TW200841702A - Adaptive image acquisition system and method - Google Patents

Abstract

A system and method for correcting optical distortions on an image acquisition system by scanning and mapping the image acquisition system and adjusting the content of output pixels. The optical distortion correction can be performed either at the camera end or at the display receiving end.

Description

200841702 IX. Description of the Invention: [Technical Field] The present invention relates to an image capturing system, particularly but not limited to, providing an image for adapting an output image to a high quality resolution camera (still camera) Or a video camera system and method therefor. [Prior Art] Due to the rapid development of high resolution sensors using Charged Couple Device (CCD) or Complementary Metal Oxide Semiconductor (CMOS) technology, Digital still cameras and video recorders have become popular and become affordable products for the general public. The technology of such sensors has been rapidly developing with the increasing density and cost reduction of semiconductors. However, the cost of digital cameras and video recorders has not been greatly reduced. The main reason is the image capture. The optical system used in the system encountered bottle 1/neck both in terms of execution and cost. A typical variable focus and variable zoom optical system has a number of lenses that increase when the image pixels are 656 horizontal lines from the horizontal resolution of Closed Circuit Television (CCTV). Up to 2,500 horizontal lines of the 10 megapixel camera and above, and the pixel resolution varies from 8 bits to 10 bits to 12 bits, equipped with optical components and optical systems Accuracy must be improved and its optical distortion reduced. The development of optical technology is not as fast as semiconductor technology. Precision optical parts with strict tolerances of 200841702, especially those of aspheric lenses, are very expensive. The current optical surface requirements are 1 micron (micro) Meters ) or higher, when multiple optical components are assembled into an optical system, the tolerance increases. Even with great care during the assembly process, it is difficult to maintain its focal length (f〇cus), spherical aberration (sphered aberration), centering (), chromatic aberration (such as ab 化 ab ab ab ab ab ab ab ab ab ab ab ab ab ab ab ab ab ), astigmatism, distortion (dist〇rti〇n), and color 〇 聚 ( (color C〇nvergence), even if the cost of the sensor is falling, and the image captures the cost of the optical system in the product However, it continues to rise. Obviously, the traditional pure optical method (appr〇ach) cannot solve this problem. It is necessary to have a lens that is not a wide-angle lens. When someone wants to take a self-portrait with a mobile phone, 'W stretch his/her arm to the farthest place. High resolution CCD or C-Tens sensors are available and cost effective. - High resolution Sensing: In combination with a very wide-angle lens system, the coverage of the SURVeillance target can be the same as for many general low resolution cameras. Therefore, replacing many low-resolution cameras with several high-resolution cameras is more cost-effective, regardless of installation, operation, and %嘈. For the traditional pure optical bed, it is very difficult to design and manufacture wide-angle lenses. Tianji everyone knows the field of view (field of

When Vlew is enlarged, the geometry of the lens will increase. In general terms, the increase in distortion is the seventh power of the slave's angle of view. This is why most digital cameras do not have a wide-angle lens. The wide-angle lens of e-Shishi is not very expensive and has a large distortion. Fish, μ a丄+mail a,* fish-eye lens A type of wide-angle lens that everyone knows. 200841702 As is well known in the art, the general formula of the optical system geometric distortion approximation can be used as a basis for correction, whether through warp table generation or immediate fixed algorithm (fixed). Algorithms, the distortion of the lens can be corrected to some extent, while the general formula cannot achieve a certain quality due to the tolerances of the lens manufacturing. And the general formula can't capture the unique characteristics of each image capturing system optical distortion. The general formula (generai formula), for example, parametric class 〇f warping function, polynomial Polynomial functions, or scaling functions (scaHng functi〇ns), are very sophisticated calculations that require expensive hardware for immediate correction. Therefore, there is a need for an efficient and cost-effective new system and method for correcting optical distortion in an image capture system. SUMMARY OF THE INVENTION The primary object of the present invention is to provide an image capture system having an adaptive device for correcting optical distortion, the corrected optical distortion including immediate correction of its geometry, brightness, and contrast variations thereof.