JP3684479B2 - Digital camera - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention reads and recognizes information such as printed characters, figures, images, or imprints imprinted on a medium such as paper, collates, and inputs non-contact document images input as image data. The present invention relates to a digital camera for use as an input.
[0002]
[Prior art]
A digital camera that is currently commercialized has a feature that a shape such as a three-dimensional object can be photographed by a simple operation by pointing the camera toward a subject to which an image is input and pressing a shutter.
[0003]
On the other hand, products for inputting images of documents and the like include flatbed scanners and sheet scanners. Both have better performance than digital cameras in terms of inputting document images, but flatbed scanners have a larger installation area and slower reading speed, sheet scanners have a smaller installation area but can only read sheet-shaped ones. There are pros and cons, such as no digital cameras, and digital cameras are superior in that they easily input images.
[0004]
As a problem when using a digital camera as an image input device for documents, etc., information such as characters, figures, and images printed on a flat surface such as paper is important, so that necessary information can be read more accurately. Requires that the object to be read be in an ideal plane as much as possible, and that the camera be pointed at the subject accurately.
[0005]
Various inventions have been proposed in which an image read from a camera other than the front is corrected to an image viewed from the front. For example, Japanese Patent Laid-Open No. 9-289611 discloses a digital camera that captures an image by correcting an oblique image of a subject to a pseudo front image.
[0006]
However, it is difficult to keep a medium such as paper in a flat state at all times, and in fact, irregularities are often generated. In contrast to such paper input, conventional flatbed scanners, etc., press the reading paper surface against a reading surface such as glass to prevent deformation of the image due to deformation and to input an image without unevenness. Yes.
[0007]
[Problems to be solved by the invention]
By the way, in the conventional flatbed scanner, scanning of the reading head takes time, and when there are many reading objects, much time is required for reading. Also, the reading environment is limited, and there are restrictions on the reading position and size. Furthermore, since the flatbed scanner is large in the lateral direction, a large area is required for installing the apparatus.
[0008]
An object of the present invention is to provide a digital camera that uses a digital camera instead of a flatbed scanner, and in this case, even a folded document or image can be captured as planar image data.
[0009]
[Means for Solving the Problems]
In order to achieve the above object, the present invention provides a digital camera including a reading unit that digitizes and reads an image, and a storage unit that stores an image read by the reading unit. Then, a deviation between each image obtained by reading a plurality of subjects is obtained, an image part having a large deviation is extracted as a feature point, and a distance between the feature point and the reading unit based on a change in brightness around the feature point The On asking Obtaining a three-dimensional position of a plurality of the feature points; The read image is divided into a plurality of polygon images based on a plane formed by connecting the plurality of feature points, and a path distance of each side of each polygon image is obtained based on a three-dimensional position of each feature point. Then, each polygon image is developed on a two-dimensional platen plane according to the determined path distance. The image correction means is provided. In this case, the corrected image data can be output to the storage means.
[0010]
According to the above configuration, the reading unit can read by changing reading conditions such as a focus position and a tilt angle. The image correcting means can detect a feature point between images from a plurality of images read under a plurality of changed conditions, and measure a distance between each point of the document and the reading means from a blur amount of the image at that time. it can. Then, if the measured distance is used, an uneven image can be developed into a planar image. In addition, it is possible to increase the speed by making the texture mapping necessary for developing the uneven image into a planar image hardware. An image distorted by oblique reading can be integrated in the texture mapping process to reduce the number of conversions, thereby improving image quality and processing speed. Further, white noise caused by photoelectric conversion can be reduced by adding and synthesizing an image in which the coefficient of the filter is changed according to the distance and the blur is restored according to the distance.
[0013]
In configuring the digital camera, the following elements can be added.
[0016]
(1) A mode in which a read image read by the reading unit is directly input to the storage unit; and a mode in which the read image is corrected by the image correction unit and then input to the storage unit. It is configured to be.
[0018]
(2) Wireless data communication means for transferring image data with the outside is provided.
