JP2003219436A - Apparatus for adjustment of pixel shift camera and measuring method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve accuracy in adjustment of a fine moving amount and remarkably reduce the time required therefor in a pixel shift camera. <P>SOLUTION: A pixel shift chart capable of differently handling the moving amounts of X and Y is employed to reduce the adjustment time of moving amount of the pixel shift camera, thereby searching for one-dimensional extreme values in two ways. Further, to reduce the time required for one trial and to avoid a problem of hysteresis, the present moving amount is measured not by repeatedly moving a CCD actually, but by repeating an arithmetic. Still further, to surely avoid the hysteresis, a series of fine moving sequences is performed in exactly the same way as that of normal photographing to pick up and store a series of moving images, and the respective moving amounts are measured. <P>COPYRIGHT: (C)2003,JPO

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a measuring method for adjusting a pixel shift camera. 2. Description of the Related Art A pixel shift camera is different from a digital camera, which has become popular in recent years, and is distributed only for business use, and is highly evaluated as a microscope image input device. The reason is that if the condition that the subject is stationary is satisfied as a condition, overwhelmingly high resolution can be obtained for the cost. In general, a CCD two-dimensional image sensor (hereinafter, referred to as a CCD) is configured so that information of each pixel can be taken out without loss by using an all-pixel reading type of a square pixel. By reading the information between the pixels by slightly moving the relative position by some method, an image having four to nine times the number of pixels of the CCD alone is constructed. However, the size of the pixel is about 4 to 7 μm, and the required movement size is one third to one half thereof.
Therefore, how to perform accurate measurement and adjustment was the biggest issue that affected the performance of this type of camera. For example, if a 2/3 inch 1.4 million pixel CCD is used, the distance to be moved is 3.35 μm, which is half the pixel size of 6.7 μm. Therefore, a high-precision displacement meter boasting an accuracy of ± 10 μm is useless. Several μm even with a laser positioning system
It is useless because it has an error of m. As a solution to this, the present inventor has previously disclosed in Japanese Patent Application Laid-Open No. Hei 9-135381 a CCD in an initial position.
Target video output signal (virtual moving image) equivalent to moving the target moving amount from the video output signal (origin image) of
And the relative movement amount (driving amount) is set such that the difference between the CCD image output signal (moving image) and the target image output signal (virtual moving image) is minimized while actually moving the incident light and the CCD relative to each other. To adjust the voltage). Although this method is epoch-making in that a discrete cosine transform and an inverse transform are used in a method for predicting a virtual moving image, there are serious problems in the required time and accuracy in actual operation. That is, if the movement amounts of X and Y are to be adjusted at the same time, a two-dimensional extremum search is performed, and a considerable number of trials must be performed until both of them reach a minimum error. Furthermore, if it is attempted to approach the target by repeating the process of actually moving the CCD and taking in the image as described above, the required time is further increased. Further, as a major problem, there is a problem of mechanical hysteresis possessed by the means for micro-moving. This problem is particularly serious when using a laminated piezoelectric element, and even when the same drive voltage is applied, the voltage is increased. The amount of movement differs depending on whether the voltage has reached the voltage or whether the voltage has reached the lower level. Therefore, there is a problem that the error is large and the adjustment cycle does not easily converge due to hysteresis.To avoid this, in each trial, return to the previous position each time and apply the target voltage again. And the time required is even greater. Further, in the above proposal, the subject is not particularly limited, and depending on the shape of the subject, even if the virtual moving image is accurately created, the result of the comparison may be the same regardless of whether X or Y is moved. There was a phenomenon that adversely affected the adjustment result. The problem to be solved is a drastic improvement in the accuracy of adjustment and a drastic reduction in the required time. According to the present invention, in order to shorten the time required for adjustment, a pixel shift chart which can separately handle the X and Y movement amounts is employed, thereby achieving the conventional two-dimensional pixel shift chart.
