TW201044317A - Method of transforming two-dimensional image into three-dimensional image - Google Patents

Abstract

A method of transforming a two-dimensional image into a three-dimensional image is provided. First, a two-dimensional image is received. Then, the two-dimensional image is transformed into a gray level image. Then, a plurality of objects in the gray level image is extracted. The plurality of objects in the gray level image is marked. Next, according to the plurality of objects in the gray level image and a distance between the first boundaries of the gray level image, a plurality of depths corresponding to the plurality of objects is determined. Moreover, according to the plurality of depths corresponding to the plurality of objects, a depth plot is generated. Therefore, the depth plot and the two-dimensional image may be combined to produce a three-dimensional image.

Description

201044317 υνιυυ^οιΓΨ 30913twf.doc/n VI. Description of the invention: [Technology

The present invention is a method for converting a two-dimensional image into a three-dimensional image, and in particular, a method for generating a depth map corresponding to a two-dimensional image and forcing a two-dimensional image and a depth map to be converted into a three-dimensional image. Q [Prior Art] ¢) In general, &3, in the type of three-dimensional (3D) image display, the Barrier Type 3D image display uses the binocular parallax method to make the human eye feel To 3D images.疋 Figure 1 is a schematic diagram of a fenced 3D image display architecture. Bamboo shooter Figure 1 'Add a fence 10 for splitting light on the backlight 3 of the LCD monitor, so that the Pixel of the liquid crystal display can be divided into two categories. The eye image, the pixel labeled 2, displays the right eye image of the person, and the image is synthesized to produce a three-dimensional image display. ' A technique is disclosed in US Pat. No. 0232666, the technical basis of which is based on the previous and current The motion vector, brightness, or color value of the face is edge detected to generate a depth map. It is worth mentioning that this type of technology is detecting objects.

It is prone to serious errors, resulting in a wrong depth map, which affects the quality of 3D images. V SUMMARY OF THE INVENTION The present invention provides a method for converting a two-dimensional image into a three-dimensional image. 201044317 uyiUU4 (Diiw 30913twf.doc/n method, which can directly extract objects from a two-dimensional image with an image and a depth map extracted You can use $#_ 'wood and map, combined with 2D shadow, you need other information, no additional invention is not like converting to 3D image. Butterfly Hanbei can provide a kind of 将-络 & Including the following steps. First, the square-dimensional image of the image of the three-dimensional image is converted into a gray-scale image. 7;, the shirt. Then 'the next two, the fine material _ pieces into the body; = corresponding to the depth value. Then, according to ===: like the depth map. Finally, combined with the depth map and the two-dimensional shadow::= order greedy feeding, the upper method, the towel in the quenching ash step includes the gray The image of the edge detection expansion (MatiGn) operation Saki @ $ 异 贝 进 进 进 填 填 填 填 Γ Γ Γ Γ Γ Γ Γ Γ Γ Γ Γ Γ 所 所 所 所 所 所 所 所 所 所 所 所 所 所 所 所 所 所 所Edge gamma for gray-scale image (Edge Esti_ 2=:Prewitt mask (Mask) for grayscale Shadow edge, the equation for edge detection is as follows: v/ = |GxxjP| + |q1;xP| where ▽/ indicates the data of the data is 3*3 matrix, 201044317 uyiUU4〇i FW 30913twf.doc/ n

Gx and Gy are 3*3 matrix Prewitt masks (j^ask)

Gy is expressed as follows: -i ο Γ

Gx=

Gy= -1 Ο 1 -1 Ο 1 0 0 0 1 1 1

O In one embodiment of the invention, the t method described above wherein the step of expanding the operation of the lean material to join the edge of the object captures the contour of the object comprises the following steps. First, 'choose a 昼 矩 a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a Then 'judge the prime money:: 21 first value.镰’ When the prime data p5 is equal to the first value, and the material ΡπίνΡ6~P9 is at least one of them - not equal to - often, when looking for a value, then 昼