[0019]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram showing a digital camera according to the present invention and its peripheral system configuration. The digital camera 1 according to the present invention includes a reading unit 11, a mode selector switch 12, a three-dimensional correction unit 2, an image correction unit 13, an image memory 14, and an auxiliary display 15. The three-dimensional correction unit 2 is provided with a focal position control unit 22 and a dimensional image correction unit 21. A personal computer (PC) 3 having a display 4 and a printer 5 is connected to the image memory 14.
[0020]
The digital camera 1 configured as described above reads a document (including graphics) placed on a desk or the like by the reading unit 11, digitizes the read image, corrects the image, and then stores the data in the image memory 14. Transfer and save. In this case, it is possible to select whether or not to perform correction by the three-dimensional correction unit 2 on the image data according to which state the mode changeover switch 12 is operated.
[0021]
When correction is performed by the three-dimensional correction unit 2, the reading unit 11 performs reading processing a plurality of times while changing the focal point based on a signal from the focal position control unit 22 during photographing. Here, the focal position control unit 22 is set in advance with a plurality of appropriate focal positions obtained from the optical system, and the reading unit 11 moves the lens based on a reading instruction from the user to perform photographing. For example, when the user gives a shooting instruction, the reading unit 11 performs the first reading in a state where the document table is in focus, and then the second time in a state where the document is in focus at a position 3 cm above the document table. A reading process such as reading is performed. Then, the three-dimensional correction unit 21 performs correction processing based on the first reading and the second reading, and outputs the image data after the correction processing to the image correction unit 13.
[0022]
When the correction by the three-dimensional correction unit 2 is not performed, the image data read by the reading unit 11 is directly output to the image correction unit 13.
[0023]
The image correction unit 13 captures image data from the three-dimensional correction unit 2 or the reading unit 11 to perform image quality correction such as luminance correction, and stores the image data in the image memory 14. The stored image data can be displayed on the display 15 attached to the camera.
[0024]
Further, the image data stored in the image memory 14 is transferred to the personal computer 3, and the personal computer 3 executes recognition of characters and graphics, image processing, etc., and displays recognition results and target images on the display 4. The data is output to the printer 5 as necessary. It is also possible to output the image data directly from the image memory 14 to the printer 5 for display.
[0025]
Details of each part of the digital camera shown in FIG. 1 will be described below.
FIG. 2 is a diagram showing the appearance of the digital camera according to the present invention. The digital camera 1 includes a CCD sensor as an image sensor and captures an image through a lens. In the camera body, a button 18, a display 15 for displaying captured images and corrected images, a mode selection switch 12 for selecting correction contents at the time of reading, and the current capture mode are displayed. Display device 17 and a microphone 16 for indexing the captured image with sound. With such a configuration, it is possible to provide an apparatus for capturing an image with a high degree of freedom and good operability using a digital camera.
[0026]
In addition, in order to transfer the captured image data to a personal computer or printer, for example, a general-purpose interface such as a USB is attached, so that the user can retrieve the image data in the camera and print it on the printer or recognize character information in the image. It is possible to perform processing. In addition, by incorporating a data transmission device in the camera and attaching a receiver to the data receiving side such as a personal computer, it is not necessary to use a cable to transfer the captured image, and it is possible to capture images with a higher degree of freedom. It becomes.