另 Another object of the present invention is to provide an image capture system with an adaptive method to instantly correct optical distortion. =I Ming has a purpose to provide a method of video content authentication at_), which is based on the correction of the video geometry, brightness and contrast during the adaptation process. Embodiments of the present invention provide a system and method for modifying video content at an affordable 7 200841702 jberg to correct optical distortion on the fly. This embodiment does not require a frame buffer, nor does it have a problem of frame delay between image frames. This embodiment is performed at a pixel clock rate, for which reason it can be described as operating in a pipeline, with one pixel output per pixel input. Embodiments of the present invention are equally well processed for upsampling or d〇wn-sampling, which does not require a uniform spatial distribution of the output pixels (uniform spatial). Distributi〇n), and, in the embodiment, only one important mathematical operation method is used: division, and it is not necessary to use the complicated and expensive floating point calculation method which is used in the conventional image adjustment system. ). In an embodiment of the invention, the method includes: placing a test target in front of the camera, extracting an output pixel centroid for a plurality of output pixels; determining a first wheeled enemy from the output pixel. The phase of the pixel is out of the pixel; the overlap of the virtual pixel is determined according to the pixel heart and the adjacent output pixel, and the virtual pixel is determined according to the content of the overlapped virtual pixel corresponding to an input t The first output image:::: then outputs the determined content to a display device. ', Valley, in an embodiment of the invention, the engine, an adjacent output pixel engine, the complex box outside ^ An output pixel centroid is coupled; - an output pixel overlap engine, a pixel centroid engine pass (four) two and - output pixel content, which is coupled with the re-engagement. The adjacent output pixel is suitable for the first-output pixel The adjacent output is determined by symplectic = several rounds of pixels - the round-out pixel overlap engine is based on the output pixel centroid extracted by 200841702 and the adjacent output image... the pixel covers the overlap of the virtual pixels, and the virtual image Haidi One round of outgoing communication; the output pixel content material is determined according to the content of the first output pixel according to the meridian, and the video display device within the j-thinking. In another embodiment of the present invention, the method:::r is a plurality of output pixels - Ο =,::〇stream), and then the output pixel centroid data, brightness, and contrast sentences are read and transmitted to a video. Displaying the pixel correction program at the end of the display device. In the variation of the present invention, for the video display device of the (four) adaptive method, the pixel centroid data and the redundancy sentence data of the camera are compatible with the pixel of the display output device. The combination of centroid data and bright and sentence-like materials requires only the use of a set of hardware to perform the operation. [Embodiment] The following description provides the general practitioners with the knowledge and practice. Months and apply to a particular application and its needs. For those skilled in the art, the different modifications made to the embodiment are simple and the principles defined here can be applied to other implementations. And the application, without departing from the spirit and scope of the invention, therefore, the invention is not limited to the embodiments described herein, and thus, the shapes, structures, features, and spirits described in the claims of the present invention vary equally. And the functions and functions produced by the inventions are not included in the scope of the invention and the scope of the invention, and are included in the scope of the invention. 9 200841702 The effects achieved can be further understood and understood. The detailed description of the drawings and preferred embodiments is as follows.

Please refer to Figure 1, which is a block diagram of a conventional camera. Referring to Figure 2A, a block diagram of an adaptive image capture system 100 is illustrated in accordance with an embodiment of the present invention. Each image capture system has a sensor 130 for capturing images. Typical sensors are like CCD or CMOS 2-dimensional sensor arrays that are common in digital cameras. Line scan cameras and image scanners use one-dimensional sensor arrays with optical lenses, which also suffer from optical distortion; others Image sensors such as infrared, ultra-violet, or X-ray may not be visible to the naked eye, but they also have their own optical lens system. There are cases of optical distortion, and therefore, embodiments of the present invention will also be useful for these image sensors. An optical lens system 170 is disposed in front of the sensor for collecting light rays emitted or reflected from the image and accurately focusing it to a sensor array 130. There is usually a camera control circuit 150 to change the shutter speed or iris opening to optimize the image capture. The output of the image sensor typically requires white balance correction, gamma correction, color processing, and various other operations to form a clear image of the captured image. The image processing 140 is usually completed by a special application integrated circuit (ASIC), but can also be executed by a microprocessor or a microcontroller with image processing capability. According to an embodiment of the present invention, an adaptive image processor 110 is used to perform optical distortion correction, brightness, and contrast correction on the image before the image is sent. Since the image processing invention is fast enough to perform continuous image processing or video processing, in the application of the present invention, image processing and video processing can be used interchangeably, and image transmission ( The output of the output or video is also interchangeable. A memory block 120 is coupled to the adaptive image processor 110 for storing the adaptive parameters as geometric and brightness corrections. In order to reduce the amount of memory stored, these parameters can be compressed first, and then decompressed by the adaptive