Instead of one-dimensional extremum search, one-dimensional extremum search is performed in two ways. Further, in order to reduce the time of one trial and to avoid the problem of hysteresis, the movement of the CCD is actually repeated and adjusted. Instead, in order to measure the current movement amount by repeating the calculation, and to ensure that the hysteresis is avoided, a series of minute movement sequences are performed exactly in the same way as normal shooting, and a series of movement images are obtained. The main feature is that the movement is collected and stored, and the amount of each movement is measured. [0008] The present invention has achieved the object of shortening the time required for adjustment and obtaining accurate measurement results only by proposing an optimal chart and improving the personal computer software. . FIG. 1 is a pixel shift chart of the present invention, wherein 1 is a vertical stripe portion and 2 is a horizontal stripe portion. At the time of use, the angle of view is set so that the entire chart including the blanks fits on the screen, and the photographing is performed by cutting out the center where only the vertical stripes and the horizontal stripes appear, and photographing. When the calculation is performed on the captured image, only the vertical stripes and the horizontal stripes are separately cut out, so that the movement amounts of X and Y can be separately handled.
In addition, fine objects and edges that can be seen or disappeared depending on the pixel shift cause measurement errors. Therefore, the brightness of the printing of the vertical stripes and the horizontal stripes smoothly changes in a sine curve. If fine particles are visible due to the printing machine, it is necessary to shift the focus of the lens to blur. FIG. 2 is a block diagram of the apparatus of the present invention. The CCD 5 is minutely moved in the vertical direction by the Y driving device 3 and minutely moved in the horizontal direction by the X driving device 4. The signal processing circuit 6 processes the video signal from the CCD 5 and
/ D conversion and output image data. The Y drive device 3 and the X drive device 4 are respectively driven by the voltages output from the drive voltage group storage device 7, and the drive voltage group storage device 7 determines the position of a regular series of minute movements for obtaining high resolution. X drive voltage data and Y drive voltage data corresponding to the indicated XY address are stored.
The voltage data at the address is called and a DC voltage is output via the D / A converter. Pixel shift sequencer 8
Outputs an XY address to the drive voltage group storage device 7 to control the order of the minute movement, and selects drive voltage data to be used. The image synthesizing device 9 is a device for performing normal photographing, and includes the pixel shift sequencer 8.
The image data is successively taken in while the CCD 5 is slightly moved via the CPU, and at the same time, appropriate pixel data is arranged at an appropriate position of the final image to construct a high-definition image. These are the blocks related to the normal operation. [0011] The following is a block relating to the adjustment for determining the X drive voltage data and the Y drive voltage data, and is a principal part of the present invention although it is not used in principle in normal operation. The drive voltage adjusting means 10 causes the origin image storage means 11 to store an image before a series of minute movements, if necessary, after pre-driving for removing hysteresis, and returning to zero voltage, and then the pixel shift The image data is successively taken into the moving image group storage means 12 while the CCD 5 is slightly moved through the sequencer 8. The origin image storage means 11 stores a vertical stripe portion in the origin image as X
The horizontal stripe portion is output to the calculating means 13 to the Y calculating means 15. The moving image group storage unit 12 stores each image separately for each XY address, and compares the vertical stripe portion of the storage image addressed by the drive voltage adjustment unit 10 with the X comparison unit 14.
Then, the horizontal stripe portion is output to the Y comparison means 16. The X calculating means 13 and the Y calculating means 15 perform the discrete cosine transform of the input image, substitute the phase coefficients for shifting the phase in the X direction and the Y direction, and perform the inverse comparison on the respective images. Output to the means 14 and the Y comparison means 16. At this time, the two calculation means obtain the calculation error information from the two comparison means, perform a calculation by changing the phase coefficient in a direction in which the error is reduced as compared with the previous calculation error, and calculate the phase coefficient by the drive The voltage adjustment means 10 is notified. The comparing means notifies the difference between the two input images to the calculating means and the drive voltage adjusting means 10 as the calculation error information. Further, the drive voltage adjusting means 10
Converting the phase coefficient when the calculation error information becomes substantially zero in both XY to a movement amount, comparing with the target movement amount of the XY address, calculating and predicting an optimal voltage as an applied voltage, and changing the applied voltage; This is because the voltage data of the XY address in the drive voltage group storage device 7 is rewritten. Discrete cosine transform and inverse transform are omitted because they are not the subject of the present invention. Since such an adjusting device and method are employed, the adjusting time is drastically shortened, and the accuracy is improved so as not to be comparable. Taking a case where shifting is performed 16 times at a 0.5 pixel pitch to obtain a high definition image as an example, the CCD is moved in the order shown in FIG. The squares in the figure indicate the movement positions at 0.5 pixel pitch intervals. The number of XY addresses corresponding to each position is 16, and according to the conventional method, adjustment of the two-dimensional extreme value search is performed 16 times.
