JP2011047808A - Method and apparatus for measuring image - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To achieve an apparatus for subjectively evaluating moire generated on a flat display panel. <P>SOLUTION: On a screen of the display panel 6, there are formed many display elements. On an image sensor built in a digital camera 1, there are formed many imaging pixels. The imaging pixels of the image sensor are smaller than the display pixels of the display panel 6. A plurality of points of the display pixels of the display panel are imaged by the imaging pixels of the image sensor. With respect to all display pixels of the display panel 6, a brightness evaluation is performed by a personal computer 5 to synthesize moire patterns. Thus, only the moire of the display image can be evaluated without being affected by moire between the display panel and the image sensor. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an image measurement method, and more particularly to an image measurement method and an image measurement apparatus that can objectively evaluate moire.

  In 3D displays without glasses, moire occurs, so measures are needed to suppress it. Therefore, in developing 3D displays without glasses, it is important to measure not only the moire contrast but also the overall shape of moire and fine texture. For practical use of 3D displays, there is often a compromise design between moiré and resolution, and the development of a method for simultaneously measuring not only moiré but also resolution is desired.

  Furthermore, it is necessary to suppress moiré caused by interference between the prism array of the backlight and the periodic structure of the liquid crystal cell even in a normal liquid crystal display (hereinafter referred to as “LCD”). Establishment was important. However, there was no appropriate means to measure and quantify the direction of moire stripes, contrast, pitch, overall shape, and even fine textures.

  In a 3D display using a two-layer LCD, two liquid crystal panels are placed on the back and front and separated by a certain distance to change the image brightness ratio between the front and back liquid crystal panels. Thus, an illusion that a stereoscopic image exists between the front and back liquid crystal panels can be given, and a three-dimensional stereoscopic image can be displayed.

  In this apparatus, two liquid crystal panels are stacked, and the magnification that appears on the retinas of the two panels is slightly different due to the difference in parallax. Therefore, moire is caused by the interference of the periodic structure of the two panel images formed on the retina. Will occur. In order to reduce this, a cross lenticular lens, a fly-eye lens array, or a diffusion layer is used. As a result, although the moire is reduced, the resolution is deteriorated. For example, “Non-Patent Document 1” or “Non-Patent Document 2” describes countermeasures against moire for a two-layer display. In addition, “Patent Document 1” is cited as a configuration in which a moiré countermeasure is described in a two-layer display device.

  In addition, the development of an autostereoscopic 3D display using a single-layer type lenticular lens has been widely promoted. In this 3D display, a lenticular lens can be attached to a liquid crystal panel, a PDP (plasma display panel), or an organic EL panel, and a three-dimensional image can be viewed without glasses. Recently, a multi-view 3D display in which this lenticular lens is obliquely attached to the panel pixels has been announced.

  As described above, whether it is a two-layer type or a single-layer type, a 3D display without glasses often requires moiré countermeasures, which sacrifices the resolution of the FPD, such as the addition of a light diffusion layer. We often rely on countermeasures. Therefore, the design and development of a 3D display without glasses requires simultaneous evaluation of moire and resolution. In addition, in order to compare the superiority and inferiority of different types of 3D displays, it is important to simultaneously measure and quantify moiré and resolution.

  Conventionally, in measuring moiré or resolution, a spot luminance measuring device is translated to measure a one-dimensional luminance distribution, and the resolution is obtained from the moire contrast or amplitude modulation degree from the distribution. However, although the measuring apparatus is large and expensive, it is difficult to measure the weak moire buried in the noise of the image by the moire measurement method based on the spot luminance measurement. For this reason, although the contrast of a strong moire can be measured, there is a problem that a weak moire that can be visually recognized by human eyes cannot be evaluated. In addition, this method has a problem that the overall shape of moire and a fine texture cannot be expressed.

  In the case of a two-layer 3D display, a spot luminance measurement device is used to measure a two-layer 3D display from an infinite point when measuring in parallel movement. Cannot measure. The display image can be optically imaged, and the spot luminance measuring device can be translated from the imaged image for measurement. However, this is a larger and expensive device. Moreover, in this case as well, it is obvious that the overall shape and fine texture of the moire cannot be expressed, and it is obvious that only a moire with a very strong contrast can be evaluated, and establishment of a new moire measurement method has been desired.

  Furthermore, it has been desired to develop a versatile method capable of measuring moire generated due to various causes in a flat panel display (hereinafter referred to as FPD). In particular, moiré caused by interference between the backlight of a liquid crystal display (LCD) and the liquid crystal cell structure is measured and quantified, including the direction of the stripes, contrast, pitch, overall shape of the moire, and fine texture. There is no appropriate means to achieve this, and it has been desired to realize a new moire measurement method capable of measuring these.

  By the way, it is conceivable to measure the moiré by taking FPD images of LCD, plasma display panel (hereinafter PDP), organic EL display devices, etc. with a digital camera (hereinafter digital camera), but the image sensor pixel structure and FPD pixel structure There is a problem that unnecessary moire appears in the captured image. No matter how much the zoom lens magnification is changed or the imaging distance is changed, the interference between the pixel structure of the image sensor and the FPD pixel structure remains, and it can only be used for imaging high-contrast moire with a large pitch. There wasn't.

Patent 3,335,998

K.Oku, et al., “Analysis and Reduction of Moire in Two-Layered 3D Display”, SID2007 Digest, pp.437-440, 2007. K.Oku, et al., “Ergo 3D Display”, ED electronic display 2008 conference, Feb.27-28 2008, Proceedings & Conference materials. Supervised by Hiroshi Miyagawa “Evaluation Technology for Television Images”, The Institute of Television Engineers, 91-93.

  As explained above, since there was no method for accurately quantitatively assessing weak moiré, moiré evaluation has to rely on the human eye, and the number of subjects increased, and statistics such as evaluation scale method and series category method increased. Moiré could only be quantified by practical treatment. Note that the evaluation scale method and the series category method are described in “Non-patent Document 3”, and it is necessary to increase the number of subjects to ensure objectivity, and it is difficult to implement a troublesome method.

  For this reason, there has been no report that the moire and resolution of 3D displays without glasses that have different methods have been evaluated. In addition, various types of 3D displays without glasses have been proposed in the past, but their moiré and resolution characteristics cannot be compared uniformly on the same ground, and therefore, a unified standard for 3D displays cannot be established. Development and popularization of 3D displays has been hindered. Measurement of moiré is troublesome, and establishment of an evaluation method for weak moiré is important for the development and popularization of 3D displays, and development of the evaluation method is necessary.