[0027]
FIG. 3 is a functional block diagram of the 3D image correction unit 21 in the 3D correction unit 2. Using a plurality of images read by the focal position control unit 22 and the reading unit 11, a feature point distance measurement unit 212 calculates a deviation between images and extracts a feature point. Here, a portion where the image differs depending on the reading condition is called a feature point. In other words, when the focus value of the camera is changed and a piece of paper on the desk is read, the image of the edge portion of the paper is blurred and changes. These points are called feature points. The distance between the camera and the feature point is measured depending on how the feature point images differ. Next, the focus correction coefficient calculation unit 213 uses the distance to measure in advance the degree of blur according to the focus position, and create and apply a filter that suppresses the blur from the measurement result. Perform the correction. Further, an image having the closest focal length is selected from images taken at a plurality of focal points according to the distance between the lens measured for each feature point and the feature point, and the images are combined. Next, in the perspective correction path calculation unit 214, the geometric deformation state of the original flat surface due to the geometrical change, inclination, document thickness, and document bending at the time of reading is determined as a feature point and a lens. Is obtained by calculating the amount of lift of the document table from the distance measurement result, the shape data is transferred to the texture mapping unit 215, the deformation is corrected so as to return the deformed shape to a plane, and the corrected image is stored in the frame memory 216. Store. Thereafter, the image data is transferred to the image correction unit 13.
[0028]
By adopting such a configuration, there is an effect that a low-cost device can be provided because a special distance measuring sensor is not required. In addition, since the distance can be measured for each feature point, it is possible to set an optimum correction parameter for a parameter that changes according to the distance, such as a blur correction parameter, for each feature point, and the bending is complicated. This is advantageous in that it is possible to set optimum parameters for the entire image when reading a form or the surface of a three-dimensional object. In addition, since the images are synthesized using images read at a plurality of focal points, there is an effect that an image with less noise and blur such as characters and symbols described in the document can be obtained. In addition, since both geometric correction and distance correction are executed, there is an effect that an image with less distortion can be provided for characters and symbols described in the document.
[0029]
FIG. 4 shows the flow of the reading operation. When the capture button is pressed (step 1001), the state of the mode selection switch 12 is searched (step 1002), and if it is the three-dimensional capture mode, a control signal is transmitted from the focal position control unit 22 to the reading unit 11 to The focal position is adjusted to the focal position 1 (step 1003). After setting the focal position, a reading operation is performed (step 1004). Thus, an image focused on the original surface is input. When the reading operation is completed, similarly, a control signal is transmitted from the focal position control unit 22 to the reading unit 11 to adjust the focal position of the lens to the focal position 2 (step 1005). After the focal position is set, a reading operation is performed (step 1006). Here, as the setting values of the focal position 1 and the focal position 2, a value obtained by adjusting a focal position value that can read a thick original is used. Further, if necessary, an image is input by changing the reading condition of the focal position. Three-dimensional correction is performed from these images (step 1007), and data is transferred to the image correction unit (step 1009).
[0030]
If it is not in the three-dimensional capture mode in step 1002, the data is read once and transferred to the image correction unit (step 1009). The image correction unit performs brightness correction or the like when the image appears dark, and is finally stored in the image memory 14 (step 1010). By adopting such a configuration, it is possible to perform a plurality of shootings at a time just by pressing the capture button 18 of the camera, and to correct and save the image data immediately after shooting, thereby improving the operability. There is.
[0031]
FIG. 5 is a functional block diagram showing an outline of the feature point distance measuring unit 212 for three-dimensional correction. Using the plurality of image data A and B in the input memory 211, the coordinate position correction unit 2121 corrects the coordinate position, and the distributed image creation unit 2122 creates a dispersed image in order to detect a deviation between the images. Further, the feature point extraction unit 2123 extracts the feature point, the distance measurement unit 2124 measures the distance with respect to the feature point, creates the distance data D for each feature point, and stores the distance data D in the input memory 211. .
[0032]
First, the correction of the coordinate position in the coordinate position correction unit 2121 corrects the change in the enlargement / reduction of the image accompanying the change in the focus. In other words, the scale inside the image captured by changing the focal point changes depending on the focal point, so correction for adjusting the scale is performed. The correction coefficient can be easily grasped in advance by reading a plurality of images of a fixed pattern in advance by changing the focal point and measuring the distance between the reference points and the degree of blur in the resulting pattern. Next, a dispersion image creation in the dispersion image creation unit 2122 calculates a dispersion value between a plurality of images. Here, the dispersed image is used to quantify the difference in pixel data at the same feature point between a plurality of images taken with different focal points. The variance value at the feature point is obtained by calculating the average value V of the brightness at the same feature point in the image read by changing the focus, and summing the values obtained by squaring the difference between V and the brightness of the image at each focal length. Is the average value. The obtained dispersion value at each feature point is saved as dispersion image data.
