JP4415699B2 - Seamless texture composition method, image composition program, and image composition system - Google Patents

Description

  The present invention relates to an image synthesizing apparatus for synthesizing an image to be printed on a printed material for building materials used for interiors such as buildings, and in particular, a seamless texture synthesizing method and an image synthesizing program for synthesizing images in which the joints of the images are not noticeable. The present invention relates to an image composition system.

  Conventionally, when a building material such as a floor, a wall, or a ceiling is used as an interior, a pattern such as an abstract pattern is printed on the surface, or the pattern is printed and pasted on a decorative sheet. Here, the used image printed on the interior product is a combination of a plurality of images, a single large image is combined, and the combined image is printed. At this time, since the image to be printed is used for the interior, it is generally desirable that the joint of the image is inconspicuous and that the handle is inconspicuous in terms of a natural finish and a good appearance.

In order to obtain such an abstract pattern, in a conventional synthesis method, when superimposing images to obtain a composite image, for example, a method of synthesizing so that joints are not conspicuous when superimposing images (for example, Patent Documents 1, 2, and 3) are provided.
JP 2003-317089 A JP 2003-317090 A JP 2003-319165 A

  However, although the image synthesizing methods disclosed in Patent Documents 1, 2, and 3 disclose a synthesizing method that stitches two images together, a single output image is stitched continuously in order to obtain an output image. The case is not specifically disclosed. Therefore, when output images are connected, a periodic texture pattern (pattern) may appear at the boundary portion.

The present invention has been made in view of such circumstances, and an object of the present invention is to provide a seamless texture that can prevent a joint between output images from being noticeable even when a plurality of output images are continuously output. A synthesis method, an image synthesis program, and an image synthesis system.
It is another object of the present invention to provide a seamless texture composition method, an image composition program, and an image composition system that perform tiling so that joints between patches do not stand out when tiling patches to an output image. .

  In order to solve the above-described problem, the present invention relates to a texture synthesis method in which patches having high similarity are randomly selected from an input image and tiled to generate an output image, and the patches are tiled from the outer periphery of the output image. It is characterized by doing.

  Further, the present invention provides a texture synthesis method in which a patch having a high similarity is randomly selected from an input image and tiled to generate an output image, and a feathering method and an image quilting method from the outer periphery to the inside of the output image, or The patch is tiled by a feathering method and a graph cut method.

  Further, the present invention relates to a texture synthesis method in which patches having a high similarity are randomly selected from an input image and tiled from the outer periphery to generate an output image. These patches are tiled by a feathering method and an image quilting method, or a feathering method and a graph cut method.

  Further, the present invention provides a texture synthesis method in which patches having a high similarity are randomly selected from an input image and tiled from an outer peripheral portion to generate an output image. The outermost peripheral patch is pasted by an image quilting method, These patches are tiled by a feathering method and an image quilting method, or a feathering method and a graph cut method.

  The present invention is also an image composition program used in an image composition system for selecting and tiling patches with high similarity from an input image and generating an output image, wherein the patch is placed on an outer periphery of the output image. A first tiling step for tiling, and a second tiling step for tiling the patch inside the output image after the outer periphery of the output image has been tiled by the first tiling step. And making the computer execute.

  Further, the present invention provides a first tie that tiles the patch on the outer periphery of the output image in an image composition system that generates an output image by randomly selecting a patch having high similarity from an input image. A ring portion; and a second tiling portion that tiles the patch inside the output image after the outer peripheral portion of the output image is tiled by the first tiling portion. To do.

  As described above, according to the present invention, patches are tiled from the outer periphery of an output image when a patch having a high similarity is randomly selected from the input image and tiled to generate an output image. Therefore, a patch having a high degree of similarity can be tiled on the outer peripheral portion, thereby improving the seamlessness around the tile (top and bottom).