  As described above, for the development of a 3D display and the like, development of a new measuring means for expressing not only the contrast of moire but also the overall shape and fine texture of moire has been desired. Further, there are many compromise designs between moire and resolution, and it has been desired to develop a method for measuring not only moire but also resolution at the same time. Furthermore, it is desired to put to practical use a simple and easy-to-use moire measuring device for quantitative evaluation and countermeasures for moire generated by the prism structure of the backlight and the periodic structure of the LCD pixels.

  An object of the present invention is to realize an imaging method of an FPD display image in which the occurrence of moire between the FPD pixel structure and the image sensor pixel structure is suppressed in order to solve the above various problems simultaneously. In addition, by using an FPD display image capturing method that suppresses the occurrence of moire, a simultaneous moire and resolution measurement method using digital camera image data is realized. This is to realize an evaluation method that objectively evaluates moire without depending only on human senses.

  The above problems are to obtain the position of each FPD pixel accurately with respect to the x and y pixel numbers of the image sensor, to obtain the average value of the internal brightness of each FPD pixel, and to determine the FPD pixel structure and the image sensor. This can be realized by obtaining an image in which the occurrence of moire with the pixel structure is suppressed. In another aspect of the present invention, a newly developed test chart group used to realize the present invention also constitutes a part of the present invention.

  It is possible to realize a measuring device that measures the FPD moire and resolution simultaneously with a digital camera shot. In addition, it is possible to provide an inexpensive measurement device that improves the efficiency of design and development when FPD is applied to a 3D display, and enables a unified standard for 3D displays without glasses.

  That is, in the conventional moire or resolution measurement, a spot luminance measuring apparatus is moved in parallel to measure a one-dimensional luminance distribution, and the resolution is obtained from the moire contrast or the amplitude modulation degree from the distribution. However, this method has a problem that the overall shape and fine texture of the moire cannot be expressed, and the apparatus is expensive. Further, spot luminance measurement has a problem that there is a lot of noise, and a moire signal with low contrast is buried in the noise level, making measurement difficult.

  By using the present invention, a digital camera shot, a tripod for fixing the digital camera, and a PC incorporating the image processing program developed this time can obtain the moire contrast or amplitude modulation degree (resolution) with a digital camera shot. Further, since the image data of the two-dimensional distribution is integrated in a certain direction and a line spread function is used, the moire signal is not buried in the noise level, and measurement can be performed even with a weak moire.

  Therefore, by using the present invention, an inexpensive and easy-to-use moiré and resolution simultaneous measurement device can be put into practical use. In addition, the practical use of this device makes it possible to uniformly compare the moiré and resolution characteristics of 3D displays without glasses of various methods in the same manner, and contribute to the promotion of the development and popularization of 3D displays.

It is a schematic diagram of the moire evaluation apparatus of this invention. It is a flowchart of the moire and resolution measurement in this invention. It is a stripe & gray scale chart in the present invention. It is a cross hatch chart. It is a white chart. It is a black chart. It is an example of a line spread function (LSF) in the x direction. This is the cumulative increase / decrease number of the x-direction luminance distribution. This is the cumulative increase / decrease amount of the luminance distribution in the x direction. It is the figure which connected the cross point calculated | required with the gravity center with the broken line. This is the standard deviation that estimates the measurement error at the cross point. It is a cross hatch lattice point used for deriving the pixel position of the FPD. This is an inner point that approximates the integral of luminance in the FPD cell with a sum. It is a cross hatch chart after image processing. It is a white chart after image processing. It is the stripe & gray scale chart which processed the image. This is the relationship between the gray scale level and the average luminance within the scale. This is the relationship between the gray scale level connected by a smooth curve and the average luminance within the scale. 6 is an image-processed stripe and gray scale chart subjected to γ correction. It is the processed image which carried out (gamma) correction of white chart imaging. This is a change in luminance in the horizontal direction in the vertical stripe pattern. This is a change in luminance in the vertical direction in the horizontal stripe pattern. It is a calculation result of an amplitude modulation degree. It is a verification white chart including a unidirectional moire. It is the processing image which imaged the white chart for verification containing the unidirectional moire. It is a figure explaining the definition of a moire evaluation function. It is the calculation result of the standard deviation with respect to an angle. It is a figure which shows the relationship between an angle and a line spread function. It is a figure which shows the coordinate dependence of the line spread function in the case of an angle of 30 degree | times. This is a measurement result of the direction of moire, contrast, and pitch at each location on the screen. It is an extraction result of a moiré pattern. It is a chart for verification including two-way moire. It is the process image which imaged the chart for verification containing 2 direction moire. It is a calculation result of the standard deviation with respect to the angle in a two-way moire. It is a figure which shows the relationship between the angle in 2 direction moire, and a line spread function. It is a figure which shows the coordinate dependence of the line spread function in an angle when a standard deviation takes the maximum value in 2 direction moire. It is a figure which shows the coordinate dependence of the line spread function in an angle when a standard deviation takes a 2nd magnitude | size in 2 direction moire. It is a measurement result of the direction, contrast, and pitch of moire in the first main direction at each location on the screen. It is a measurement result of the direction, contrast, and pitch of moire in the second main direction at each location on the screen. It is an extraction result of a moiré pattern.

  Hereinafter, the contents of the present invention will be described in detail by way of examples.

  In the following, FPD display images are captured while suppressing the occurrence of moiré between the FPD pixel structure and the image sensor pixel structure due to the imaging of the four test charts shown in FIGS. 3 to 6, and slight moiré such as 3D display is performed. An outline of a procedure for obtaining and displaying the moiré contrast, the moiré pattern, and the resolution (amplitude modulation degree) in a certain image will be described.

  FIG. 1 shows an embodiment of a resolution and moire evaluation apparatus according to the present invention. The apparatus of the present invention is composed of a digital camera 1, a tripod 2, a remote shutter 3, an image data transfer cable 4, a PC 5 for image processing, evaluation, processing image display, and evaluation result display. In FIG. 1, reference numeral 6 denotes a flat display panel which is a target for evaluating moire. FIG. 2 shows a flowchart of resolution and moire measurement using the apparatus of FIG. FIGS. 3 to 6 show examples of test charts used for resolution and moire measurement.

  This chart is designed and developed by the inventor for measuring resolution and moire. A test chart can be output using a test chart output program according to the size of the area to be evaluated. Note that data for test chart output is also input in a moire / resolution evaluation system, and image processing is performed according to the test chart used for measurement. The moire / resolution evaluation system can output test charts, and can handle any image format. However, if the images of FIGS. 3 to 6 are displayed on a part of the FPD screen and the moiré and resolution at that part are evaluated, the same test chart can be used even if the image formats are different. it can.