image processor before being applied. Before being sent to the outside world, the processed image is processed. An output formatter 160 is encapsulated into a different output format. For the NTSC format, it is a typical analog transmission standard for Q, and the processed image is first encoded into an appropriate analog format; In the case of Ethernet, processed images are formatted into Ethernet packets. Previously, it is compressed by MPEG-2, MPEG-4, JPEG-2000 or other different commercial compression algorithms, and then the Ethernet packets are further encapsulated to match them. Transmission protocols, such as wireless 802.11a, 802.11b, 200841702 802.llg, or 100M wired Ethernet; the processed image can also be packetized to be USB, Bluetooth, IEEE 1394 , Irda, HomePNA, HDMI, or other commercial video transmission protocol standards are transmitted. Video output from the video output from the image capture system is input to the typical display device 190; the image is presented before the screen, in the display The device 19 0 is further formatted in conjunction with a specific display output device such as a CRT, LCD or projector. [Camera calibration] p A typical image capture system displays barrel distortions, such as 1st. According to the image of the checker board pattern, it can be computed throughout the image space ( Entire image space ) The centroid of the black and white squares of the checkerboard intersection, and the brightness of each grid is measured; the resulting geometric and brightness/contrast distortion image is essentially a “fingerprint” of a particular image capture system (finger print) The fingerprint contains distortions from the lens, tolerances in assembly, differences in coating on the substrate, differences in the passivation on the sensor, and errors caused by other manufacturing/assembly. . Since the light wavelength of an light affects the degree of distortion through an optical system, the distortion centroid can be collected three times, once in red, once in green, and once in blue, enabling correct adjustment of its lateral color. distortion. A checker board pattern test target with a 25-inch inches width, as shown in Figure 4A, can be fabricated using photolithography 12 200841702 (photolithography) to achieve high precision in commercial The upper system achieves a micro-inches accuracy of 25 inches in total width, which produces dimensional accuracy of 0.00004%. For a linear dimension of 2500 pixels for a 10 megapixel camera, the accuracy of the checkerboard can be expressed as 0.1% of a pixel. As shown in Fig. 4C, the checker board test pattern does not need to be placed completely perpendicular to the camera, and the offset angles can be directly accurate from both sides of the a/b. The degrees are calculated and the camera offset angles are removed from the calibration error; no precise mechanical calibration is required in this calibration procedure and there is no need to move the target (calibration plate). By using a typical cross shaped or isolated squares fudicial patterns, camera correction accuracy can be achieved by about 1/4 to 0.1 pixels. In the checker board pattern, black squares can be used (J-squares and white squares intersect each other for higher accuracy. See Figure 9, which is calibrated through a camera and A graphical method is used to determine the intersection of two diagonally placed black squares 905 and 906 to display a severely defocused image captured by the checkerboard pattern. Sensor array 900 superimposed On the collected image, line 901 is on the right edge of the square 905, using the calculation of the inflection point from white to black transition. , or 13 200841702 疋 use linear extrapolation to calculate the intermediate point from white to zero transition, and then determine the edge. Line 9〇2 is on the left edge of the square 906 In an optical system capable of focusing clearly, the line 901 and the line 902 should be coincident, and the key feature of the checkerboard pattern ("board pattern") is that even if there is a light with a flaw The system, the aperture optimization or focus optimization, or the alignment of the optical axis perpendicular to the calibration plate (the calibrati〇nplate) is flawed, the vertical transition (vertical transition) The line) can be accurately calculated, when it is equidistant and parallel to the line 901 and line 902. Using the same principle (token), line 903 is the lower edge of the square 905, while line 904 is The upper edge of the square 906; when the square of the square is formed by lines 901, 902, 903 and 904, the edges of the two black squares 9〇5 and 9〇6 can be Therefore, the camera calibration accuracy can reach 0. 025 pixels or better, which is characterized by the optical distortion of the entire image 0 capture system, which is a progressive varying function (smooth varying) Functi〇n), therefore, having a 4 (M_square pattern of ch_r pattern) in a linear area (3) (four) dimension is sufficient as the (8) pixel in one area ((1) claw (3) in cents) 10 megapixel camera The distortion feature, and the test pattern similar in shape to a checkerboard pattern, will have similar results, for example, a diamond-shaped checkerboard pattern can also be used. 14 200841702 The checkerboard pattern test Target ) can be fabricated on a Mylar film with a black and transparent grid using the same process as the printed circuit board. This test target can be set to one

