added to each line data and transmitted to the line unit processing unit 106. The content of the flow 201 is the processing content given from the overall control unit 100, and the data content 202 is the content given from the beam scan control part of the overall control unit 100.

  The flow of the line unit processing units 106, 107, and 108 will be described using the line unit processing unit 106 that performs integration processing.

  As illustrated in FIG. 3, the line unit control unit 106 receives reception data 301 that is data in line units from the input unit 105. The received reception data 301 is processed according to the flow 201. First, the flow 201 is taken out, and the presence / absence 206 of accumulation, which is processing of the line unit processing unit 106, is confirmed. If the integration flag is set in this integration presence / absence 206, the line data 304 is integrated to generate line data 305. At this time, the number of bits per pixel of the line data 305 and the line data length are calculated from the image size 209 and the data length 211. If the integration flag is not set, the line data 304 is used as the line data 305 as it is. Transmission data 303 is generated from the flow 201, the data content 202, and the line data 305, and transmitted from the processing order 205 in the flow 201 to the next processing of integration, that is, to the line unit processing unit 107 which is image quality improvement. Data 303 is transmitted. As shown here, integration presence / absence 206 indicating whether or not processing is performed, an image size 209 that is the size of the image, and a data length 211 that is an information source of a bit configuration per pixel are generated from the received data 301. In order to change these, only the change is set for the input unit 105, and it is not necessary to set for other processing units.

  Note that the flow of processing in the line unit processing unit 106 is the same in the line unit processing units 107 and 108, and only the processing performed on the line data 203 is different.

  The output unit 109, which is the final stage of the image preprocessing unit 102, receives line unit data transmitted from the line unit processing unit 109 and looks at the image size 209 and the line number 210 of the data content 202 to generate one image. Then, one image is transmitted to the image processing unit 103. In the case where the image processing unit 103 supports line-by-line transfer, data composed of at least data content 202 and line data 203 can be transmitted for each line. When the image processing unit 103 supports line-by-line transfer, transmission is possible with a delay of scan time + 5 lines as shown in FIG.

  When there is a change in the image algorithm, for example, when an image algorithm for performing the image quality improvement processing 2 shown in FIG. 5 is added, a line unit processing unit 601 is added to the switched fabric 104, and the flow of the initial setting file The image quality improvement 2 presence / absence 606 is added to 201, and the data content pointer 602, the line data pointer 603, the processing number 604, and the processing order 605 are changed and transferred to each line unit processing unit. When the number of line unit processing units is reduced, when the number of line unit processing units is reduced, basically the same as when the number of processing units is increased, and when the line processing unit is left, the processing number 204 and the processing order 205 are changed to correspond. Only the line processing unit is set to be skipped.

  As described above, according to the present invention, after the scan is completed, the processing of the preprocessing unit is completed with the line delay corresponding to the number of the line processing unit, the input unit, and the output unit, and the image can be transferred to the image processing unit. It is possible to cope with multiple image sizes and processing orders by simply changing the processing contents and data contents generated by the input unit, as well as for significant image algorithm changes that accompany changes in the line processing unit. It can be handled only by changing the file.

  Next, a second embodiment of the present invention will be described with reference to the drawings.

  FIG. 7 shows an image preprocessing unit 703 that performs integration processing, image synthesis processing, image quality improvement processing, and image reduction processing, and an electron beam apparatus including the image preprocessing unit 703. The image pre-processing unit 703 includes a switched fabric 715, an input unit 706 and an input unit 707 connected to the switched fabric 715, a line unit processing unit 708 that performs integration processing, a line unit processing unit 709, and a line unit processing unit that performs image synthesis. 710, a line unit processing unit 711 that performs image quality improvement processing, a line unit processing unit 712 that performs image reduction processing, an output unit 713, and an output unit 714.