The procedures (1) and (2) of (I) test chart imaging (PC external processing) in the flowchart are shown below. When taking an image, the image recording mode is set to JPEG or RAW (preferably RAW), the manual mode is set, the flash is turned off, and the shutter and the camera are used remotely.

  (A) Display the stripe & gray scale chart on the FPD screen where you want to evaluate, place the camera fixed on the tripod at a distance of about 40-50cm from the FPD screen, and display the surrounding area in the image on the digital camera monitor. Set the zoom ratio and position of the digital camera so that the square frame just fits. Adjust the focus as needed. This is done by using a function that automatically adjusts the approximate focus, automatically scans the vicinity of the position, and detects an accurate focus position. In many cases, the zoom ratio of the digital camera can only take a discontinuous value, so that the position of the tripod needs to be finely adjusted back and forth in order to make the surrounding square frame just fit in the image.

  (B) From this state, change the FPD display to a white chart, turn off the autofocus, set the F value to a high resolution of the optical lens, and display the indicator with the digital camera brightness level at the center when displaying white Set the shutter speed so that

  (C) Also, change the FPD display to the stripe & gray scale chart and adjust the focus again. This is done using a function that automatically scans and detects the exact focus position.

  (D) The camera position, zoom ratio, F value, shutter speed, and focus determined as described above are fixed, and the four types of stripe & gray scale chart, cross hatch chart, white chart, and black chart in FIGS. The pattern is imaged sequentially. At the time of imaging, there is no movement of people around, and the position of the human body part when pressing the shutter is fixed over the details. This is to make the reflections of the surrounding light the same as much as possible. This is because the reflection of surrounding light is corrected by subtracting image data obtained by black chart imaging, but it is necessary to prevent the reflection from changing when the four patterns are imaged. Note that when using a dark room, it is possible to suppress the reflection of light from the display image from entering the FPD screen and reflecting the incident light again, so that the black chart imaging may be omitted.

Next, the image data (JPEG or RAW data) of the above chart is transferred to a personal computer. Furthermore, as shown in the flowchart (III) of FIG. 2, the moiré between the FPD pixel structure and the image sensor pixel structure is further improved by the image processing method devised and developed this time (FIG. 2 (II) (3) ). An FPD display image with reduced generation is taken, and a moiré contrast and a moiré pattern are obtained from an image with slight moiré such as resolution (amplitude modulation degree) and 3D display, and these are displayed.

  The image processing method devised and developed this time will be described below. In actual image processing, the cross hatch pattern JPEG or RAW data is first used to determine the cross position of the cross hatch pattern, the position of each pixel of the FPD is calculated, and the average value of luminance within each pixel is determined. To obtain images with minimal moiré. A procedure for calculating the position of each pixel of the FPD will be described. The numbers in parentheses below correspond to the numbers in the flowchart of FIG.

(4) Conversion of imaging data from binary data to real luminance distribution data
The RAW image data is expressed in 65536 gradation of two-dimensional distribution in which x and y take discrete values (or in 256 gradation in the case of JPEG data). This image sensor image data has coordinates (x, y) that take only discrete values (x _i , y _i ) on an equidistant grid, so that I, (x _i , y _i ) = I _y Assume that j starts from 1 and takes values up to the upper limit (i _max , j _max ) of the number of pixels in the x and y directions. Further, continuous coordinates (x, y) give coordinates such that the lower left corner of the lower left pixel of the raster is (0, 0) and the upper right corner of the upper right pixel of the raster is (i _max , j _max ).

This (x, y) is used for the calculation inside the program, but in the graph display of the calculation result, the center of the image is (X, Y) = (0, 0), and the left and right edges Use standardized coordinates (X, Y) such that X = ± 1 and Y = ± 1 at the upper and lower ends. Distinguish between uppercase normalized coordinates (X, Y) and lowercase coordinates (x, y). Similarly, the normalized grid point coordinates (X _i , Y _i ) are determined for the grid point coordinates (x _i , y _i ).

When discrete coordinates (x _i , y _i ) are expanded to continuous coordinates (x, y), I (x _i , y _i ) = I _y (x, y) is displayed. However, since I (x _i , y _i ) = I _y is 65536 (= 2 ¹⁶ ) gradation (from 0 to 65535), I (x, y) is set to I (x _i , y _i ) = I _y It is defined by a value normalized by dividing by 65535 (or defined by a value normalized by dividing by 255 for JPEG data).

(5) Noise reduction of imaging data (Gaussian distribution convolution)
Since image data I (x, y) has a finite gradation and variations in pixel sensitivity of the image sensor, quantization noise and random noise (white noise) are superimposed. In many cases, the calculation process does not work because it is difficult to say that it is a continuous function. In order to avoid this difficulty, a noise removal method using a low pass filter (LPF) by convolution of Gaussian distribution is adopted. That is, noise removal is performed using the image distribution Id (x, y) subjected to LPF processing defined by the following equation.

Here, R _x and R _y are 1 / e radii of a Gaussian distribution in the x and y directions. By adjusting these radii, a program created assuming a continuous function will not malfunction. This radius may be about 3 to 6 times the pixel pitch of the image sensor for searching the FPD pixel position. The x and y coordinates are also discrete values (x _i , y _i ) as the center of the image sensor cell. Numerical integration was performed for each cell of the image sensor. The integration for each cell was performed using a 16th-order Gaussian integration, and the integration in the x and y directions was performed up to the image sensor cell covering an ellipse having a radius three times that of R _x and R _y . Other than that, the Gaussian distribution is small, so it contributes little to the integration, so it is omitted.

(6) Approximate calculation of image sensor position corresponding to each cross point of the cross hatch chart First, from the luminance distribution Id through LPF in the vicinity of the coordinate axis at the normalized coordinate (X, Y), x Alternatively, integration over a range in the y direction (actually sum: integration over a range of ± 10% of the entire image) to obtain an averaged line spread function (LSF), the cross hatch lattice spacing, The approximate position of the center line on the top and bottom of the cross hatch and the left and right sides, and the number of black lines on the X and Y axes contained in the captured image are obtained. FIG. 7 shows the line spread function LSF (X) in the x direction. In this figure, the portion where the value on the vertical axis is large is the white portion of the cross hatch, and the portion where the value is small is the black portion of the cross hatch. Further, x corresponds to discrete normalized coordinates X _i corresponding to the local minimum value. It can be seen from FIG. 7 that the white portion with high luminance is still noisy and the black portion is less susceptible to noise. The reason for making the cross hatch pattern black on a white background instead of white on a black background is to avoid the influence of noise as much as possible.