前方 The front of the illumination source is as shown in Figure 4B. 'For the calibration of brightness and contrast, the colorimetric method of each black square and white square in the checkerboard test pattern (c〇1) 〇rmetry ) can be measured using sophisticated instruments such as the cs_1〇〇A colorimeter manufactured by Sakamoto Konica Minolta Corporati〇n (c〇i〇rime) For example, typical commercial instruments measure brightness tolerances as low as 〇·2〇/. The typical image that can be manipulated can display the gradient of brightness. As shown in Figure 5, when the brightness readings (luminance readings) obtained from the instrument are compared, the brightness and contrast distortion map (4) of the sensor can be recorded. Down, this is a "fingerprint, Uingerprint" or feature outside the geometry of a particular image manipulation system. [In the Streaming Streaming Feature] A preferred embodiment of the present invention embeds features in a video stream. Signature information' and performing adaptive image correction at the end of the display, see section 2, which illustrates a block diagram of the preferred embodiment. In this embodiment, in the image capture system The adaptive image processor (1) embeds a feature in the video stream, and the adaptive image processor 181 performs an optical distortion correction on the __ display device 191. Please refer to the section, which displays 4 pixel video data. A schematic diagram of an image having red, green, and blue within 15 200841702 and each having 8 bits. A preferred embodiment of the embossed feature for embedding optical defects in geometry and/or brightness is As shown in Fig. 7, both features can be drawn by the distortion difference between the two, which can reduce the storage requirement. Since the embedded optical distortion feature is the brightness data at the end 2, the target is not installed. Even if optical distortion correction cannot be performed, the data will be translated into very weak video data, and the embedded features are not visible on the display device, but for display devices that can perform optical distortion correction, Converting the video (] back is as if there is virtually no distortion in both geometry and brightness. This is also important for security applications, because if all video images are not distorted, The identification of the execution target can be more precise and quicker; while the %(4) data is not corrected and it will be difficult to tamper with, because both geometry and brightness will change before the display, but for Any correction of the pre-corrected data will result in failure to conform to the characteristics of the original image capture device and will be highlighted. In a Sti11 camera, the entire optical signature must be embedded in each picture before, or transmitted in advance as a feature of that particular camera; for continuous video, the whole is Optical features do not have to be transmitted all at once, there are many ways to decompose the feature and transmit it through multiple video frames, and there are many ways to encode the optical features to make them more difficult to restore ( [Representation of video compression] [Video compression before transmission] 16 200841702 The standard compression algorithm can be used before the 友 友 隹 , , , , , , , , , , , , , , , , 劣 劣 劣 劣 劣 劣 劣 劣 劣 劣 劣 劣 劣 劣 劣 劣 劣罟 Be especially careful to ensure that the optical features are not distorted during compression. [Optical Distortion Correction] In terms of geometry and brightness, the use of 弁 胜 胜 , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,进一步 〇 Further description of the image processor 110 is as follows. Specifically, the image processor m maps an original input video frame to an output video frame by matching the output pixels on the screen with the virtual pixels corresponding to the original input video frame pixels. The image processor 110 utilizes memory 12 to store pixel centroid data and/or any operations that need to be temporarily stored. The image processor 110 can be implemented by a software or a circuit system. For example, the details of the application-specific integrated circuit (ASIC) device 110 will be described in detail later. The memory 12 can include a flash memory or other memory format. In an embodiment of the invention, the system HH includes a plurality of image processors, each of which is used for - color (red, edge, blue) and/or other inner valleys (eg, : Brightness) 'Adjust the image to output in sync. Please refer to Fig. 8, which is a block diagram showing the image processor (10) described in Fig. 2A. The image processor 110 includes a round of pixel centroid engine illusion, an adjacent output pixel engine 220, an output pixel overlap 5 丨 engine 23 〇 and 17 200841702 an output pixel Μ 240 240 ° the output pixel centroid engine 21〇read (read 〇m) the centroid position to the FIFO buffer module (heart “η(10)—, impotence) memory (eg, inside the image processor or other location) corresponds to (Levant Iines) o#^„>€(lines);&〇^ Additional moments~ are stored at the same time, so the required memory can be additionally reduced. The adjacent output pixel engine 22 determines which output pixel is diagonally adjacent to the desired output pixel by diagonally adjacent output pixel positions in the s memory. Further details of the output pixel overlap engine 23 will be described in detail later. The output pixel weight (four) engine 23 () determines which virtual pixel is overlapped by the output pixel. The details of the step-by-step output of the output pixel content engine 240 will be described in detail later. The output pixel content engine (10) determines the content of the output pixels (e.g., color, brightness, etc.) based on the content of the overlapped virtual pixels. Ο i'Fig. 10 is a description of the corrected video display area 310 of the camera before the corrected display range 7 (10) and (4). Before the geometric correction, the camera output using the two wide-angle lens usually has barrel distortion, which occupies less display area than the surface 妒τ / 、, , & The corrected