[0033]
Next, when the feature point extraction unit 2123 compares the values of the dispersed image data at the feature points with a specific threshold value, the feature point can be easily detected as a feature point exceeding the preset threshold value, that is, the deviation between images. Can be extracted as feature points. For example, the edge of a lifted piece of paper can be cited as a candidate.
[0034]
With such a configuration, there is an effect that distance measurement can be realized only from a plurality of images with different reading conditions, and a special distance sensor can be dispensed with.
[0035]
FIG. 6A shows a distance measurement flow in the distance measurement unit 2124 in the feature point distance measurement unit 212. First, based on the brightness value around the feature point, the brightness change around the feature point is approximated by a straight line (step 1021). This operation is created only by changing the focus, and this is used. Then, an equal brightness straight line that produces the same brightness is created (step 1022). The distance from the camera from the position where the focal point is formed is calculated (step 1023), and three-dimensional position information is calculated from these positional relationships (step 1024).
[0036]
FIGS. 6B and 6C are diagrams conceptually showing how an image is observed depending on the focus position of the optical system and the document reading position. A and B are positions where documents are placed, and these documents are read from the left side (not shown). The dotted line in the figure indicates the optical axis, and the solid line indicates a part of the light beam, which are collected at the focal position. In FIG. 6B, when the original is placed at the reading position A and the focus of the optical system is at the position A, the boundary between white and black is read without blur as in the image 1026. Can do. When the original is at the position B at the same focal position, the boundary between white and black of the original is blurred, and the read image becomes an image with no clear boundary such as 1025. In this case, an image can be obtained by using the image 1026 as it is. FIG. 6C shows a case where the focal point is in focus at a point close to the original position B, and both the original position A and B are blurred as images 1027 and 1028. In this case, a high-quality output image can be obtained by combining the images using these two images 1027 and 1028. With such a configuration, there is an effect that distance data can be measured with high accuracy from images having different focal points.
[0037]
Next, the focus correction function 213 will be described in detail. FIG. 7 is a conceptual diagram showing how the blur f (Z) at the feature point in the captured image changes according to the height position Z where the document is lifted from the document table. . Here, the blur f (Z) is considered as a line width when the dispersion image is binarized with a preset threshold value. The unit is the number of dots of a pixel. Here, a change in the blur width when the focal position 1 is fixed at a focal position that matches the height of Z0 and the focal position 2 matches the height of Z1, and the document is lifted from the document table is shown. Since the image at the focus position 1 is in focus at Z0, the blur width is 0, but the blur width changes as the focus shifts by floating the document from Z0, as shown by a solid line graph The bokeh width increases monotonically. The image at the focal position 2 has a blur width that changes like a graph with a broken line, and has a shape similar to a quadratic curve that takes a minimum value at the focused Z1. By using this graph, if the amount of blur can be measured, the focal length can be obtained, and the distance can be estimated.
[0038]
Next, the fluoroscopic correction route calculation function 214 will be described in detail. FIG. 8A shows a conceptual diagram of the route calculation in the perspective correction route calculation unit 214. This is a figure showing a characteristic point when a certain piece of paper is read, that is, a distance measurement result of one side of the piece of paper. The horizontal axis of the graph indicates the distance from the apex of the paper piece, and the vertical axis indicates the amount of lift of the paper piece from the document table. In this example, a situation is shown in which the apexes of a piece of paper are lifted at both ends. First, as shown in (b), only the end points P, Q, R, S, and T of the straight line are extracted, and the distance between the end points as the path distance is measured. For example, the distance between the end points P and Q is observed on the read image data in the form of the distance between P and Q mapped on the X axis. By measuring the distance between the end points P and Q on the paper using, for example, the three-square theorem, the path can be grasped. Based on this distance, based on the reference point P ′ as shown in (c), position information developed for each end point, such as the end points Q ′, R ′, S ′, and T ′, according to the respective distances. Deploy them side by side on a plane while taking over. This is applicable to the texture mapping technique used in 3D graphics.