In order to determine the local minimum point, the local minimum point is determined by using the indicators of cumulative increase / decrease in FIG. 8 and cumulative increase / decrease in FIG. From the luminance distribution Id cumulative count change N (X) is through the LSF 7 in FIG. 8, by sequentially moving the point from X _i has a low to large, before the value of the luminance at that point It is obtained by comparing with the value at the point. The initial value of N (X) is 0, and the difference in luminance is 0 or negative. If N (X) at the previous point is negative or 0, N (X) at this point is the previous point. If N (X) at the previous point is positive, N (X) is reset to 0. Similarly, if the luminance difference is 0 or a positive value and N (X) at the previous point is positive or 0, N (X) at this point will set 1 to N (X) at the previous point. Add and if N (X) at the previous point is negative, N (X) resets to 0. This procedure was repeated to obtain an index N (X). It was determined that the point with the smallest value of N (X) within the range of ± 25% of the center of the captured image corresponds to the lattice point closest to the Y axis.

  It can be seen that the value of N (X) is oscillating at a fine pitch in the range of ± 5 in the vicinity of 0 on the vertical axis, but this is determined by the relationship between the test chart and the number of pixels of the image sensor. ing. The reason why the vibration is oscillated at a fine pitch in the range of ± 5 is that the size of the test chart is set so that about 10 × 10 image sensors are included in one FPD pixel. In other words, in the imaging of the white part, since the FPD cell is darkest near the BM and the center of the cell is brightest, the brightness generally increases with five image sensors corresponding to half of the FPD cell, and the other half of the image sensor 5. This is because the brightness decreases with the individual (although there are exceptions everywhere).

  In FIG. 9, when X is changed from a small value to a large value, when N (X) is continuously the same sign (excluding 0), the luminance difference is added, and N (X) If the sign changes, the cumulative increase / decrease amount H (X) reset to 0 is indicated. The x shown in FIG. 7 is the result of searching for the local minimum value in the vicinity when H (X) becomes smaller than a certain value (−HCUT). The numerical value of this HCUT is usually about 0.1 to 0.15, but there is no problem. However, if the signal vibrates greatly near 0 on the vertical axis or the signal level is too dark, it is necessary to adjust the value. There is. This HCUT value is entered by the program. The above is for the X direction, but the Y direction number of the image sensor closest to the center of the crosshatch test chart is similarly obtained for the Y direction.

  Next, pixel numbers i (k, l) and j (k, l) of the image sensors at the respective cross points are obtained. Here, (k, l) represents the number of the cross hatch cross point, and (k, l) = (0,0) is the cross point in the center. Rough cross position information i '(0,0), j' (0,0) at the center of the screen of the cross hatch pattern obtained in the above procedure, the approximate interval Δn (in units of the cross hatch near the center of the screen) Is the pixel pitch of the image sensor: the average value of four intervals in the ± X and ± Y directions from the center of the screen) as the initial value, and the pixel number i (k, l), j (k, l)

The program starts with a more accurate pixel from the approximate pixel number i ′ (0,0), j ′ (0,0) of the image sensor at the cross point at the center of (k, l) = (0,0). The numbers i ′ (0,0) and j ′ (0,0) are obtained. Integrate in the x or y direction within the range of ± 2/3 of the approximate spacing Δn of the cross hatch near the center of the screen (this value 2/3 can be selected as the input data of the program). Spread functions LSF _X (x) and LSF _Y (y) are obtained, and i ′ (0,0) and j ′ (0,0) are obtained by the same algorithm as described above.

  Next, the rough pixel number of the cross point of (k, l) = (0,1) is set to i ′ (1,0) = i ′ (0,0) + Δn, j ′ (1,0) = j ( 0,0), and an accurate pixel number i (0,1), j (0,1) is obtained by the above algorithm. Thereafter, i ′ (k + 1,0) = i (k, 0) + [i (k, 0) −i (K−1,0), j ′ (k + 1,0) = j (k, 0): As k = 2, 3, 4,..., accurate i (k + 1,0), j (k + 1,0) is obtained by the same algorithm as described above.

  Further, i '(0,1) = i (0,0), j ′ (0,1) = j (0,0) + Δn, and i (k, l), j (k , l): k = 0, 1, 2, 3,. After that, i ′ (0, l + 1) = i (0, l), j ′ (0, l + 1) = j (0, l) + [j (0, l) −j (0, l−1)] Similarly, i (k, l + 1), j (k, l + 1): k = 0, 1, 2, 3,..., L = 2, 3, 4,. With the above procedure, i (k, l), j (k, l): k = 0, 1, 2, 3,..., L = 0, 1, 2, 3,. The same procedure is repeated in the second quadrant, the same procedure is repeated by connecting the third and fourth quadrants, and i (k, l), j (k, l) in all quadrants: k =... -3, -2 , −1,0,1,2,3,..., L = ...− 3, −2, −1,0,1,2,3,.

The above is the rough calculation of the image sensor position corresponding to each cross point of the cross hatch chart.
Note that the cumulative increase / decrease amount H (X) when X increases at each crossing point is negative and minimum, but the value is h ⁽⁺⁾ (k, l), and similarly when X decreases. Let h ^(-) (k, l) be the minimum negative value of the increase / decrease amount. Further, h ⁽⁺⁾ (k, l), h ⁽⁻⁾ (k, l), the smaller absolute value is defined as h (k, l). The h (k, l) is obtained in the x direction and the y direction, and stored as h _X (k, l) and h _y (k, l), respectively. The index x x (k, l) of the brightness x of each cross point and the depth index h _X (k, l), h _y (k, l) of the cross point x _c (k, l) l), y _c (k, l) is used for calculation.

(7) Accurate calculation of the image sensor position corresponding to each cross point of the cross hatch chart The image sensor pixel number i (k, l), j (k) corresponding to the cross position of the cross hatch pattern with the prescription of (6) , l): k = ...- 3, -2, -1,0,1,2,3, ..., l = ...- 3, -2, -1,0,1,2,3, ... However, in order to increase the accuracy of the cross position, the coordinates x _c (k, l) and y _c (k, l) of the cross position are obtained.
Specifically, the luminance distribution Id that has passed through the low-pass filter LPF around the image sensor numbers i (k, l) and j (k, l) at each cross point is ± (2/3) Δn in the y or x direction. Integration is performed in the range of (rounded down after the decimal point), and line broadening functions LPF _X (x) and LPF _Y (y) in the x and y directions near each cross point are obtained.
Next, for the x coordinate x _c (k, l) of each cross point, the index h _X (k of the valley depth in the x direction is calculated from the minimum value LSF _{X min} of the x-direction line spread function LPF _X (x). , l) minus 1 / e or more h _x (k, l) (1-1 / e), that is, LSF _{X min} −h _X (k, l) (1-1 / e) Is the reference value, LPF _X (x) inverts the part below the reference value with the positive and negative inversion, the valley is the mountain, the center of gravity of the mountain is obtained, and the coordinate x _c (k, l) of the cross position is obtained (here e Is the base of the natural logarithm (= 2.71828), but it is not necessarily limited to that value). That is, the x coordinate x _c (k, l) of each cross point can be expressed by (Expression 2).

