TWI358231B - Auto-focus method for camera and digital camera - Google Patents

Description

1358231 S3U07-0002 23194twf.d〇c/n IX. Description of the invention: [Technology of the invention] The present invention relates to a camera autofocus method, and more particularly to a method for automatically taking a camera focus using a single image. The method of location. [Prior Art] With the development of technology and the universal use of digital devices, the relationship between people and digital products is becoming more and more inseparable. Among digital products, digital cameras (Digital Camera, DC) occupy a very important place. In order to determine the pros and cons of a digital camera, in addition to the image quality of the camera, the camera technology is neglected. Listening not only needs to be fast, but also the speed of the 二1 difocal speed and the focusing effect will directly affect the imaging speed. No focus: And: 'Use the feature function to discriminate the camera. Please refer to the figure! The camera is in the schematic diagram of the camera. Do not take images, the over-the-counter stitches, and the framing images that are captured by multiple focus points must pass the position of ^^^^^. Therefore, the process of taking a point, looking for the blur of the point - and then returning to the clearness of this in this way is quite a waste of time. Not only does it have to be used in multiple focus points. However, the current power consumption of the Xiao 2 camera is naturally very particular. For the m not f, the battery life of the camera and the battery life of the camera are very heavy. 4 1358231 S3U07 -0002 23194twf, doc/n view. Therefore, if you can focus more in a power-saving manner, you can greatly improve the battery life of the camera. The manufacturer of this camera is not in a hurry to seek an appropriate solution. The method of the present invention is to overcome the above problems. [Invention] The present invention provides an autofocus method for a camera to save the convergence time, and the _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The body cost, improve the enamel and reduce the volume of the digital camera. The invention provides a camera autofocus method comprising the following steps: setting a first parameter related to the camera in one step P and the second parameter q. In another step, a unique image is acquired. In a further step /, the object distance is calculated according to the image, the first parameter P and the second parameter q. In a further step, 'based on In the embodiment of the present invention, in the camera autofocus method, the step of setting the first parameter P and the second parameter q includes the following steps: at the variable distance Di, Shooting the same point source separately, and obtaining the corresponding image Fi; using the Gaussian distribution and calculating the corresponding diffusion parameter σί according to the image ;; establishing the data set according to the variable distance Di and its corresponding diffusion parameter (Di, 〇i According to the data set (Di, (yi), the first parameter is set? and the second parameter q, where i is the corresponding number. In an embodiment of the invention, the above camera autofocus method, 5 S3U07 -0002 23194twf.doc/n The step of calculating the object distance according to the image, the first parameter P and the second parameter q comprises the following steps: _ boundary algorithm 镰 image boundary; estimated target diffusion parameter σ; according to target diffusion parameter σ, First reference The number p and the second parameter q calculate the object distance. First, the invention provides a digital city, including the silk component, the video image broadcast and the digital signal processing 11. The optical focusing component is used to adjust the difficult position of the digital position. The 彡 龄 单元 配置 配置 配置 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学 光学And the second parameter q. The image capturing unit captures only the second parameter: the image. The digital signal processor is based on the image, the first parameter P and the first signal. The distance is from the digital signal processor and the focusing light is sent according to the object distance. 70 pieces are focused according to the focusing signal. 2^Invented in the -step towel, set the oil machine to reduce the first, - parameter p and the second step in the second step - in the other step - collect the only image. In another step, the object distance is calculated based on the image, the first parameter p and the second parameter q. Only V Cai's adjusted the position according to the object distance. Therefore, the focal length can be adjusted by analyzing the ~ image, and the focusing time is greatly reduced. For the above (four) above (four) and the excellent position is more obvious, the following is only a case, with the accompanying drawings, as detailed below. [Embodiment] FIG. 2B is a flowchart of a method of the present invention, which is an embodiment of the present invention. FIG. 2B is a flowchart of an autofocus side S3U07-0002 23194twf.doc/n method according to an embodiment of the present invention. Referring to Figures 2A and 2B in combination, the light: focusing element 2 is used. , image manipulation unit 3. With η 40. Optical suspicion, component 2 〇 (4) adjustment of digital phase 2 2 capture unit 30 置于 has been placed in the optical focusing component 2 〇 I, position. The shirt is like an image. The image manipulation unit 30 is, for example, a charge=piece: a coupled device CCD), an optical focus component, and a shadow. First, the first step is to set the phase-parameter and the correlation parameter by the digital signal processor 40. Two parameters q. Step (4), the image is operated by a single 30-inch image. In step S2, the digital signal processor 40 calculates the object based on the image, the first parameter p and the second parameter q. Step S204 ’ The digital processor 4G assigns a difficult signal to the optical focus το 2G according to the object, and the optical dynasty 2 () is focused according to the signal. This makes it difficult to perform only a single image, greatly reducing the pair f and _ of the digital camera. Then take a more detailed description of the steps. Figure 3 is a schematic illustration of an imaging system in accordance with an embodiment of the present invention. Please refer to FIG. 3 , wherein the object distance u is a distance between the light source 〇 and the lens 2 〇 1 's is the distance between the defocus plane and the lens 2 〇 1 , f is the focus of the lens 2 〇 1 , and D is the aperture of the lens 201 , d Imaging the point source in the defocus plane. From Fig. 3, the object distance u and the aperture D have the following relationship (丨): w = 7^b^.·. Formula (1), where F is the aperture coefficient of the lens 2〇1. Considering the effects of diffraction (Diffracti〇nEffect) and S3U07-0002 23194twf.doc/n lens imaging, the point of the good surface is Gaussian. Where r is the Gaussian distribution leaving the mid-range parameter. ^, and k is a constant of the mosquitoes made by the characteristics. The relationship between the formula (1) and the point spread can be as shown in the following formula (2): / double σ 9~( -f-Fd S - V . · formula (2), first Parameter P and