[0039]
FIG. 9 is a conceptual diagram in which this is expanded to a two-dimensional plane. When the bent form as shown in FIG. 9 (a) is observed from above, it looks like (b), but the two-dimensional route calculation described in FIG. 8 is expanded to three dimensions, and the route is calculated based on the distance information. When developed on a plane, a composite image as shown in (c) is obtained. In this way, a folded form or a character written on a curved surface can be pressed against a plane to obtain a composite image observed from the front.
[0040]
FIG. 10 summarizes these processes in a flow. As described with reference to FIG. 8, feature point distance data is input, and the feature points are linearly approximated (step 1401), and distance data from which unnecessary feature points are deleted is generated. Next, as described with reference to FIG. 9, a plane created by these feature points is approximately created (step 1402). In this case, three adjacent points are selected, and the distance between the plane composed of the three points and the peripheral feature points of the three points is calculated. If the distance is smaller than a preset threshold value, the process of approximating the plane and merging is repeated. These are approximated as triangles and described as polygons (step 1403), the coordinate relation and connection relation are checked for consistency, and texture mapping from the three-dimensional model created with distance data to the two-dimensional platen plane is performed. A development coefficient such as a coordinate position and an enlargement / reduction ratio is calculated (step 1404). With such a configuration, there is an effect that an image can be developed in a plane using distance data obtained for feature points, and a document can be corrected and read correctly even on a folded form or a thick book surface.
[0041]
Next, perspective transformation will be described. When reading from a free position, such as diagonally above, rather than above the front of the document, the near area is large and the far area is small. It is perspective transformation to change this as seen from the front. FIG. 11A shows a conceptual diagram of perspective transformation. If the image observed from the front is a rectangular PBCQ, it looks like a trapezoid ABCD when observed from diagonally above. The side AD is a near side, and the side BC is a far side. The oblique reading optical system is designed so that the reading resolution can be ensured even at a far side. Therefore, if the side BC is aligned in common, the side AD is longer than the side PQ, and the side AB is also the side PB. Longer than that. The purpose of the perspective transformation is to convert the observed trapezoid ABCD into a rectangular PBCQ. FIG. 11B shows variable definitions for conversion. The length of the side AD is Src_Width, the length of the side BC is Dest_Width, the difference between the coordinates A and B in the x direction is Dx_Left, and the difference in the y direction is Dy_Left. The hatched portion is a reading area.
[0042]
FIG. 12 shows a flow of perspective transformation. First, the image extraction start position of the reading area in the input memory 211 is calculated (step 1501). The start position xi is initially a point A, but is a function of the sub-scan line y, and can be expressed as y * Dx_Left / Dy_Left. Next, since it is necessary to align the length with Dest_Width in line units, the reduction ratio is calculated for each line (step 1502). First, the length x_width of the target line is calculated. This can be expressed as x_width = (Dest_Width + ((Src_Width−Dest_Width) * (Dest_Length−y) / Dest_Length)) by simple proportional calculation. The reduction ratio is Dest_Width / x_width. However, since the calculation using the decimal point takes time, the actual processing uses DDA (Digital Differential Analyzer) (step 1503). Finally, it is determined whether or not the processing is completed (step 1504). By adopting such a configuration, it is possible to realize a texture mapping process that takes a long processing time with simple hardware and to provide a low-cost and high-speed processing system.