Here, x ₁ and x ₂ are coordinates of x at which −LSF _X (x) + LSF _{X min} −h _X (k, l) · (1-1 / e) = 0. Here, the value of LPF _X (x) or LPF _Y (y) is given on the grid points of x and y corresponding to the cell position of the image sensor, but the other x and y points are linear or cubic. It is assumed that x _c (k, l) is obtained as a linear function at both ends of (a) a piecewise linear function and (b) a piecewise cubic function. For the y coordinate y _c (k, l) of each cross point, the y-direction line spread function LPF _Y (y) and the y-direction valley depth index h _y (k, l) at each cross point are used. Obtain the same formula as equation (2). However, the calculation results were not significantly different in either case (a) or (b).

From this, the accuracy of position detection is not the exactness of the integration of (2), but rather the noise superimposed on the luminance distribution (difference in sensitivity for each pixel = fixed pattern noise, amplifier 1 / f noise, thermal noise, photodiode It can be understood that it is dominated by dark current, shot noise, etc. Therefore, the condition (a) that shortens the calculation time as much as possible is adopted (either (a) or (b) can be selected in the program). FIG. 10 displays a lattice figure connecting the cross points x _c (k, l) and y _c (k, l) thus obtained.

In order to verify the calculation accuracy of the lattice points, the adjacent 3 × 3 lattice points are targeted, and from x _c (k, l), y _c (k, l) at the four corner lattice points, x, The y position is a function of the parameters s and t, and is expressed by a first-order polynomial for each of s and t. s = 0, t = 0 is the lower left lattice point, s = 1, t = 0 is the lower right lattice point, s = 0, t = 1 is the upper left lattice point, s = 1, t = 1 is the upper right lattice point From the four grid point positions, the x and y positions of the other five grid points were obtained from the interpolation function of a first-order polynomial and compared with the x and y positions obtained from the center of gravity.

  FIG. 11 shows a result obtained by performing this operation with combinations of all 3 × 3 lattice points adjacent to the lattice points in FIG. 10 and obtaining standard deviations of errors in the x and y coordinates. In terms of standard deviation, the cross point is obtained with an error within 5% of the image sensor pixel pitch (an error within 0.5% of 1/10 of the FPD pixel pitch). It can be judged that considerably high accuracy was obtained even though the error was judged by strict evaluation.

(8) Calculation of luminance average inside each FPD pixel
For the derivation of the FPD pixel position, interpolation of the cross hatch x, y coordinates with (a) first order, (b) second order, (c) third order, and (d) fourth order polynomials was considered. FIG. 12 shows cross hatch lattice points used for interpolation in each case.

  The shaded part is the range to be interpolated. The lower left corner is s = 0, t = 0, the upper right corner is s = 1, t = 1. It is obtained by interpolating the position coordinates (x, y) of the image sensor with (i, j) expanded to real numbers.

The coefficients a _nm and b _nm of the polynomial in this equation are the image sensor position coordinates x _c (k, l), y _c (corresponding to (N−1) × (N−1) crosshatch lattice points. k, l) can be obtained uniquely. The cross hatch black line is two or four FPD pixels thick, so the coordinates x _c (k, l), y _c (k, l) of the center of gravity correspond to the end face of the FPD pixel. Yes. Although the cross hatch pattern can be generated at an arbitrary interval, in the example of FIG. 4, the interval is 10 FPD pixels.

  If the cross-hatch cross point is 10 FPD pixels, substitute 0, 1/10, 2/10, 3/10, ..., 9/10, 1 for s and t in equations (3) and (4). Then, the position of the FPD pixel located between the cross hatch lattices is obtained. Furthermore, an arbitrary position in the FPD pixel can also be obtained by Expressions (3) and (4). By substituting values for s and t in 1/10 units, the outer frame of the FPD pixel can be obtained, but if values other than 1/10 unit are substituted, the coordinates of the image sensor position corresponding to any point in the FPD pixel x and y can be obtained.

  Next, using the mapping of equations (3) and (4) and interpolation with the polynomial of the (x, y) coordinates of the luminance distribution of the image sensor, the average luminance in the pixel = the sum of the light emitted from the pixel ( However, the luminance distribution corresponding to each FPD pixel is obtained. One pixel of the FPD is divided into small rectangular subcells of 20 × 20, for example, and the average luminance in the pixel is obtained by dividing the sum of luminance values at the center of the subcell (dots in FIG. 13) by the number of subcells. . The number of subcells can be selected arbitrarily, but the accuracy is almost saturated at about 20 × 20.

  Although the calculation was performed by changing the upper limit order N of the polynomial from 1 to 4, even if the upper limit order was increased from N = 1, the result was hardly affected. It is estimated that the noise superimposed on the signal from the image sensor dominates the calculation error. If the noise from the image sensor is greatly reduced, a higher-order interpolation function will be meaningful, but N = 1 is sufficient if the current digital camera is used. The above is a mapping method in which the outer frame position of each pixel of the FPD, an arbitrary point in the pixel, and the coordinates (x, y) obtained by realizing the pixel number (i, j) of the image sensor correspond one-to-one. It is.

  Further, the luminance with respect to the coordinates (x, y) of the image sensor is expressed in a format similar to Expressions (3) and (4) using values on grid points corresponding to the pixels of the image sensor as shown in FIG.

The coefficient c _nm can be expressed using the luminance I (x _i , y _j ) = I _ij of the original image that is not subjected to LPF processing on the grid points corresponding to the pixels of the image sensor. Here, the calculation was performed by changing the upper limit order M of the polynomial, but even if the upper limit order was increased from M = 1, the result was hardly affected. It is estimated that the noise superimposed on the signal from the image sensor dominates the calculation error.