The second parameter q is a relatively systematic parameter, so the correct estimation of the object distance II is correctly estimated by correcting σ. This is followed by a more detailed description of the second parameter q. 7 Further parameter P ... Fig. 4 is a flow chart for setting the first parameter p and the first parameter q of the invention. Referring to FIG. 4 and FIG. 2a and FIG. 3, firstly, in step S401, the digital camera 1 captures the same point light source at a plurality of variable distances to obtain a corresponding plurality of images Fi. In step S4〇2, the Gaussian distribution is used and the corresponding diffusion parameter 〇i is calculated according to the corresponding plurality of images Fi, respectively. Step S403, establishing a data set (Di, ai) according to the plurality of variable distances Di and their corresponding diffusion parameters σί. In step S404, the first parameter p and the second parameter q are set according to the data set (Di, cii), where 丨 is the corresponding number. In other words, the more data collected by the data set (Di, ai), the more accurate the first parameter p and the second parameter q are estimated. Then, according to the data set (Dipi) and the cost function = -^^-) 2, the first parameter P and the second parameter q are estimated using an iterative method, for example, using the Newton's method *q-σΐ (Newton's Method). It should be understood by those skilled in the art that the "Newton descent method" exemplified in this embodiment is only a specific embodiment. In another embodiment, the iterative method can still be 1358231 S3U07-0002 23194 twf.doc/n descent method Interleaved with the steepest descent method... = f is set for this particular embodiment. Finally, the number of times P > q is set in the digital signal processor 40. The digital camera 1 (four) can calculate the object distance u by the single image and the formula (2).

It is worth mentioning that 'Steps S4〇1~4〇4 can be set before the digital camera ι〇 factory. 'The consumer can set the first parameter P and the second parameter q when making the camera (4) Trouble. It should be understood by those skilled in the art that the steps S4G2 to S4() 4 can be performed by the digital signal processor 40. Other external operators may also be utilized in another embodiment, such as using a personal computer for operations. Therefore, the invention is not limited thereto. When the first parameter P and the second parameter q are set, when the consumer uses the digital camera 10, only a single single image can be captured to focus. Next, a more detailed description of how to calculate the object distance u based on a single single image is given.