[0043]
FIG. 13 shows an image composition method and hardware configuration diagram in the texture mapping unit 215. FIG. 13A shows an image g0 (Z) at the first focal length Z0, an image at the second focal length Z1, and g1 (Z) according to the distance Z from the measured lens position to the feature point of the document. ). The broken line from Z1 on the Z axis to 1 on the gn (Z) axis indicates the mixing ratio of the image g1 (Z) at the first focal length Z1 to the distance Z, and on the Z axis from Z0 to the gn (Z) axis. The straight line up to 0 indicates the mixing ratio of the image g0 (Z) at the second focal length Z0 with respect to the distance Z. If the focal length is Z0, only the image g0 (Z) at the first focal length Z0 may be used, and if the focal length is Z1, the image g1 (2 at the second focal length Z1 ( Only Z) may be used. If the focal length is between Z0 and Z1, the image g0 (Z) at the first focal length Z0 and the image g1 (Z) at the second focal length Z1 may be used in half. Generally speaking, when the measurement distance is Zp, the image at the first focal length is mixed with g0 (Zp), and the image at the second focal length is mixed with g1 (Zp). Is generated.
[0044]
FIG. 13B is a hardware configuration diagram of the image composition unit. The image 1301 at the first focal length in the input memory 211 is read, the ratio is adjusted by the distributor 1307 through the blur correction filter 1304, and the image 1302 at the second focal length in the input memory 11 is read and blurred. After the ratio is adjusted by the distributor 1308 through the correction filter 1305, the ratio is synthesized by the adder 1310, the timing is adjusted by the latch 1311, and the output G_Out 1312 is output. Here, the blur correction filter 1305 is a filter that repairs the blur generated in the image in accordance with the distance, and in general, an edge enhancement filter can be substituted. The coefficient of this filter needs to change according to the distance, and is realized by changing the coefficient of the center value and the peripheral value according to the distance. Not only this blur correction filter coefficient 1306 but also the ratio setting 1309 of the distributor is calculated based on the tables of FIGS. 7 and 13A and set via the bus 1303. With such a configuration, there is an effect that white noise can be reduced by synthesizing an image obtained by correcting a blur of a plurality of images according to the distance, and a high-quality image can be synthesized.
[0045]
FIG. 14 shows a hardware configuration for executing the texture mapping 215. The output G_Out of the image composition unit generates data while averaging with multi-value data so that the data is not lost when the perspective transformation is reduced, and obtains output data MW_DAT. In this case, the ratio of input data is adjusted by a distributor 1331 according to the reduction ratio, and the adder 1332 synthesizes the input data. As with the input data, the output is adjusted by the distributor 1333, temporarily stored 1334, and accumulated by the adder 1332. When the variable parameters shown in FIG. 11B are set in the registers 1341, 1343, 1345, 1351, and 1349 of FIG. 14, hardware perspective conversion is executed. First, the length x_width 1343 of the line of interest is generated by subtracting the displacement DX_Width 1341 by the subtracter 1342 for each line. On the other hand, the output length Dest_Width 1345 is added by the adder 1346 using the temporary storage register X_DDA 1348 for DDA every clock. However, if X_Width 1343 is subtracted from the output of X_DDA 1348 by subtracter 1344 and X_DDA 1348 can be subtracted, the subtraction result is set to X_DDA 1348 via selector 1347. The borrow of the subtracter 1344 is used as the control signal X_Skip. The read start address X_Start 1351 is added for each line by the displacement DX_Start 1349 by the adder 1350 and loaded to the x-direction counter X_Count 1352. This counter is used as a read address. The address for writing to the memory is managed using another memory counter M_Count 1355. By controlling this count-up control using X_Skip, the output length can be made uniform. With such a configuration, there is an effect that the texture mapping process with a perspective transformation function can be realized only by setting several registers from the CPU.
[0046]
FIG. 15 is a configuration diagram of a digital camera to which an image index function is added. The digital camera 1 is provided with a voice input unit (microphone) 16 for adding a memo or the like to a captured image and storing the image data.