  Further, according to the present invention, since the patch is tiled by the feathering method and the image quilting method, or the feathering method and the graph cut method from the outer peripheral portion to the inner portion of the output image, the position of the patch tiling An image synthesis method according to the above can be applied, and this makes it possible to make the joints between patches more inconspicuous.

  Also, according to the present invention, the outermost patch is attached by the feathering method, and the internal patch is tiled by the feathering method and the image quilting method or the feathering method and the graph cut method. The seamlessness can be further improved and the speed can be increased.

Hereinafter, an image composition system according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a schematic block diagram showing a configuration of an image composition system 1 according to an embodiment of the present invention.
In this figure, the image synthesizing system 1 selects and extracts a patch having a high similarity at random from the input image input from the image input unit 10 by the patch extraction unit 20, and extracts the texture from the outer periphery of the texture by the tiling unit 30. While the tiling is performed, the composition processing unit 40 sequentially synthesizes a portion where the tiled patch and the image already tiling on the texture overlap, and the texture output unit 50 outputs the texture. Hereinafter, each part of the image composition system 1 will be described.

  The image input unit 10 inputs a sample input image. This input image is, for example, an image as shown in FIG. 2A, and is a basic image for generating a texture output from the texture output unit 50 described later.

  The patch extraction unit 20 selects, from the input images input from the image input unit 10, a patch that is an image having a size as a unit of an image to be pasted on the texture, and outputs the selected patch to the tiling unit 30. . Here, for example, the patch is extracted at a predetermined size from the input image as shown by reference symbol a in FIG. 2A and pasted to the output image as shown in FIG. 2B. Is synthesized as a texture.

  In addition, when selecting a patch, the patch extraction unit 20 compares the pixel values of each pixel in the boundary area where the image already pasted as the output image and the next patch to be pasted overlap, and the degree of similarity A patch with a high value is selected from the input image. FIG. 3 is a diagram illustrating an example of a boundary region. As shown in this figure, when the patch B1 in FIG. 3A is pasted on the output image, the boundary area has the width of the boundary area between one end of the patch B1 and the other end of the patch B2 to be pasted. This is a region where an arbitrary number of pixels shown is overlaid (FIG. 3B).

ここでは、Ｎは境界領域の画素数である。ｐ_Ｂ１は重ね合わせる一方の画像の画素の画素値、ｐ_Ｂ２は重ね合わせる他方の画像のうち重ね合わせる一方の画像の画素に対応する画素の画素値である。ｉは、境界領域内における距離を示す座標を示す値である。具体的には、図４に示すように、パッチＢ１とパッチＢ２において、境界領域となる部分の符号ａで示す画素と符号ｂで示す画素が重ねられるので、符号ａで示す画素の画素値がｐ_Ｂ１、符号ｂで示す画素の画素値がｐ_Ｂ２として計算される。この類似度の計算は、境界領域内の画素毎に行われる。 The similarity referred to here indicates the degree of similarity between pixels corresponding to a patch, and is calculated by the following equation (1).

Here, N is the number of pixels in the boundary area. p _B1 is a pixel value of a pixel of one image to be superimposed, and p _B2 is a pixel value of a pixel corresponding to a pixel of one image of the other image to be superimposed. i is a value indicating a coordinate indicating a distance in the boundary region. Specifically, as shown in FIG. 4, in the patch B1 and the patch B2, the pixel indicated by the reference symbol a and the pixel indicated by the reference symbol b are overlapped, so that the pixel value of the pixel indicated by the reference symbol a is The pixel value of the pixel indicated by p _B1 and symbol b is calculated as p _B2 . This calculation of similarity is performed for each pixel in the boundary region.