Figure 5 is a flow chart for calculating the object distance based on a unique single image, a first parameter P, and a second parameter q, in accordance with an embodiment of the present invention. Please refer to FIG. 5 and FIG. 2A'. In the present embodiment, the following steps are performed by the digital signal processor 4 to illustrate. First, in step S501, the boundary algorithm is used to capture the boundary of a single image. Step S502' estimates the target diffusion number σ. In step S503, the object distance u is calculated according to the target diffusion parameter σ, the first parameter p and the second parameter q, for example, the object distance u is calculated using the formula (2). Next, step S501 will be described in more detail. 6A is a flow diagram of obtaining a boundary of an image using a boundary calculus 9 S3U07-0002 23194 twf.doc/n method in accordance with an embodiment of the present invention. Figure 6B is a schematic illustration of a horizontal operand of a boundary algorithm in accordance with an embodiment of the present invention. Figure 6C is a schematic illustration of a vertical operand of a boundary algorithm in accordance with an embodiment of the present invention. Referring to Figures 6A, 6B and 6C in combination, in the present embodiment, a Sobel boundary algorithm is described. Gradient boundary algorithms can also be used in another embodiment. First, in step S601, a plurality of regions of the unique single image captured are respectively convolved with a horizontal operation unit 601 to obtain a horizontal edge response (Edge Response) of the plurality of pixels of the single image. The operand can also be called a mask or a kernel. Step S602, the plurality of regions of the unique single image captured are respectively distributed with the vertical operation unit 6〇2 to obtain a vertical edge response of the plurality of pixels of the single image. It is worth mentioning that the 'horizontal operation element 601 has the characteristic of strengthening the horizontal boundary. Vertical operand 602 has the property of enhancing vertical boundaries. In order to combine the characteristics of the horizontal operation unit 601 and the vertical operation unit 602, the larger of the horizontal edge response and the vertical edge response of any element in the image can be taken as the output of the pixel by step S603'. value. In this way, the line connecting multiple elements with the maximum value is considered to be the boundary of the unique single image. The extreme value described here is, for example, the upper limit value of the gray scale value that the pixel can represent. A person having ordinary knowledge in the art can also determine a set value as a basis for determining whether the output value of the pixel is an extreme value. In addition, the "operating element" exemplified in FIGS. 6B and 6C of the present embodiment is only a specific embodiment. In another embodiment, the operating element can still use a gradient operation element, or other weights and sizes (Size) operation elements. Therefore, the present invention 1352831 S3U07-0002 23194 twf.doc/n should not be limited to a more detailed description of such a specific embodiment. . Next, for one of the embodiments of step S502, an iterative method is used to estimate the flow of the target diffusion parameter σ. Fig. 8A is a schematic view showing the division of the image of the Xiangbian County into two regions in accordance with the present invention. The graph is based on a flow-flow graph that estimates the target diffusion parameter σ in an iterative manner.凊合幷参日g圄7, βΔ你〇-Μ鲜, in this embodiment, iterate

The example is explained. In another embodiment, the iterative method also allows for the fastest rate of money. First, by step Shan, it is estimated that ν==(§1'β2'σ) 'hypothetically--the hypothetical boundary divides the early reading s彳 into the first-imaginary_gl and the second imaginary area g2, and then step S702 , define the cost C(v>= Σ 2φΓ^ΒΆγ , (〇, _ σ σ JJ , whose number /(a,b) = gVφ[^^)]_g2φτ£(α>ό>Ί.藤藤<c<;ni ,