[0047]
A user inputs a description of a captured image, a record of a memo, a title, and the like through a microphone 16 during or after the image is captured. After the image read from the reading unit 11 is corrected by the three-dimensional correction unit 2 or directly transferred to the image correction unit 13, the input record is added to the image data and stored in the image memory 14. As a data format for addition, it is possible to record by adding audio data to the end or the beginning of the image memory, and to separate and reproduce the audio and image data when reading. With this function, it is possible to easily perform data management after taking an image, so that it is possible to perform an efficient operation of acquiring image data.
[0048]
FIG. 16 is a configuration diagram of a digital camera capable of capturing an image such as a huge plane. When the division reading mode is selected by the mode switch 12, the image captured by the reading unit 11 is stored in the image memory 20 after performing the above-described three-dimensional image correction. In this case, when the user designates the number of divisions in advance, the plane is divided into quadrilaterals according to certain rules and indexed, and the result is displayed on the display 15. The user shoots after declaring, using the microphone 16 or the like, which area to shoot when shooting.
[0049]
When the number of divided images specified at the beginning is stored in the image memory 20, the image data is transferred to the image pasting unit 21. The image pasting unit 21 joins adjacent images and outputs the images to the image correcting unit 13 as one image. The image stitching method has already been realized by general software, and can be realized by using these methods.
[0050]
With this configuration, the digital camera of the present invention can be realized simply by incorporating a memory and image pasting software.
[0051]
FIG. 17 shows a flow of the division reading function. First, when the user inputs the number of divisions (step 1601), an index is automatically created in accordance with a certain rule inside the camera, and the division result is displayed on the display (step 1602). Next, the user inputs the displayed index number (step 1603) and takes an image at the corresponding position (step 1604). This operation is repeated for the number of divided sheets (step 1605). When the divided number of images are prepared, the image data are combined by the image joining unit (step 1606), and the data is stored in the image memory of the main body.
[0052]
With this function, when shooting figures and characters drawn on huge planes such as large murals, the restrictions on the shooting location can be greatly reduced, so that it is possible to perform highly flexible reading operations. .
[0053]
FIG. 18 is a configuration diagram of a digital camera having a function of automatically setting a reading range using feature point extraction. When automatic setting processing is selected with the mode selection switch 12, the feature point extraction unit 30 extracts feature points in the distance measurement for each feature point from the read image, and stores the feature point information. Further, when extracting feature points, the same result can be obtained even if edge detection processing generally used in image processing is used. From this feature point information, a rectangular portion in the image is extracted as a reading range, and only the image in this range is subjected to three-dimensional image correction and output as a reading result. As a result, when it is desired to read only the inside of a page such as a book, the user can save the trouble of photographing the reading range in accordance with the page at the time of photographing, so that an image can be captured with good operability.
[0054]
An application example when sequentially reading a document such as a book that covers a plurality of pages will be described. A continuous reading mode and a page storage unit are added to the camera by the mode switch 12. In this mode, when consecutive pages are read sequentially, information that the page number is written in the upper or lower center of the image is stored in advance, and the number at that position is read using OCR at the time of reading. The change of the number is monitored, and a warning is given when there is skipping or the like.
[0055]
This operation flow is shown in FIG. When the reading button is pressed (step 1701), a reading operation is performed (step 1702), and a page number recognition process in the image is performed (step 1703). It is determined whether or not the page is the next page of the number read immediately before (step 1704). If it is not the next page, a warning is given to the user using a buzzer or an auxiliary display (step 1706). Then, the user is asked to select whether or not to read again (step 1705). As a result, the user is less likely to skip skipping when sequentially reading successive images, and there is an effect that the user can capture with good operability.
[0056]
In the above processing, the distance between feature points has been measured and corrected based on image data captured under a plurality of conditions. However, the shape of the plane to be captured in advance is determined to be a document of A4 vertical size or B5 book size. If a memory for storing the shape of the frame of the feature point is provided and the information is used for capturing, the image can be processed only by image processing without using the distance measurement means of the feature point. Is possible. The method can be achieved by combining the feature points of the image and the size of the specified shape, and combining processing such as image enlargement / reduction based on the ratio of the sizes.