viewing zone 73G (_ton (10)a) (here, the virtual pixel coordinate virtual pixel coffee) contains one of the input video frames (i叩 plus *(3) frame). By multiplying 7 arrays (for example, each line has X virtual pixels and each frame has y lines), the corrected viewing area 〇 virtual pixel system exactly coincides with the round-out video frame, in the present invention In one embodiment of 200841702, the viewing zone may have a 1 6 · g orientation ratio of 1 2 8 〇 x 7 20 pixels, or a 4 : 3 aspect ratio for 640 x 480 pixels. In the optical distortion display area 310 of the screen, the actual number of output pixels is consistent with the output resolution, and in the viewing area 73, the number of virtual pixels / input resolution is consistent; in other words, The resolution of the input video frame is matched, that is, the pixel ratio of the virtual pixel to the input video frame is equivalent to 1 · 1, however, the virtual pixel to output pixel ratio is not necessarily equivalent to 1:1. For example, at the corner of the viewing zone 730, there may be several virtual pixels per output pixel, and in the center of the viewing zone 730, the virtual pixel to output pixel ratio may be equivalent to i ·· i (or less). In addition, the spatial location and size of the output pixel are different from the virtual pixel system in a non-linear manner U non_linear fashi〇n). In an embodiment of the invention, the actual output pixel is mapped to the virtual pixel such that the virtual pixel appears to be the same as the input video. Then, the mapping mode is used to resample the input video, so that the output pixel is displayed such that the virtual Q pixel looks exactly the same as the input video pixel; in other words, the output video frame is matched with the input video frame. In order to be able to view the same image. Please refer to FIG. 1 1 for a schematic diagram of mapping a virtual pixel grid 730 (virtual pixel grid) from an output pixel to an image area 31〇. In an embodiment of the invention, the output pixel content can be used to create the viewed virtual pixel, the mapping of the output pixel being represented in virtual pixel mode (or unit), for this purpose, the virtual pixel Array 73A can be viewed as a conceptual conceptual grid in which the position of any output pixel in the coordinate box 73〇 can be represented by horizontal and vertical coordinates. 19 200841702 It is like 辛Γ/ is by finding out the difference in the description of the mapping and the relative size difference in the virtual pixel coordinate box, and the (4). Big V: to any exact amount. For example: - the first output is four times larger than the first output pixel 420, and the mapping description of the first input + 410 can be χ + 2·5 / output image corresponding to the sketch, and likewise, the mapping of the rounded pixels The 遴 is x + 1 2 · 5, y.2 · 5. = The output pixel centroid engine 21 is required for other engines to communicate with each other, and the data can be stored in the memory directory m in the search directory (in _up-table f〇rm) format or other formats (for example :Lianmei = for other treatments. All the other shells needed to adjust the image;·', inferred or obtained from the contents of 3H video, this will be described in detail later. The amount of material required for the output pixels in the virtual coordinate square is obviously quite large. For example, if the virtual resolution is i 2 8 〇x7 2 Ο 〇', approximately 2 4 bits are needed to fully track each The output pixel centroid. However, the mechanism itself can easily provide significant compression _p:Ctlon) (for example, a method can find the entire pixel in each output line) and then use incremental Change to find other pixels). In an embodiment of the present invention, the operation of determining the centroid of the pixel by the imaging device 'can provide a separate guide for each pixel color'. This allows horizontal adjustment of the image while adjusting Color correction. Read more in Figure 12, which shows that the operation of the input centroid diagram from the calibration procedure is performed immediately - each centroid is operated from external storage II in a pre-computed format, m is recorded M12Q. 20 200841702 Conceptually, when the centroid engine 21 captures the centroid, the engine 210 stores the centroid in a set of line buffers, these line buffer registers. Also represents a continuous FIF〇 (which has a special interpolation in the case of the boundary), each incoming centroid is input to the beginning of the first FIF〇, and the stream is repeated from the end of each FIFO to the beginning of the next one. . The use of the centroid FIF 〇 S2 as a line buffer register enables the adjacent output pixel engine 220 to simplify the position of the adjacent centroids to determine the position of the corner, since the line is operated before and Subsequent line buffers eliminate the extra "corner" and "cornerholder" and the angular center of the total is found in the same FIFO position relative to the operated centroid. See Figure 1 3 ' It describes an output pixel angular position calculation method (c〇rner