σ ^ σ \ divides the image into the first-real area gl by the boundary of the image obtained in step S501, and the second actual area g2, M is the image-area. (a, b) is the responsibility of the region - the symplectic coordinate ed_ is the distance between the atomic coordinates (a, b) and the imaginary boundary of the image. d, (a, b) is the distance between the coordinates of the pixel (a, b) and the boundary of the image β φ is a Gaussian distribution function. In other words, 'f(a,b) is a known value] is an estimated value obtained by convolving the first imaginary region with a Gaussian distribution function, and estimating the convolution of the second imaginary region g2 and the Gaussian distribution function. value. The cost function C(v) is the best target diffusion through the minimum mean square error. 11 1358231 S3U07-0002 23194twf.doc/n Parameter σ. In the above, the calculation of step 703, ΏΏ盥ac(7).^ initial dg2 do. Step S704 calculates Δ▽. ▽匸(7), that is, a set of adjusted values of V, based on the calculated value of 22, ^2 disk dC(V) initial dgl I and the current v value substituted into the cost function c(v). In step S7〇5, V is corrected according to ν==ν+Δν; in step S706, the iterative method is completed when I Δν h, at an error value or the number of iterations reaches the setting side. It is worth mentioning that there are people in the field who can set the error value and set value according to their needs. If | Δv | is not smaller than the error value and the number of iterations does not reach the set value, then return to step S703 to execute the iterative method. In this way, a better target diffusion parameter σ can be obtained. Since the first parameter p, the second parameter ^ and the target diffusion parameter σ are all known at this time, according to the formula (2, the object distance u can be obtained. Furthermore, it is worth mentioning that although the above embodiment has already There is a possible pattern for the ancient ten-objects from the U, but those who have the knowledge in the technical field should know that the waste merchants have different methods for calculating the object distance. The need to replace the above algorithms with a simpler formula 'to save computing time and reduce hardware costs. In other words, f if the method of calculating her u meets the above #_, ^ is in line with the spirit of the present invention. Referring to FIG. 2A again, after the object distance U is obtained, the digital signal processing benefit 40 can output the focusing signal to the optical focusing component 2 according to the object distance U. The focusing signal includes the mirror of the digital camera. It depends on the distance to be adjusted and the direction of adjustment. The optical age component 2G focuses on the difficult signal, 12 1358231 S3U07-0002 23194twf.doc/n Makes the image class 3 〇, the quality of your image, and reduces ' Second consumption image. Borrow Enhancing the captured image The embodiment of the present invention has at least the following advantages: The object camera image can be used to obtain the object distance u, thereby adjusting the number (10) between the acceleration objects, and more. 2: Example: The position camera does not require a mirror It can be..., power b, hunting to save hardware cost and power, improve picture quality, and can reduce the volume of digital camera. Although the present invention has been disclosed in the preferred embodiment as above, Tie Yue's in any technical field Those who have the usual knowledge, no, the scope of the patent application is as follows: [Simplified illustration of the drawing] Figure 1 is a schematic diagram of a conventional camera focusing process. Figure 2 is in accordance with the present invention 1 is a flow chart of an autofocus method in accordance with an embodiment of the present invention. Fig. 3 is a schematic diagram of an imaging system in accordance with an embodiment of the present invention. A flowchart for setting a first parameter ρ 13 S3U07-0002 23194 twf.doc/n and a second parameter q according to an embodiment of the present invention. FIG. 5 is a single sheet according to an embodiment of the present invention. FIG. 6A is a flowchart of obtaining a boundary of an image by using a boundary algorithm according to an embodiment of the present invention. FIG. 6B is a flowchart according to the present invention. FIG. 6C is a schematic diagram of a vertical operation unit of a boundary algorithm according to an embodiment of the present invention. FIG. 7 is a schematic diagram of a vertical operation unit according to an embodiment of the present invention. A flow chart for estimating the target diffusion parameter σ by an iterative method. Figure 8A is a schematic diagram of the imaginary boundary dividing the scene into two regions according to an embodiment of the present invention. ', ' a, i Figure 8B is an implementation in accordance with the present invention A flow chart for estimating the diffusion parameter σ by an iterative method. < [Description of main component symbols] 10: Digital camera 20: Optical focusing component 30: Image capturing unit 40: Digital signal processor 201 • Lens 601: Horizontal operation unit 602: Vertical operation unit 1358231 S3U07-0002 23194twf.doc /n U : object distance S: distance between defocus plane and lens f: focus D: aperture

d: imaging of the defocus plane gl: first imaginary area g2: second imaginary area gl': first actual area g2': second actual area S201 to S204: steps S401 to S404 of an autofocus method of FIG. 2B FIG. 4 is a step S501 to S503 for setting the first parameter p and the second parameter q: each step S601 of FIG. 5 for calculating the object distance according to the unique single image, the first parameter P and the second parameter q. S603: FIG. 6A is a step of obtaining a boundary of an image by using a boundary algorithm