[0057]
In the case of this method, when a regular document is always read, it is not necessary to prepare a distance measuring means, the processing load is reduced, the reading operation is speeded up, and the user can quickly capture an image. effective.
[0058]
The digital camera shown in FIGS. 1, 15, 16, and 18 may be provided with wireless data communication means for transferring image data with the outside.
[0059]
【The invention's effect】
As described above, according to the present invention, it is possible to input an image of a document on a plane such as a folded form or a thick book with a simple operation of the digital camera without fixing the subject and special distance detection. Since an image can be input with high image quality without using a sensor, the operability at the time of taking information drawn on paper or a plane as digital can be greatly improved.
[0060]
Further, it is not necessary to scan a reading head such as a flatbed scanner, and the capture time can be increased. Furthermore, there are few restrictions on imaging conditions, and images can be captured from any position. When not in use, it does not require a large installation area like a flatbed scanner, and does not hinder the work of the user.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of a digital camera according to the present invention.
FIG. 2 is a perspective view showing an appearance of a digital camera according to the present invention.
FIG. 3 is a functional block diagram of a three-dimensional image correction unit.
FIG. 4 is a diagram illustrating a flow of a reading operation.
FIG. 5 is a diagram showing an outline of a distance measuring unit for each feature point for three-dimensional correction.
6A and 6B show the principle of three-dimensional correction, in which FIG. 6A shows a distance measurement flow, and FIGS. 6B and 6C show how an image is observed depending on the focal position of an optical system and an original reading position. It is the figure which showed these conceptually.
FIG. 7 is a diagram showing the relationship between the height position of a document and line blur.
FIG. 8 is a diagram conceptually showing that path correction is performed in the order of (a) → (b) → (c).
FIG. 9 is a diagram specifically illustrating the path correction in FIG. 8;
FIG. 10 is a diagram showing a processing flow of three-dimensional correction.
11A is a conceptual diagram of perspective transformation, and FIG. 11B is a diagram showing variable definitions for transformation.
FIG. 12 is a diagram showing a flow of perspective transformation.
13A is a diagram showing an image composition method, and FIG. 13B is a hardware configuration diagram.
FIG. 14 is a hardware configuration diagram of a texture mapping unit.
FIG. 15 is a configuration diagram of a digital camera to which an image index function is added.
FIG. 16 is a configuration diagram of a digital camera to which a division reading function is added.
FIG. 17 is a diagram illustrating a processing flow in a divided reading mode.
FIG. 18 is a configuration diagram of a digital camera to which a function for automatically setting a reading range using feature point extraction is added.
FIG. 19 is a diagram showing a processing flow of continuous reading by page number recognition.
[Explanation of symbols]
1 Digital camera
2 3D correction unit
3 Personal computer
4 Display device
5 Printer
11 Reading unit
12 Mode selector switch
13 Image correction unit
14 Image memory
15 display
21 3D image correction unit
22 Focus position controller

Claims (3)

In a digital camera comprising reading means for digitizing and reading an image, and storage means for storing an image read by the reading means,
The reading means changes a focal position and obtains a deviation between images obtained by reading a plurality of subjects, extracts an image part having a large deviation as a feature point, and based on a change in brightness around the feature point, the feature After obtaining the distance between the point and the reading means, the three-dimensional positions of the plurality of feature points are obtained, and the read image is divided into a plurality of polygon images based on a plane formed by connecting the plurality of feature points. Image correction means for determining the distance of each side of each polygon image based on the three-dimensional position of each feature point and developing each polygon image on a two-dimensional platen plane according to the determined distance. A digital camera characterized by that. The digital camera according to claim 1, wherein
A mode in which a read image read by the reading unit is directly input to the storage unit; and a mode in which the read image is corrected by the image correction unit and then input to the storage unit. A digital camera characterized by being. The digital camera according to claim 1, wherein
A digital camera comprising wireless data communication means for transferring image data to the outside.

Images