For one extreme value search, an average of 60 pixel shifts and image capture are required even if the criterion is loosened so that convergence is easy, and it takes 960 times in total, which takes about 250 seconds. When this is performed by the adjusting device and method of the present invention, the adjustment is completed in 10 seconds or less. This is because, unlike the conventional method, the number of calculations is relatively large, but the number of pixel shifts and image captures is extremely small. Moreover, the shift method needs to be performed at the same timing as normal shooting. (Shift 16 times)
Is performed at once without any calculation, so that the time reduction effect is so great. Table 1 shows a comparison of required time. Regarding the number of operations of the present invention, since there are two types of measurements, X and Y, at 16 movement positions, the basic number is set to 32, and one measurement converges. Since the average number of computations before the calculation is 7 to 8, it is set to 8. Although the number of times the pixel shift sequence is rotated is 2 to 5, it is, on average, three times: the first time, after the voltage data is updated, and the confirmation. [Table 1] As described above, the pixel shift camera according to the present invention does not use any large-scale jigs or devices in adjusting the amount of pixel shift movement, which is the most important point.
Since processing can be performed in a short time using only the camera body, the personal computer, and the pixel shift chart, a low-cost, high-quality pixel shift camera can be provided.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a diagram showing a pixel shift chart. FIG. 2 is a block diagram illustrating a configuration of a pixel shift camera. FIG. 3 is an explanatory diagram showing an example of a pixel shift order. [Description of Signs] 1 Vertical stripe portion of pixel shift chart 2 Horizontal stripe portion of pixel shift chart 3 Y drive device 4 X drive device 5 CCD two-dimensional image sensor 6 Signal processing circuit 7 Drive voltage group storage device 8 Pixel shift sequencer 9 Image synthesis device Reference Signs List 10 drive voltage adjusting means 11 origin image storing means 12 moving image group storing means 13 X calculating means 14 X comparing means 15 Y calculating means 16 Y comparing means

Claims (1)

Claims 1. A CCD two-dimensional image sensor receives reflected and scattered light from a subject via a lens system, and
In a pixel shift camera, a plurality of images are taken by moving the CCD two-dimensional image sensor regularly in a two-dimensional direction parallel to a light receiving surface to take a plurality of images, and these are combined to obtain a high-resolution final image. In order to perform the adjustment, a pixel shift chart having a vertical stripe portion and a horizontal stripe portion near the center is prepared as a subject, and an origin image storage means for storing an origin image before the movement, and the image after the regular series of minute movements Moving image group storing means for storing a vertical stripe portion of the origin image, and an X calculating means for creating a virtual X moving image by performing a Fourier transform or a discrete cosine transform, performing a phase shift in the X direction, and performing an inverse transform. And comparing the vertical stripe portion of the image from the moving image group storage means with the virtual X moving image and changing the phase in the X direction until the difference is minimized. X comparison means for repeatedly starting the calculation means, and Y calculation for creating a virtual Y moving image by performing Fourier transform or discrete cosine transform on the image of the horizontal stripe portion of the origin image, shifting the phase in the Y direction, and then performing inverse transformation. Means for comparing the horizontal stripe partial image of the image from the moving image group storing means with the virtual Y moving image, and repeatedly activating the Y calculating means while changing the phase in the Y direction until the difference is minimized; And a method for adjusting a pixel shift camera, wherein the brightness of the printing of the vertical stripes and horizontal stripes of the pixel shift chart changes smoothly in a sine curve.