S701~S706: FIG. 7 is a step of estimating the target diffusion parameter σ by an iterative method. 15

Claims (1)

1358231 | cto years old. Month J Day Amendment 100-10-3 X. Application Patent Range: 1. A camera autofocus method, including the following steps: Shooting the same point source separately at multiple variable distances Di, and obtaining corresponding multiple sheets Image Fi; using a Gaussian distribution and calculating corresponding diffusion parameters σί according to the images Fi respectively; And establishing a data set (Di, Gi) according to the variable distance Di and the corresponding diffusion parameter σί; setting the first parameter p and the second parameter q according to the data set (Di, oi), wherein i For the corresponding number; acquiring an image; calculating an object distance according to the image, the first parameter P and the second parameter q; and adjusting the focus position according to the object distance. 2. The camera autofocus method according to claim 1, wherein the step of setting the first parameter p and the second parameter q according to the data set (Di, ai) comprises the following steps: The first parameter p and the second parameter q are set according to the data set (Di, (?i) and a cost function, wherein the cost function is q-Gi ρ(ρ^) = Σφί-3 The camera autofocus method according to claim 2, wherein the iterative method is a Newton's descent method. 4. The camera autofocus method according to claim 1, wherein the image is based on the image, The first parameter P, and the second parameter q count 16 1358231 100-10-3 to calculate the object distance, comprising the following steps: using a boundary algorithm to capture the boundary of the image; estimating a target diffusion parameter σ; And calculating the object distance according to the target diffusion parameter σ, the first parameter ρ and the second parameter q. 5. The camera autofocus method according to claim 4, wherein the boundary algorithm is utilized Steps to capture the boundaries of the image, package The following steps: convolving a plurality of regions of the image with a horizontal operation unit to obtain a horizontal edge response of the plurality of pixels of the image; and respectively convolving the regions of the image with a vertical operation element Obtaining a vertical edge response of the pixels of the image; and taking the larger of the water edge response of the pixel and the vertical edge response of the pixel as the output value of the pixel, wherein the output value is The connection between the plurality of elements of the extreme value is intended to be the boundary of the image. 6. The camera autofocus method according to claim 5, wherein the horizontal operation unit and the vertical operation element are Sobel operations. 7. The camera autofocus method according to claim 5, wherein the horizontal operation unit and the vertical operation unit are gradient operation elements. 8. The camera automatically as described in claim 4 a focal method, wherein the step of estimating the target diffusion parameter σ comprises the steps of: estimating the target diffusion parameter σ by a delivery method. The camera autofocus method described in Patent No. 8, The sigma iterative method is a gradient descent method. 17 1358231 100-10-3 10. The method of estimating the target diffusion parameter σ by the iterative method according to the camera autofocus method described in claim 8 The method includes the following steps: estimating a set of initial values of 30 V, where v==(gl, g2(6), an imaginary boundary dividing the image into the -first-imaginary region gl and the second hypothetical region g2; defining a cost function , where system = Buzhou ^], the boundary of the image divides the image into a -first-real area gl' and a second actual area g2, where M is a region of the image, and (a, b) is the region Any of the pixel coordinates, d(a, b)g, the distance between the pixel coordinates and the imaginary boundary of the image, d, (a, b) is the distance between the pixel coordinates and the boundary of the δ hai image. φ is the Gaussian distribution number calculation tgl ^ do 'calculate Δν = ν(:(ν), that is, according to the calculated @, $22诳dC(V) initial ^2 and substituting the V value into the cost function Q&gt ;) Calculate the value obtained, obtain a set of adjustment values Δν of v; correct ν according to ν = ν + Δν; and when | Δν An error value is smaller than the number of iterations reaches or - the set value of the iteration method is completed. The camera autofocus method of claim 4, wherein the step of calculating the object distance according to the target diffusion parameter σ, the first parameter ρ and the second parameter q comprises the following steps: 18 1358231 100-10-3 Calculate the object distance 'where u is the object distance. 