Schematic of CalCUlati°n) The embodiment of the image line and method of adjustment depends on some hypothetical conditions: μ • Between adjacent pixels The change in size and shape of the output pixels is not significant. 〇 Φ is between adjacent pixels, and the offset of the output pixel in the direction of "X" or "y" is not large. * The size of the output pixels and the coverage of the content can be approximated by the quadrilateral enough to simulate (approximated) and the output quadrilateral. These hypothetical conditions are usually accurate in a back projection television. If the above hypothetical condition can be achieved, then when each output pixel content is prepared for daylight, any output pixel approximates the quadrilateral (quadHiaterai app) corner location (the virtual pixel coordinate 21 200841702 can be adjacent output) The pixel engine 220 is calculated on the fly by placing the halfway point 610 at the center of all diagonal output pixels (e.g., the output pixel 62A). After establishment, the overlap of the swallowed pixels is established by the output pixel overlap engine 23〇, thereby creating a direct (same) overlap of the input video. Note that in the above example, the approximation of the output pixel quadrilateral covers the ^ A virtual pixel, but it may also be so small that it can be completely placed in a virtual "", for example, the output pixel 420 (see Figure 11) can also be completely placed in a virtual pixel. 'In order to be able to perform pipeline operation, by the adjacent output pixel engine 220, the approximate angular value of each upcoming output pixel can also be advanced. One or more pixel clocks are calculated from 0—*. After the spatial relationship between the output pixels and the virtual pixels is established, the content of the content can be determined by the output pixel content engine 24 (resampled sampling) The technique is calculated. Since the output pixel size/density varies from position to location in the viewing area 310, some areas will perform augmentation a, 9 suitable for sampling (uP-samPled), while other areas are subjected to power sampling ( D〇wn_sampled). This is

At , y P” will require additional filtering power (eg, smoothing, etc.), and the filtration depends on the degree of optical distortion. The generation of the optical distortion also provides a unique mechanism for improving the resampling. For example, in a partial region of the screen 73 0, the 5 HM output pixels in A are rarer than the avatar 22 200841702, while in other Location, this relationship will be reversed, so the same resampling algorithm is feasible.刼Using: the material is made to fit in a new pixel in each virtual pixel. The new angle is known.) The change in the method can include the use of "virtual, pixel partial coverage area." The specific gravity will be further described in detail later. Please refer to Fig. 14 for a schematic diagram illustrating the pixel branch (s'divisi〇n) and the approximate value UpPn>XimatiGn). As mentioned before, - The default method is to estimate the output of the pixel on the area covered by the virtual pixel, and calculate the content value of the input according to the specific gravity value of each overlapping virtual pixel. However, with hardware Accurately calculating the percentage of overlap requires significant speed and processing power. This is not compatible with the requirements for projecting TV sets. In order to simplify the operation of the hardware, the output pixel overlap engine 230 is transmitted through the L virtual pixel. The finite branch of the square 31Q (sub_divisi.) determines the overlap (for example, divide each virtual pixel into 4 by 4 branch coordinates, or any other sub-division), and use Minute "Ub_divisi〇n) to estimate the number of overlapping coverage area of an output pixel. The overlap calculation of the output pixel overlap engine 230 can be simplified using some of the branch sampling characteristics, as follows: The output pixel is approximately within the largest rectangle defined by qUa (jriiateral approximation), and all branch samples are within the overlap region. 23 200841702 ♦ In this output pixel, the approximate quadrilateral is defined outside the minimum rectangle, all the knives The sampling system is not within the overlap region. • The % of the total number of samples taken between the two defined rectangles is an effective approximation of the number in the overlap. Next, the output pixel content engine 240 is used by Each virtual pixel is subtracted by the number of related overlapping branches, and after summing up the result, the content of the output pixel is determined by dividing by the total number of weights. Then, the output pixel content engine 24G outputs The content of the mosquito to the light engine is used to display the content of the decision. A more detailed description of the method for optical distortion is shown in Figure 15. In one embodiment of the present invention, the image processor is Executing - Hai method 800. In an embodiment of the invention, the image processor or a plurality of image processor 11 executes a plurality of conditions of the method (for example: Each one is used for any of red, green, and blue colors. First, the extraction (step 8ig) outputs the pixel centroids, which are read from the memory and placed in the FIFOs (for example: I 5 After capturing (step 810), the position of the diagonally adjacent pixels in the FIFOs memory, the shirt (step 82Q) the desired input: the diagonal adjacent output pixels of the pixel Next, determining (step (4)) is between the diagonal adjacent pixel and the desired pixel (hW - Ο; next, determining (step 840) that the output pixel covers the overlap of the virtual pixel; and according to Overlap, decision (step 85) to rotate the pixel content; then, the determined output pixel content is output to a light engine for projection to a display. The =800 will repeat the above steps for additional output pixels. Until all the contents of the output are completed by mosquitoes (steps are called. It must be noted that the pixel weight 24 200841702 new mapping procedure (pixel remapping procees) is a single pass process, and the pixel remapping procedure (pixel Remapping p Rocees) does not require information on the optical axis position. [concatenate adaptive algorithms for projection displays] for flat panel displays using LCD or plasma technologies ( In the case of flat panel displays, there is no geometric distortion of the image from the display itself, but not to the projection display. Projection optics f) (projection optics) will magnify images from a digital light modulator 50-100 times to a typical 50" or 60" projection display, which introduces a focal length ( Focus ), spherical aberration, chromatic aberrations, astigmatism, distortion, and color convergence, just like the optics of an image capture device. Although the physical distortion will be different, the principle of the centroid can still be used. Therefore, it is feasible to simultaneously connect the centroid principle to enable simultaneous correction of image capture and display distortion. Taking the point 420 of Figure 16 as an example, it can summarize the display geometry correction of [X+3.5, Y+1.5] and the image acquisition geometric correction of the [Χ+2·5, Υ+1.5], and string Connected to [Χ+6, Υ+3], and the final centroid is point 430, the tanned centroid image can be calculated early, using the same mark, brightness and contrast distortion corrected image Can be calculated. The above description of the various embodiments of the present invention is for illustrative purposes only, and other variations and modifications of the foregoing embodiments and methods are possible in accordance with the teachings herein. 25 200841702 2 wins the parts of the invention can be used - the stylized general purpose digital computer, the application of the integrated circuit, or the connection of the traditional parts of the network and electricity to perform 'this connection can be wired, wireless , data machine, and so on. The invention is not intended to be limited or limited, and the invention is limited only by the scope of the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS The unrestricted and unresolved embodiments of the present invention are described in conjunction with the following drawings, wherein the same reference numerals are used for the similar parts of all the different views unless otherwise specified. . BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a block diagram of a conventional video capture system. Figure 2A is a block diagram illustrating an image capture system in accordance with an embodiment of the present invention. The second embodiment is a block diagram of an adaptive image capturing system according to another embodiment of the present invention.

Ci f 3 A is a schematic image obtained from a conventional image couching system. Figure 3B is a schematic diagram of the image taken from a wide-angle adjustable image. Figure 4A shows a checker board pattern in front of the lightbox (lightb〇x) as geometry and brightness. Correction. Figure 4B shows the relative position of the camera in the calibration procedure. Figure 4C shows that in a typical calibration setup, the checker board pattern is not exactly perpendicular to the camera. 0 26 200841702 Figure 5A shows the barrel effect exhibited by a typical camera/lens system. ▲ Figure 5B shows the lens system shown by a typical camera. Brightness fall off. Figure 6 is a schematic diagram showing a 4-pixel video data image with red, green and blue content and each having 8 bits. Figure 7 shows a 4-pixel video data image. Schematic diagram, which has red, green, and blue content and each has 8 bits, and an additional 2 bits of plane, for (for storage brightness and contrast correction, and geometric correction data. Figure 8 shows Block diagram of the image processor. Figure 9 shows the discoid pattern (such as ~b〇ard _ - the severe out-of-focus image, and the -graphic method to determine the intersection of two black squares placed diagonally Figure 10 shows the image of π bright distortion (4) and the corrected distortion-free display range. Figure 11 is a schematic diagram showing the output pixel mapped to the virtual pixel coordinate of the image. Figure 1 2 shows the Schematic diagram of the calibration procedure inputting the centroid. 3 Diagram Description - Schematic diagram of the calculation of the output pixel angular position (Cai er). The brother 1 4 diagram illustrates the schematic diagram of the overlapping approximation of the pixel branch (_·-). An adaptation method for optical distortion. σ Figure 16 illustrates the display of an output pixel mapping to a virtual pixel coordinate square of a display, and then mapping to the virtual image of the image capture device 27 200841702 Prime coordinates [Main component symbol description] Component number English Chinese 10 0 Adaptive image acquisition system Adjustable image capture system 110 Adaptive image processor Adjustable image processor 111 Adaptive i Mage processor Adjustable Image Processor 12 0 Memory Memory 12 1 Memory Memory 13 0 Sensor Sensor 13 1 Sensor Sensor 14 0 Image processing Image Processing 14 1 Image processing Image processing 15 0 Camera control circuit Camera control circuit 15 1 Camera control circuit Camera control circuit 16 0 Output formatter Output formatter 16 1 Output formatter Output formatter 17 0 Optical lens systems Optical lens system 17 1 Optical lens systems 18 1 Adaptive image processor Adjustable image processing 19 0 Display device Display device 19 1 Display device Display device 28 200841702 2 10 Output pixel centroid engine Output pixel centroid engine 2 2 0 Adjacent output pixel engine Adjacent output pixel engine 2 3 0 Output pixel overlay engine Output pixel overlap engine 2 4 0 Output pixel content engine Output pixel content engine 3 10 Distorted imaging area Distorted image area 4 10 First output pixel First output pixel 4 2 0 Second output pixel Second output pixel 6 10 Halfway point Midway point 6 2 0 Output pixel Output pixel 7 3 0 Corrected display area Corrected display area 8 0 0 Method Method 8 10 Acquire output pixel centroid Capture output pixel centroid 8 2 0 Determine diagonally adjacent output pixels Angle adjacent output pixel 8 3 0 Determine halfway point between centroid and adjacent output pixel Determines the midway point between the centroid and the adjacent output pixel 8 4 0 Determine overlay Determines the overlap 8 5 0 Determine content Decides content 9 0 0 Sensor Array sensing array 9 0 1 Line 9 0 2 Line 9 0 3 Line 29 200841702