  FIG. 14 shows an image obtained by performing the above processing on the cross hatch image. FIGS. 15 and 16 show processed images of data obtained by imaging the stripe & gray scale chart and the white chart of FIGS. It can be seen that a processed image in which the generation of moiré is suppressed as much as possible is obtained.

15 and 16, the luminance distribution I _B (x, y) of the black chart image is subtracted, and the local average value around the cross point (the average is around the cross point in the x and y directions). This was done in the range of ± Δn / 2) to correct the difference in sensitivity at each location of the image sensor. In other words, the black level correction and the shading correction are performed simultaneously (however, when the image is taken in a dark room, the black chart is not taken, so the black level is not corrected). From the local average around the grid-like cross points, the local average brightness including other than the grid points of the white chart and black chart

If the local average values at the center of the screen are I _MW0 and I _MB0 , the correction value I ′ (x obtained by performing black level correction and shading correction for an arbitrary image I (x, y) in Expression (5) , y) is expressed as follows.

  The charts of FIGS. 3 and 5 have a + mark, but this + mark is at the same position as a part of the cross of the cross hatch. The cross points marked with + are obtained by the same method as the cross hatch cross points, and when the images of FIGS. 3 and 5 are taken from the difference of the cross hatch cross points and when the images of FIG. 4 are taken. This corrects a slight misalignment between the digital camera and the monitor. Using a set of cross points located at the four corners of the rectangle, the position shift at the cross point of the cross hatch in the rectangle is interpolated with a first-order polynomial of horizontal coordinates and vertical coordinates, and corrected. . However, the points outside the cross points are corrected by extrapolation.

(9) Optimization of RGB image weighted overlay coefficient (required when using RAW data)
In the case of using 256-gradation JPEG image data, there was no problem when the RGB image was superimposed 1: 1: 1, but 65536 tone RAW data using the RAW image was 1: 1: 1. When superposed on 1, a moire with a large pitch was generated although it was light. This moire is determined to be invisible if the mixing ratio of RGB images is appropriately selected, and the moire is reduced by automatically adjusting the overlay coefficient.

When an RGB image is I _R (x, y), I _G (x, y), I _B (x, y), the three images are basically two parameters except for the normalization constant. Can be overlapped. If the coefficient of _R is W _R , the coefficient of _G is W _G , and the coefficient of _B is WB,

However, it is not W _R , W _G , W _B for optimization of the superposition coefficient,

The image of G must be included, the superposition coefficient W _G is 1,

Optimize the values of W _{R / B} and W _{RB / G} to minimize the standard deviation from the average luminance value, and minimize the occurrence of moiré between the FPD pixel structure and the image sensor pixel structure. Such operations are automatically performed in the program. When optimizing the mixing ratio, select the 1 / e radius of the Gaussian distribution to be convolved as R _x = R _y = 3 with the FPD cell pitch as the unit for the processed image of the white chart, and reduce the noise to optimize. It has become. The Davidson-Flecher-Powell method, which is a kind of quasi-Newton method, was used for this optimization. In this case, the optimized values of W _{R / B} and W _{RB / G} are (0.163, 0.448).

In the optimization, since the calculations (5) to (8) in the flowchart of FIG. 2 are repeated many times without outputting image data, a considerable calculation time (CPU time) is required. However, if the digital camera to be imaged and its aperture / shutter speed, FPD filter, or phosphor, and the brightness are fixed, the optimized W _{R / B} and W _{RB / G} values are reproducible and optimized every time. do not have to.

(10) Calculation of grayscale average luminance and calculation / display of γ-corrected image Comparing Fig. 16 with Fig. 3 of the original image, it can be seen that the level close to black in gray scale is crushed. . In order to correct this, the gamma characteristic is determined. 3, the gray scales above and below the stripe are displayed in 256 gradations, and when the black level is 0 and the white level is 255, the levels are 0, 16, 32, 48, 64, 80. 96, 128, 176, 224, 240, 255. A value obtained by dividing this value by 255 is obtained on the horizontal axis, and the average luminance in the gray scale cell is obtained on the vertical axis, and is shown by a broken line in FIG. The reason why there are two broken lines in FIG. 17 is that there are two pairs of gray scales on the chart.

  Although this graph is a polygonal line, the two polygonal lines are averaged, each point is interpolated with a cubic spline function, the vicinity of the black level is approximated with an exponential function, and the γ characteristic is represented by a smooth function. FIG. 18 shows the result. When expressed by a smooth function, the average brightness on the vertical axis is the independent variable, and the horizontal axis is the dependent variable. When approximated as an inverse function and the vertical axis value was input to the inverse function subroutine, the horizontal axis value was obtained. If this subroutine is used, γ correction can be performed. The results are shown in FIGS. As a result of the γ-corrected image processing, the gray scale portion is not crushed, and it has been confirmed that an image display close to the original image can be obtained when FIGS. 19 and 20 are displayed on the LCD.

  However, γ correction is performed only when an image display close to the original image is confirmed, and image data without γ correction is used when evaluating resolution and moire contrast. In addition, when accurately obtaining the resolution and moiré contrast, it is necessary to use a raw image (RAW image) that has not undergone image processing, instead of a JPEG image from a digital camera.

(11) Calculation and display of resolution (amplitude modulation degree) FIGS. 21 and 22 show changes in luminance corresponding to the pattern of the black and white stripe portion. The stripe portion is obtained by averaging the average luminance of each FPD pixel along the stripe and displaying it in the direction perpendicular to the stripe. FIG. 21 shows a change in luminance in the horizontal direction with vertical stripes (vertical stripes), and FIG. 22 shows a change in luminance in the vertical direction with horizontal stripes (horizontal stripes). Six graphs with different amplitudes and pitches are shown in one figure. From left to right, the pattern is alternately black and white every 1, 2, 3, 4, 5, and 6 FPDs. Originally, this type of graph vibrates symmetrically in the plus (white) direction and minus (black) direction around 50%, and the amplitude of the difference with respect to the difference between the white level (100%) and the black level (0%). The ratio is called amplitude modulation AR (Amplitude Response), and the resolution is often evaluated by this.

22, the graph vibrates symmetrically in the plus (white) direction and the minus (black) direction around 50%, but in FIG. 21, the plus (white) direction vibration is crushed. I understand. There is also such white crushing, so here we use the lower amplitude of the 50% level,

It was judged that evaluating the amplitude modulation degree (AR) was excellent.