The prime material Ρ 5 is adjusted to a second value. Connect the I a double value test and repeat the above steps to perform the expansion operation on the transport data. In an embodiment of the invention, it is. , the method of returning the second value W, wherein the first value is in one embodiment of the invention, :; r = region step = step == 4 of its ===: get: first: scan shadow, prime data The left 昼 资料 资料 ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' 、 ' 、 、 、 、 、 、 、 、 、 、 、 、 ' ' 2010 2010 2010 2010 2010 2010 2010 2010 2010 Then, from right to left, the edge image is scanned by ^ to obtain the second scanned image, and when the ^ is equal to the first value, the second pixel data and the lower pixel data are equal to the second pixel data. 7 The material is adjusted to the second value. Then & 隹 ^. Once the image of the image is taken to fill the surface, the image is scanned and the second sweep is inserted. In the second embodiment, the above-mentioned first value is . In the example of the object tra, the above steps are included in the image in the grayscale image. First, the regions corresponding to the grayscale components are sequentially marked. Then, adjust the image of each object to convert the 2D image into a 3D image, and obtain the 2D image, expand, fill, and mark the object into 3D images. U-pictures ^ need other information'. The above characteristics and advantages and disadvantages are more obvious (4). The following is a specific example and is described in detail below with reference to the drawings. [Continuous application method] Wei Ying = Fang = ^ ^ Example _ two-dimensional image conversion to three to the machine private map. Please refer to FIG. 2, first, in step S2〇i, 201044317 ^>〇913twf.doc/n f to receive two two-dimensional images. Next, in step S2〇2, the two-dimensional image is converted into a grayscale image. Then, in step S203, the object in the grayscale image is taken. Next, in step S204, the objects in the grayscale image are marked. Then, in step S205, the deep LU value corresponding to each object is determined according to the distance between each object in the grayscale image and the -th boundary of the image of the image. Then, in step S207, the depth-value map is based on the depth value corresponding to each object. Finally, in step S207, the depth map and the two-dimensional image are combined to generate a three-dimensional image. 3 is a flow chart illustrating an embodiment of step S2〇3 in accordance with an embodiment of the present invention. Please refer to FIG. 2 and FIG. 3 together. In the present embodiment, the step s2〇3 includes steps S301 to S303. For example, edge detection is performed on grayscale images to generate an operational data. The computational data is then expanded to join the edges of the object to obtain an outline of the object to produce an edge image. ^, fill the area corresponding to the outline of each object to get a filled surface. Those having ordinary skill in the art can implement step S301 in any manner according to their needs, such as gradient concept detection, wavelet detection, and operator detection. Here, only the Prewitt mask is used as an example of the image two-dimensional differential operator detection image edge, and the Prewitt mask is a three-by-three matrix, and the judgment of detecting the image edge is as follows: +l(Qix^+^ 2x^+^3><^+C^x^+(^x^+(^xi>^7xi?^x^+G>9X^)| =| GaP( +| QrxP\ where Gx and Gy are Prewitt mask: 201044317 0910046ITW 30913twf.doc/n

Gxl 〇χ; ~-ι ο Γ % Gy2 ~-i -i -Γ Gx 2Gx, - 1 1 0 1 ^Gy = Gy4 Gys = 0 0 0 Gxl Gx9_ -1 0 1 1 1 1 P is the pixel corresponding Mask position: P3 P6 P9 P2 P5 P8 PM P4 P7 L_p Figure 4 is a schematic diagram illustrating the expansion operation in accordance with an embodiment of the present invention. Figure 5 is a schematic diagram showing the direction of operation of the expansion operation in accordance with one embodiment of the present invention. Please refer to FIG. 3 to FIG. 5 together. As shown in Fig. 5, in the calculation data 501, an expansion operation is performed from left to right and top to bottom to obtain an edge image by connecting the edge of the object to generate an edge image (step 302). The operation method is as follows. First, the pixel matrix P in the operation data 501 is selected: Pi Pi Pi Pa Ps P6 Pi Ps P9 , wherein the pixel data in the pixel matrix P is 1, 〇, 1, 〇, respectively. 〇, 1, 〇, 〇 and 〇. First, it is judged whether or not the pixel data p5 in the pixel matrix P is 〇. When the pixel data p5 is 0, and Ph, P6 to 9 are not all equal to 0, P5 is changed to 1. Edge matrix P' after expansion operation:

P\ Pi P P\ P's P P'l p\ P The matrix data of the edge matrix after operation are 1, 〇, 1, 〇, 1, Bu 0, 0 and 0, respectively. In other words, when the pixel data P5 is equal to the first value, and 昼 201044317