12. A digital camera comprising: an optical focusing component for adjusting a focus position of the digital camera; an image capturing unit disposed on an optical path of the optical focusing component for capturing an image; and a digital position The signal processing thief, the optical focusing component disk is connected to the image capturing unit; ' ' wherein the digital camera captures the same point light source at a plurality of variable distances Di, so that the image capturing unit obtains correspondingly The image is Fi, and the digital signal processing state utilizes the Nans distribution and respectively according to the diffusion parameters (1) corresponding to the images, and the digital signal processor according to the variable distance Di and its corresponding diffusion parameter Σί establishes a data set (〇ι, σι) 'the digital signal processor and sets the first parameter p and the second parameter q according to the data set, where 丨 is the corresponding number & The digital signal processor further calculates an object distance according to the image, the first parameter ρ and the third parameter q, and the digital signal processor according to the object distance ^ = withered 5fL number, § hai optical focusing component according to the adjustment The signal is focused. 13. The digital camera according to claim 12, wherein the digital signal processor sets the first one according to the data set (Ι) ί, σί) and a cost function. Parameters: 卩 and the second parameter q, where = the valence function is c(p, ==)2. ^代 / q-〇i 14. As for the digital camera described in claim 13, the iterative method is the Newton's descent method. ", 中中15. The digital camera described in claim 12, wherein 19 1358231 100-10-3 digital signal engraving _-boundary algorithm _ takes the boundary of the image, and the digital signal processing The target estimates a target diffusion parameter σ, and the digital signal processor calculates the object distance according to the target diffusion parameter σ, the first parameter ρ and the second parameter q. 16. The digital camera of claim 15, wherein the digital signal processor convolves the plurality of regions of the image with a horizontal operation element to obtain a plurality of horizontal edges of the oblique image. Responding, and the digital signal processor convolves the regions of the image into a vertical transport element to obtain a vertical edge response of the pixels of the image, and the digital signal processing takes any command The horizontal edge of the element should be larger than the vertical edge response of the element, and the line of the plurality of pixels whose output value is the extreme value is regarded as the boundary of the image.数. The digital camera of claim 16, wherein the digital signal processing 1 is a water conversion unit and a Sobel operator. 18. The digital camera according to claim 16, wherein the horizontal operation unit and the vertical operation unit are used: the number is specifically for the reading (four) of the item 15 of the range, wherein the number is Subtraction processing (4) - Iterative method estimates the target diffusion parameter 〇. 20: The digital camera according to claim 19, wherein the iterative method of digital signal processing 1 is a gradient eight 21= application for the camera described in item 19, and the digital signal processing 11 estimate 1 - the initial value of the group, its h = (gtg2 & i 20 1 * 358231 100-10-3 Yi Yiyi a cost function C (v) = where / (also head, the shadow:; the actual edge of the image The shirt image is divided into a first intersecting area gi' and a second actual area, where is a region of the image, (a, b) is any of the pixel coordinates of the region, and d(a, b) is The distance between the pixel coordinates and the imaginary boundary of the image, d, (a, b) is the The distance between the pixel coordinates and the actual boundary of the image, φ is a Gaussian distribution function, and the digital signal processor calculates, dC(V)^dC(y), bit signal processing 1 lining Δν=ν(χνρ也| y j is calculated according to the calculation, wing * interesting (four) day, # ^ system and the v value of the target is substituted into the value of the like function c (v) = to obtain the value of ν - _ integral value Λ ν, the digit Signal New H and ν = ν + Δν for ν into the positive bit signal processor _ whether to complete the iterative method, t 丨 ^ difference or a number of similar phase - set the dare _ iterative method. The digital signal processor divides the image into a first rice area gl and a first imaginary area g2 by using an imaginary boundary, and the digital signal processing 22. The digital camera according to claim 15 of the patent, the bean In the present and the present, the heart is reduced according to "six calculating the object distance, and its towel u is the object distance: 21