  8 includes an input unit 706, an input unit 707, a line unit processing unit 708, a line unit processing unit 709, a line unit processing unit 710, a line unit processing unit 711, a line unit processing unit 712, an output unit 713, and an output unit 714. The data structure transmitted / received via the switched fabric 715 is shown. The data structure includes processing content 801, data content 802, and line data 803. The processing content 801 includes a data header 804, an execution offset 805, an integration processing content 806, an image composition processing content 807, an image quality improvement processing content 808, and a reduction processing. The data content is composed of an image type 810, an image size 811, an image number 812, a line number 813, and a data length 814. The processing unit connected to the switched fabric has an offset 815, a size of the number of bytes 816, and a configuration content of the data content 802, and reads the offset 815 and the number of bytes 816 at the head of the data structure. Since the offset and the number of bytes of the execution offset 805 are included here, each processing unit can know how many pairs of data header size, offset, double, and number are included. Therefore, since the last pair of the offset and the number of bytes indicates line data, the structure shown in FIG. 9 can be understood. The execution offset 815 stores the offset value of the process to be executed. For example, in the case of integration processing, the execution offset stores the offset value of the integration processing content 806.

  A series of flows will be described.

  As shown in FIG. 7, the data output from the detector A 701 is input to the input unit 706, and the processing content 801 and the data content 802 are added to each line data and transmitted to the line unit processing unit 708. The method of adding the processing content 801 is the same as that in the first embodiment.

  The operation of the line unit processing unit 708 will be described with reference to FIG. After receiving the reception data 301 which is data in line units, the head offset 815 and the number of bytes 816 of the data structure are read. As described above, this allows the data structure shown in FIG. 9 to be understood. Next, the execution offset value is called, and the head of the indicated offset value is called. Since the offset value of the integration process is entered here, the integration attribute value of the integration process content 806 is read. Since 2 is included in this, the next integration presence / absence and post-integration processing are included. The attribute value required for each process is included in the information distributed in advance as a setting value from the overall control unit 100 together with the size of the offset and the number of bytes.

  Processing is performed according to the flow 201 based on the integration processing content 806 which is a series of information. The processing method is the same as in the first embodiment. At the end of the processing, transmission data 303 is generated. At this time, the content of the execution offset 805 is rewritten to the post-integration processing of the integration processing content 806. Since the contents of the process to be sent next are included here, the transmission data 303 is transmitted based on this information to the line unit processing unit 710 that performs the image composition process.

  Similarly to the above operation, the detector B 702, the input unit 707, and the line unit processing unit 709 operate.

  Next, the operation of the line unit processing unit 710 for image synthesis processing that receives two inputs will be described. When the reception data 1001 shown in FIG. 10 is received, the data structure in FIG. 9 is recognized and the image composition processing content 807 is read out in the same procedure as the line unit processing unit 708. The number of combined images is read out, it is recognized that two lines are combined, and processing is waited until received data 1002 having a different image type 810 having the same number as the line number 813 is received. After the reception data 1002 arrives, the process is the same as in FIG. The number of inputs exceeding 2 inputs is basically the same as in the case of 2 inputs, and only the process is waited until the required number of lines is reached. Further, even if different line numbers 813 of the same image type 810 are required in time series, the process does not change to waiting for input, and basically the same flow is performed.

  The operation of the line unit processing unit 711 that outputs two transmission data will be described. The processing after receiving the received data 1101 shown in FIG. 11 is the same as that in FIG. 3 until data transmission. At the time of data transmission, two types of post-image quality improvement processing 1 and post-image quality improvement processing 2 are defined so that there is no image quality improvement processing content 808 in the processing content 1105 of the transmission data 1103. Referring to this, the data transmission in the flow 1104 recognizes that the data is transmitted to two places, and transmits transmission data 1102 and transmission data 1103 to each of them. Even when there are more than two outputs, only the item of the image quality improvement processing content 808 is increased.

  The line unit processing unit 712 that is an image reduction process has the same operation as the line unit processing unit 706 that is an integration process, and the output unit 713 and the output unit 714 recognize the processing content 801 as the line unit processing unit 706. The flow is the same as that of the output unit 109 of the first embodiment.

  If it is necessary to change the settings for the input unit 706 and the input unit 707 at the same time, it can be performed simultaneously using multicast communication of the switched fabric 715. As a result, it is possible to cope with a plurality of inputs / outputs in the same way as in the first embodiment without being conscious of the image size and the timing of processing switching due to the timing difference of the setting to each input unit.

  Further, the expandability is basically the same as that of the first embodiment, but the processing content 801 corresponds to the addition of a new function, and is dynamic for each image size for each line unit processing unit. It is possible to deal with changes in processing.