FIG. 23 shows the amplitude modulation degree (AR) evaluated by the equation (14). The horizontal axis represents black and white for each FPD pixel, the spatial frequency is 1, and the other spatial frequencies are displayed. The vertical axis is the amplitude modulation (AR). Although there are two polygonal lines, these are amplitude modulation degrees (resolution) for vertical stripes and horizontal stripes (a slightly larger value on the horizontal axis 1 is the amplitude modulation degree for vertical stripes). From FIG. 23, it can be seen that the amplitude modulation (AR) for vertical stripes and horizontal stripes is almost equal. The amplitude modulation degree (AR) given here is the scattering of light in the substrate glass of FPD (in this case, LCD). This is a result including scattering of light within the digital camera optical system, geometric aberration of the digital camera lens, wavefront aberration, parasitic aberration due to lens system tolerance, etc., and does not represent absolute amplitude modulation Please be careful. Therefore, it is recommended that the value of amplitude modulation (AR) here be used for relative comparison when lenticular lenses such as 3D display devices, light diffusion layers, and the like are different. In other words, it is only necessary to evaluate the influence of the relative ratio of AR on the additional resolution of the diffusion layer, the lenticular lens, or the like, based on the AR value in the FPD with almost no resolution degradation as a reference.

  The above is the image processing method in which the generation of new moire due to the interference between the periodic structure of the FPD pixel and the periodic structure of the image sensor is minimized, and the measurement method of the amplitude modulation degree (AR). Next, an evaluation method for a moire image appearing in an FPD will be described in detail by using an image processing method that minimizes the generation of this new moire. Regarding the flowcharts (12) and (13), the case where the moire is one direction and the moire is two directions will be described.

(12) Moire direction, pitch, contrast calculation, moire pattern extraction
Part 1 (one-way moire)
FIG. 24 shows a white chart for verification including a slight unidirectional moire. However, in this figure, the moire contrast is exaggerated by a factor of three compared to the actually used white chart for verification. This is because moire cannot be visually recognized even if the chart actually used is printed and displayed. The effectiveness of the present invention was verified by imaging a chart including such a simulated moire. In addition, a simulated moire chart in which moire is superimposed on the charts corresponding to FIGS. 3 and 4 is prepared and imaged, the black chart in FIG. 6 is used as it is, and the same procedures as in the flowcharts (1) to (11) are followed. After image processing, moire was evaluated. FIG. 25 shows the result of image processing in which the moire between the FPD pixel and the image sensor pixel devised and developed this time is reduced as much as possible to the image obtained by displaying the test chart including the simulated moire of FIG. 24 on the LCD. (No γ correction). However, FIG. 25 also shows the contrast exaggerated at the same magnification so that it can be compared with FIG.

In the processed image that is not subjected to the γ correction, the moire appears to be emphasized, and it can be seen that the visibility is higher than the direct observation. It can be clearly seen that the pitch at the lower left of the screen is coarse, the pitch is fine at the upper right, and that the moire is unequal, and the fringe angle is changing. Next, the direction, contrast, and pitch of the moire are obtained from the image processing data of the moire imaging.

In FIG. 27, a rectangular area at (u, v) coordinates of 80 × 40 (unit is FPD cell pitch) is rotated in the central portion of the image data of FIG. 25, and the standard deviation σ (θ of the square root of variance σ ² (θ) is rotated. ) Is obtained. The horizontal axis is the angle θ (degrees) and the vertical axis is the standard deviation. From this figure, it can be seen that the standard deviation has the maximum value when the angle θ is perpendicular to the moire fringe θ = 30 ° (the moire fringe vertical direction of the chart is set to θ = 30 °).

  FIG. 28 shows the dependence of the function f (θ, u) on the coordinate u every 1 ° when the angle θ is between 20 ° and 40 °. 27 and 28, it can be seen that the amplitude of the function f (θ, u) is the largest when the standard deviation has the maximum value.

  FIG. 29 shows the dependence of the function f (θ, u) on the coordinate u when the standard deviation takes the maximum value (precisely, the calculated value θ = 30.0 °). This graph suggests a method for obtaining the pitch and contrast of moire. If this graph is approximated as a sine wave, the amplitude (peak to peak: pp value) is obtained by dividing the square root of the variance (standard deviation σ) by the average value fm (θ) and multiplying by 2√2. Therefore,

It is represented by

Further, the valley of the moire waveforms u coordinate u of the point cut the average waveform down in Figure 29 ^(-) the average of the up u coordinates of a point that exceeds the average value u ⁽⁺⁾ (u ^{(- )} + U ⁽⁺⁾ ) / 2, and the moire pitch was expressed by the interval between the valleys. In addition, the portion obtained by subtracting the average value in the portion below the average value of the waveform was inverted in sign to form a mountain shape, and the moire pitch was obtained by expressing the valley of the moire waveform at the center of gravity of the mountain shape. When the mean value (u ⁽⁻⁾ + u ⁽⁺⁾ ) / 2 represents the valley of the moire waveform and the pitch was evaluated, and when the moire valley was represented by the center of gravity, almost the same result was obtained.

  In FIG. 30, the moire pitch (θ), the moire contrast (CR), and the moire pitch expressed by the center of gravity are shown by five points on the screen. The measurement result from the image picked up by giving the moire image in which the angle and pitch change depending on the location and the condition of the moire chart correspond well.

  Furthermore, a moiré pattern can be extracted by locally performing a method for obtaining the position of the moire waveform valley and the direction of the moiré and scanning the operation on the screen. The result is shown in FIG. In this case, square (u, v) regions of FPD pixels 40 × 40 are arranged vertically and horizontally by overlapping [an integer part of (40 / √2) (28) −4]] = 24 FPD pixel intervals, Rotate each square (u, v) area to find the angle that maximizes the standard deviation and the pitch in that direction, and draw the moire valley line in the square of 24 × 24 FPD pixels. The moiré pattern is expressed in minutes. However, the squares have the same size, but the squares overlap each other by reducing the pitch of the square movement amount at the left and right and top and bottom ends in the x and y directions.

  As can be seen from this example, luminance is integrated in one direction in the rectangular area to obtain a line spread function (LSF), and the angle θ that maximizes the standard deviation from the average value of LSF while rotating the rectangular area is set. The method of determining the moire direction, drawing the local moire pattern by obtaining the moire pitch and contrast from the vibrating LSF waveform, and obtaining the overall moire pattern while scanning the rectangular area is extremely versatile. Even moire patterns with unequal pitches could be extracted easily. This local moire extraction method and the overall moire extraction method based on this method are also important points of the present invention.