Ubf i um〇i Γ W 30913twf. doc/n Data biW ice at least shots - when the shirt is at the first value, the pixel poor P'5 is set to the second value. Repeat the above steps to expand the operational data. Since Ρ5 is equal to 〇, and ρι~ρ4, ¥ & no: is equal to 〇, so Ρ, ι 十9 is and 0. After performing the above expansion operation on the entire screen (arithmetic data), the edge of the object can be sealed to obtain (4) rounds such as generation-edge miscellaneous. It should be noted that, in this embodiment, the first value is 〇, the second value is i, the center 1 indicates that there is a gray scale value, and 0 indicates that there is no gray scale value (turns to the visible surface is black) 'but this The invention is not limited to this. Next, fill in the edge image to get the fill-up book. This step clearly indicates the area and location of the (4) piece. Fig. 6 is a schematic diagram of the step S3°3 in Fig. 3 of the implementation of the child. Please refer to: Figure 6 ‘100 first, the edge image (10) is left to right and from the top 扫 - sweeping 05' where P5 is equal. And two, then P5 is adjusted to repeat the above numerical adjustment mode is the scanning unit's scan from left to right and top to bottom. After =1 edge image 6G1 and the numerical adjustment, you can get a 2 Image 603. Next, changing the scanning direction, scanning the edge image from right to left and from bottom to top with the same number ς, the image of the scanning method is as shown in the first 606, wherein ρ5Λ η η, Bu W When the field is in the way, the T-field P5 4 is in the 〇 and the p6 and p8 are both, then it will be ruined. Then 'the first broom image 6G3 and the second sweep image ^ 9 201044317 0910046ITW 30913twf.d〇c /n Take the parent set, you can fill the face 6Q2, and fill the picture (10) 2 A can clearly see the area where the object is located. Next, step-by-step instructions on how to mark the object, Figure, the present invention - real 。 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 First, mark the grayscale image area, if the grayscale image 701 is a book, ^ If the left side is greedy, then the new number will be marked. i The area of the object (the area consisting of 255 is marked as image 7 (^=, respectively, Η, followed by the mark, adjust the mark image 7〇? If the above data is not 0, the left two marks are edited), then the upper mark is copied: the car side 255, and the upper or left side of the mark; ^ X8). * The 昼素 data is modified according to the above mark When the material ^ 〇 (such as the marked point X9), 5), at the same time modify the left side of the 昼 资料 ^ ^ § § § point X9 value is modified to 5). Next, re-mark all the ', mark X6, The value of the change is changed to the number. After the modification, the mark number (7) which is continually marked as _' will be repaired. 'The mark X7 can clearly find that there are 6 objects, ', 八* this, recorded The above-mentioned marking method is only 1 to 6 of the present invention. It is worth noting that the present invention is not limited to 10 201044317 \jy i\j\j^\jxTW 30913twf .doc/n This Figure 8 illustrates a schematic diagram of two images generated in accordance with the present invention. Please refer to Figure 2, Figure 3 and Figure

t ^ 801 〇 , ;〇;,^S2〇J =r is the second r image' and then the gray image is extracted by the two ^ o 进行 for the face ^,: ^ 乂 咖 咖). Then, the object surface is obtained for the edge of the computational material 802 to produce an edge image (^%02). Next, fill the area corresponding to the inside of each edge image 803 to obtain the 仃 ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' Remove to obtain the depth of two depths of field: In addition, 'notable is the object of the image diagram of the horse 1 whose depth value is divided into two layers, that is, the background and the object. Step _9 is an embodiment according to the present invention Explain the depth map and y image plots that produce multiple depths of field. The eyes are combined with reference to Figure 2, Figure 3, and Figure 5. The two-dimensional image 9〇1 / Zhongxie eve objects will be all after quenching individual objects. The object is carried out, and then the ice value corresponding to the object is determined according to the coordinate position of the individual object. In Fig. 9, the method for generating the data 9 〇 2, the edge image 9 〇 3 and filling the 面 surface 904 and FIG. 8 The similarities in the above will not be described here. Then, 'the areas corresponding to the objects in the filled 904 are marked in sequence' and the objects that are too small are removed, and the filled objects 904 are marked as 1 to 5 (steps S201 to S204). Then, depending on the individual The distance between the object and the lower boundary of the grayscale image (昼面) determines the individual object 11 201044317 09100461TW 30913twf.doc/n piece ^ pearlity map (step S2〇5), the farther the distance, the smaller the depth value (representing the distance Far), the closer the distance is, the larger the depth value (representing the closer distance). If the distance between the two objects is equal to the lower boundary, the depth values are the same. Then, a depth map is generated according to the depth value corresponding to each object ( Step S206), as shown in the depth map 905, which includes 5 objects, and the lowermost objects B1 and B2 have the same distance from the lower boundary, so the depth values are the same. The distances between the objects B3, B4, and B5 and the lower boundary are both Different from each other, they have different depth values. If the depth value of the background is added, the depth map 905 has five depth values, and the value of the respective depth values can be set by the user in advance, for example, The height of the surface corresponds to the maximum value, and the distance between the object and the lower boundary can be calculated according to the ratio of the depth value. It is worth noting that the depth value of the embodiment is according to the object and the lower part. The distance between the boundaries is calculated, but the calculation method is not limited to the above method. Alternatively, the basin depth value may be calculated by using different boundaries, such as the upper boundary of the kneading plane or the left and right boundaries as the reference boundary. The embodiment is not limited. Finally, after obtaining the depth map, the two-dimensional image can be combined to generate a three-dimensional image (step S207). In general, the present invention provides an object and a depth map that can be directly extracted from the two-dimensional image. The method combines the two-dimensional image with the extracted depth map to form a two-dimensional image. Thus, the present invention does not require additional information, and can convert the two-dimensional image into a three-dimensional image without additional image capturing equipment. The present invention has been disclosed in the above embodiments, but it is not intended to limit the invention, and those skilled in the art can make some modifications and refinements without departing from the spirit and scope of the invention. Therefore, this article 12 201044317 uyiuu4〇xFW 30913twf.doc/n The scope of protection of the invention is subject to the definition of the scope of the patent application. [Simple Description of the Drawings] FIG. 1 is a schematic diagram of a three-dimensional image display architecture using a fence. 2 is a flow chart illustrating a method of converting a two-dimensional image into a three-dimensional image in accordance with an embodiment of the invention. FIG. 3 is a flow chart illustrating an embodiment of step S203 in accordance with an embodiment of the present invention. 4 is a schematic diagram illustrating an expansion operation in accordance with an embodiment of the present invention. Figure 5 is a schematic diagram showing the direction of operation of the expansion operation in accordance with one embodiment of the present invention. FIG. 6 is a schematic diagram showing the step S303 of FIG. 3 according to an embodiment of the invention. Figure 7 is a block diagram showing the step S204 of Figure 2 in accordance with an embodiment of the present invention. FIG. 8 is a diagram showing an image of a depth map for generating two depths of field according to an embodiment of the invention. FIG. 9 is a diagram showing an image of a depth map for generating multiple depths of field according to an embodiment of the invention. [Description of main component symbols] 10: Fence 13 201044317 09100461'I'W 30913twf.doc/n 20 : Alizarin 30: Backlights 501, 802, 902: Operational data 601, 803, 903: Edge images 602, 804, 904 Filling the face 603: the first scan image 604: the second scan image 605: the first scan mode 606: the second scan mode 701: the grayscale image 702~704: the mark image 801, 901: two-dimensional Image 805: depth map of two depths of field 905: depth map XI to X9 of multiple depth of field: marker point P: pixel matrix P': edge matrix S201 to S207 after expansion operation: flowchart steps S301 to S303: flowchart step 14