  As described above, according to the present invention, with respect to an image preprocessing apparatus having a plurality of inputs and outputs, the processing of the preprocessing unit is performed after the scan is completed with a line delay corresponding to the number of the line processing unit, input unit, and output unit. It is possible to transfer the image to the image processing unit after finishing, it is possible to cope with a plurality of image sizes and processing orders only by changing the processing content and data content generated by the input unit, and the line processing unit It is possible to cope with a large image algorithm change accompanied by an increase / decrease only by changing the processing content 801.

  This image pre-processing apparatus is an apparatus capable of creating one image from a one-dimensional image apparatus from a detector and performing complex image pre-processing for image generation in real time. Therefore, it can be used not for an electron beam but for an optical system.

The figure explaining the apparatus structure containing the 1st Embodiment of this invention. The figure explaining the data structure in the 1st Embodiment of this invention. Explanatory drawing of the line unit data processing part of 1st Example of this invention. FIG. 3 is a detail view of setting items in the first embodiment of the present invention. An explanatory view of line unit processing operation of the present invention. Explanatory drawing of expansion of the line unit data processing part in 1st Example of this invention. The figure explaining the apparatus structural example containing the 2nd Embodiment of this invention. The figure explaining the data structure in the 2nd Embodiment of this invention. Explanatory drawing of the data header part in the 2nd Embodiment of this invention. Explanatory drawing of 2 input line unit data processing part of 2nd Example of this invention. Explanatory drawing of 2 output line unit data processing part of 2nd Example of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Overall control part 101 Detector 102,703 Image pre-processing part 103 Image processing part 104 Switched fabric 105,706,707 Input part 106,107,108,601,708,709,710,711,712 Line unit processing part 109, 713, 714 Output unit 201 Flow 202, 802 Data content 203, 803 Line data 204, 604 Number of processes 205, 605 Processing order 206 Integration presence / absence 207 Image quality improvement presence / absence 208 Reduction processing presence / absence 209, 811 Image size 210, 813 Line number 211, 814 Data length 301, 1001, 1002, 1101 Received data 302 Line unit processing unit image processing flow 303, 1003, 1102, 1103 Transmission data 401 Processing content pointer 402, 602 Data content pointer 4 3,603 own processing pointer processing order pointer 405 processing content in the pointer 404 processing content of the line data (for line unit processing unit of the integration process)
406 Image size pointer 407 Line number pointer 408 Data length pointer 606 Image quality improvement 2 presence / absence 700 Overall control unit 701 Detector A
702 Detector B
704 Image processing unit A
705 Image processing unit B
801 Processing content 804 Data header 805 Execution offset 806 Integration processing content 807 Image composition processing content 808 Image quality improvement processing content 809 Reduction processing content 810 Image type 812 Image number 815 Offset 816 Number of bytes 1004 2 Input line unit processing unit image processing flow 1104 2 Output line unit processing section image processing flow

Claims (5)

Based on the output of the detector of the scanning electron microscope, an image is generated and processed in the image processing apparatus.
An image input unit that generates a signal in units of scanning lines from the output, a line unit processing unit that calculates an output of the image input unit, an output unit that outputs an output signal of the line unit processing unit, and the image A switched fabric connecting the input unit, the line unit processing unit, and the output unit;
Image processing wherein data is transferred through the switched fabric between the image input unit, the line unit processing unit, and the output unit, and the data includes processing contents, data contents, and line data apparatus. The image processing apparatus according to claim 1, wherein the switched fabric includes a shared memory accessed from the image input unit, the line unit processing unit, and the image output unit. In either claim 1 or claim 2,
An image processing apparatus, wherein a plurality of line processing units that perform operations on a plurality of lines are connected to the switched fabric.   In an image processing apparatus that forms an image by performing predetermined processing on data output from a detector of a scanning electron microscope, the predetermined processing is performed in units of scanning lines of the scanning electron microscope, and then the scanning line units An image processing apparatus characterized in that the data is synthesized to form an image. In claim 4,
The image processing apparatus characterized in that the predetermined processing includes at least one of integration processing, image improvement processing, and image reduction processing.

Images