Part 2 (in the case of two-way moire)
An image obtained by displaying the test chart including the simulated moiré of FIG. 32 on the LCD is subjected to image processing in which the moiré between the newly developed FPD pixel and the image sensor pixel is reduced as much as possible, and is shown in FIG. FIG. 32 also exaggerates the moiré contrast three times that of the actually used white chart for verification, and FIG. 33 also exaggerates the contrast at the same magnification. FIG. 33 shows that the image including the moire shown in FIG. 32 is well reproduced. However, since the pitch of this moire is very long, the local average range is set to 5 × 5 times the normal, and shading correction is performed from the imaging information on the white chart with moire. In the processed image that is not subjected to the γ correction, the moire appears to be emphasized, and it can be seen that the visibility is higher than the direct observation.
Next, the moire direction, contrast, and pitch are obtained from the image processing data of the moire imaging. The result of reducing noise through the low-pass filter LPF due to the convolution of the Gaussian distribution with 1 / e radii _Rx and _Ry of the Gaussian distribution as 1 is further used for this image data in units of FPD pixels. This is because, in such a large pitch moire, the moire signal information is not lost even through LPF with a Gaussian distribution having a 1 / e radius about one pitch of the FPD pixel, and the calculation accuracy is improved.

  FIG. 34 shows the result of obtaining the standard deviation by rotating the rectangle at 80 × 40 (u, v) coordinates in the center portion of the image data passed through LPF in FIG. From FIG. 34, it can be understood that the standard deviation basically has two peaks. A peak at θ = 30.0 ° and a peak in a direction substantially perpendicular to the angle (θ = −60.1 °). FIG. 35 shows the dependence of the function f (θ, u) on the coordinate u in the vicinity of θ = 30 ° at which the standard deviation has the maximum value for the image data of FIG. 33 subjected to LPF processing. Comparing the case of FIG. 28 (fine pitch) and FIG. 35 (coarse pitch), it can be seen that the dependence of the amplitude of the vibration waveform on the angle θ is relaxed in the case of FIG.

  In FIGS. 36 and 37, θ = −60.1 ° (the first standard deviation has the second largest maximum value in the case where θ = 30.0 ° (the first main direction) with the maximum standard deviation and the direction almost orthogonal thereto. (2 main directions) shows the u dependence of the line broadening function f (θ, u).

  In FIGS. 38 and 39, the moire angle (θ), the moire contrast (CR), and the moire pitch expressed by the center of gravity obtained by rotating a 40 × 40 (u, v) rectangular area are shown by five points on the screen. However, the difference between the angles in FIGS. 36 and 37 (θ = 30.0 °, θ = −60.1 °) and the angle at the center of the screen in FIGS. 38 and 39 is about 0.1 ° (u, v) rectangular region. This is because the sizes of are different. The measurement result from the image captured by displaying the moire image on the FPD and the condition of the moire chart correspond well. With just a little LPF, noise could be reduced and the moiré contrast and pitch could be determined more accurately.

  In FIG. 32, the ratio of the moire contrast between the direction of θ = 30 ° and the direction of θ = −60 ° orthogonal to the central portion is set to 2: 1. Even the output of 39 is 0.0394: 0.0224, which is almost 2: 1. In FIG. 32, the moire pitch in the two directions is equal to 10 FPD pixels at the center of the screen, and almost the same result is obtained in the outputs of FIGS. 38 and 39 based on the imaging of the digital camera.

  FIG. 33 shows the result of extracting the moire pattern from the image data that has passed through the LPF. It can be seen that the moiré pattern is accurately extracted. Here, the moire pattern extracts the overall moire by separately extracting and drawing the moire in the first main direction and the moire in the second main direction.

  Note that the present invention is basic and can be used to capture an image that has just been transferred to the FPD without moiré between the FPD and the image sensor, without being used for moiré or resolution measurement. In this case, the test charts of FIGS. 3 to 6 are displayed on the FPD and sequentially imaged, and then an arbitrary image is displayed and imaged on the FPD, and an arbitrary image can be obtained by calculating the sum of luminance for each FPD pixel. An image in which the occurrence of moire is suppressed can be obtained.

  DESCRIPTION OF SYMBOLS 1 ... Digital camera, 2 ... Camera fixing tripod, 3 ... Remote shutter, 4 ... Image data transfer cable, 5 ... Personal computer, 6 ... Flat display panel.

Claims (7)

A cross hatch chart is projected on a flat display panel having a plurality of pixels, and the cross hatch chart is imaged by a digital camera having an image sensor having a plurality of pixels smaller than the pixels of the flat display panel;
Using the captured image captured by the digital camera, detecting the position of the pixel of the image sensor relative to the pixel of the flat display panel,
When other images are projected on the flat display panel, the flat display panel and the image are detected by detecting the luminance of a plurality of points in each pixel of the flat display panel and detecting the average of the luminances of the plurality of points. An image measurement method, wherein moiré due to sensor interference is suppressed and moiré in the image is detected. An image measuring device for measuring moire generated in a flat display panel having a plurality of pixels,
Projecting a cross hatch chart on the flat display panel, imaging the cross hatch chart with a digital camera having an image sensor having a plurality of pixels smaller than the pixels of the flat display panel;
Using the captured image captured by the digital camera, detecting the position of the pixel of the image sensor relative to the pixel of the flat display panel,
When other images are projected on the flat display panel, the flat display panel and the image are detected by detecting the luminance of a plurality of points in each pixel of the flat display panel and detecting the average of the luminances of the plurality of points. An image measuring apparatus that suppresses moire caused by sensor interference and detects moire in the image.   The image measurement apparatus according to claim 2, wherein the other image is a white chart.   3. The other image is a plurality of separate images of vertical stripes, horizontal stripes, or full white, and moiré and resolution generated in the image of the flat display panel are simultaneously measured. Image measuring device.   3. The image measuring apparatus according to claim 2, wherein a white chart or a black chart is imaged on the entire surface, and white level or black level correction is performed using information on the captured image. A rectangular area is defined in the flat display panel, luminance is integrated in one direction in the rectangular area to obtain a line spread function (LSF), and a standard from an average value of the line spread function while rotating the rectangular area The angle that maximizes the deviation is defined as the moire direction,
The image measuring apparatus according to claim 2, further comprising obtaining a pitch and contrast of moire from the oscillating line broadening function waveform. In the rectangular area, luminance is integrated in one direction to obtain a line spread function, and while rotating the rectangular area, an angle that maximizes a standard deviation from the average value of the line spread function is defined as a moire direction,
Furthermore, by drawing the local moire pattern corresponding to the rectangular area by obtaining the pitch and contrast of the moire from the waveform of the vibrating line broadening function,
The image measuring apparatus according to claim 2, wherein an overall moire pattern is obtained while scanning the rectangular area.

Images