Ο 9 0 4 Black square Black square 9 0 5 Black square Black square 30

Claims (1)

200841702 X. Patent application scope: 1 · A method for image capture, comprising: taking a centroid of the output pixel for a plurality of output pixels; determining a first output image out pixel from the plurality of output pixels; The adjacent input is based on the extracted output pixel centroid and the adjacent output pixel ～ the first output pixel covers the overlap of the virtual pixels, and the virtual image is: corresponding to an input image;决定 determining the content of the _th output based on the content of the superimposed virtual pixel; and outputting the determined content by the tiling. 2. The method of claim 4, wherein the extraction system reads three rows of output pixel centroid data to the memory. 3. The method of claim 2, wherein the adjacent output pixels determine the adjacent output pixels of the diagonal. 4. The method of claim 3, wherein the adjacent output of the diagonal c, the adjacent memory location of the diagonal in the hidden body. 5. The method described in the fourth paragraph of the patent application scope, wherein the decision of the overlap is as follows: the virtual pixel to the minimum of 2 Fenshan - the eclipse 2 by 2 sub-area (Sub-regi〇n And the sub-area in which one of the virtual pixels in the virtual pixel is overlapped by the output pixel, and the method of determining the content is used for the early morning color. 7. If the method of claim 6, the method of the content, and the method of the content of the patent, the content of the content and the transmission of the 2008 41702 is repeated for additional colors. The determination of the content is such that the method described in item 1 of the patent application is added and divided. 9. The method of claim r, wherein the method operates in a pipeline. 10. As described in the first paragraph of the patent application, Xiangshan is included in the determined valley to incorporate the overlap as brightness data. f) U. As described in item 1 of the scope of the patent application, wherein the output further includes an optical distortion characteristic embedded in or exempted from the output. 12·—A system for image capture, comprising: an output pixel centroid engine, which converts the helmet to a plurality of output pixels to capture the centroid of the output pixel; - an adjacent output pixel engine, the system Communicating with the output pixel centroid engine, the adjacent output pixel engine can determine the adjacent output pixels of the output pixel from the plurality of output pixels; (j - output pixel overlap engine, which is adjacent to the output pixel The output pixel engine passes m light = according to the extracted pixel centroid and the adjacent output pixel, the fourth output pixel covers the overlap of the virtual pixels, and the virtual pixel corresponds to an input video. And, the Bayi output pixel content engine, which is coupled to the output pixel overlap engine, can determine the inner valley of the first output pixel according to the content of the overlapped virtual pixel, and can output the determined content. The system of claim 12, wherein the output pixel centroid engine system reads the three columns of output pixel centroid data to the memory. 32 200841702 14. If applying The system of claim 13, wherein the phase engine determines a phase = pixel 15 by determining a diagonal phase (four) (four). The system is as described in the patent application scope, wherein the adjacent wheel image; The system is a system for reading the adjacent memory locations diagonally opposite to the memory in the memory, and the two diagonally adjacent output pixels. The system of claim 12, wherein the engine determines the The overlapping portion is obtained by subdividing the virtual pixel to a minimum of 2 by 2: 〇 =: (, -), and the number of sub-regions overlapped by each virtual (four) ^ ^ pixel. As described in claim 12 The system, wherein the output pixel content engine determines a content for a single color. The system of claim 17, wherein the output pixel content engine determines content and outputs for additional The system of claim 12, wherein the output pixel content engine utilizes addition and division only. ◎ 20. The system of claim 12, wherein System department The system of claim 12, wherein the output pixel content engine is immersed in the determined content as brightness data. The system of claim 12, further comprising an adaptable image processor for embedding geometric or luminance optical distortion features in the output. 23·-- capturing image system, comprising: , which is a plurality of output pixels for extracting the centroid of the output image 33 200841702; one determines the adjacent output pixel to determine the U output image I ^ #, the self-complex number of output pixels, and the adjacent output of the pixel Pixel; a 疋 疋 , , , , , 发 发 发 发 发 发 发 发 发 发 发 发 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 根据 决定 决定 决定The virtual pixel system skew τ sub-head corresponds to an input image; a decision content device, the post-pot g corps according to the content of the overlapped virtual pixel, Used to determine the content of the first round of pixels; and - the rounding device ' is used to output the determined content. ϋ 34