Claims (1)

201044317 uyiuu^oiiW 30913twf.doc/n See the profit range: 1. A type of silk that converts a two-dimensional scene into a three-dimensional image. The method receives a two-dimensional image; converts the two-dimensional image into a grayscale image; extracting an object in the grayscale image; ❹ 标记 marking the object in the grayscale image; and converting between the object of the (4) image towel and the & Determining the depth value of each object; according to each object, touching the red button-shirt; and combining the depth map with the two-dimensional image to generate a three-dimensional image. 2: If you want to use the method described in the Wei Wei Di 丨 item, the steps in the object in the fire P white scene include: · Producing one: image for edge detection (Ε·EStimation) to produce noodles, Filling the area corresponding to the object outline (10) to obtain the method described in item 2 of the second-order profit range, wherein the step of the gray material is used to generate the computing mask. (Mask) to extract the gray age _ object edge, the equation for the edge detection of Jining is as follows: 201044317 uyiUU4t> ut 309l3twf.doc/n W = \Gxx P\ + \Gyx where ▽/ indicates that the operation data 'p is The data of the 3*3 matrix, the prewitt mask of the Gj Gyg 3*3 matrix, the above G and Gy are respectively as follows: 4. The method of claim 2, wherein the step of expanding the operation data to join the edge of the object to obtain the contour of the object comprises: 乂 selecting a halogen matrix p= Λ Ps P6 Pi Pi P9 where Pi~P9 indicates that the element data is judged whether the element data I>5 is a first value - P~P when ί ί ί 1 bucket P5 is equal to the first, and the pixel data P1 to P4, P5 are adjusted to - The second value; and the job h _ 昼 贫 贫 骤 骤 该 该 该 该 该 该 该 该 该 该 该 该5. If the value of the value stated in item 4 of the scope of the patent application is 〇, the second value is h macro ', the first number of 5 hai. 6. The method described in claim 2, the object The step of the corresponding area in the contour comprises: scanning the edge image from top to bottom to obtain a first image, wherein # scanned - the first element data ^_|_^ 16 TW 30913twf. Doc/n 201044317 value, the left side of the first 昼 资料 资料 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联 联Down to the top of the sweep, the edge is the second value; and the image, when the scanned second image is taken to obtain the second value of the right side of the second sweep data, the first value is The second number 'the second 昼 资料 i i 均 均 & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & The method of claim 1, wherein the object in the (four) square white shirt is marked by the method described in claim 1 Including: /, the mark is different according to the _ field corresponding to each piece of the gray-scale image towel: 标记 The mark corresponding to the object, so that each object corresponds to the 彳 彳 of the j is arranged in a sequence. 17