ここで、ｐ_ｏｕｔは、パッチＢ１とパッチＢ２の対応する画素における画素値の平均値であり、εは、０から１の任意の値である。そして、類似度ｄが閾値ｄ_ｍａｘ以下である場合に類似度が高いパッチとして選択される。この選択は、例えば、入力画像の中からランダムにパッチのサイズでパッチの候補として抽出し、類似度が高いか否かを参照し、類似度が高い場合にその候補のパッチを選択し、類似度が低い場合に再度入力画像の中から候補となるパッチを選択する。なお、類似度が閾値以下のパッチがない場合は、最も類似したパッチが選択される。また、ここでは、εの値を任意に変更することにより、類似度を度合いを変更することが可能である。 Whether or not the similarity is high is determined based on a threshold value calculated by the following equation (2).

Here, p _out is an average value of pixel values in corresponding pixels of the patch B1 and the patch B2, and ε is an arbitrary value from 0 to 1. When the similarity d is equal to or less than the threshold value d _max , the patch is selected as a patch having a high similarity. For this selection, for example, a patch candidate is randomly extracted from the input image as a patch candidate, referred to whether the similarity is high, and if the similarity is high, the candidate patch is selected and similar When the degree is low, candidate patches are selected again from the input image. If there is no patch whose similarity is equal to or less than the threshold value, the most similar patch is selected. Here, the degree of similarity can be changed by arbitrarily changing the value of ε.

  The composition processing unit 40 includes a feathering processing unit 41, a graph cut processing unit 42, and an image quilting processing unit 43. In accordance with image composition instructions from the outer peripheral tiling unit 31 and the internal tiling unit 32, A process for synthesizing the image in the boundary region is performed.

The feathering processing unit 41 performs image composition by the feathering method. That is, the feathering processing unit 41 calculates a pixel value by assigning a weight proportional to the distance of each patch to each pixel in the boundary region in the pixel where one patch overlaps the other patch. Here, as shown in FIG. 5, the feathering processing unit 41 sets the number of pixels P after image synthesis in the patch B1 and the patch B2 as P = w × P _B1 + (1−w) × P _B2
It calculates based on the following formula. Here, w is a value determined based on one patch,
w = (number of pixels indicating the distance from one edge) / (number of pixels having the width of the border area)
As a result, for each pixel where one patch and the other patch overlap, the pixel is calculated in consideration of both pixel values according to the distance from the edge of the patch, and is proportional to the distance from the edge. The boundary areas can be mixed together by applying the weights, so that the patches can be smoothly joined together.

The graph cut processing unit 42 performs image composition by the graph cut method. That is, the graph cut processing unit 42 calculates the difference between two adjacent pixels that are adjacent pixels when the one patch and the other patch are overlapped with each other. A cut line is generated by calculating and selecting the minimum pixel for each column. Specifically, this will be described with reference to FIG. In FIG. 6, when the patch A and the patch B are overlapped with the pixels 1 to 9 as the boundary region, the cost M of the patch A and the patch B for the pixel 4 adjacent to the pixel 1 is 3) Calculate based on the equation.

  Here, s is a pixel value in one of the adjacent pixels, and t is a pixel value in the other of the adjacent pixels. Therefore, the cost M between the pixel 1 and the pixel 4 in FIG. 6 is the difference between the pixel value of the patch A and the pixel value of the patch B at the position of the pixel 1, and the pixel value and the patch of the patch A at the position of the pixel 4 Calculated by the sum of the pixel value difference in B. The boundary line passing through the route with the least cost M is set as a cut line, the patch is cut, and the patches A and B are connected. As described above, since the cost for two adjacent pixels n is obtained, it is possible to separate the patch at a portion where the pixel values are close to each other between patches. Can be improved.

  The image quilting processing unit 43 performs image composition by an image quilting method. That is, the image quilting processing unit 43 detects a cut line with the smallest difference between the two upper and lower patch images to be superimposed, and stitches (joins) the patches along the detected cut line. Here, when selecting the lower pixel row, the smallest pixel is selected from the three pixels at the lower position of the previously selected pixel.

FIG. 7 is a diagram for explaining the processing of the image quilting processing unit 43. In this figure, coordinate values e _{i, j} correspond to each pixel in the boundary region, and the boundary region that overlaps among the lower three pixels from each e _{0, j} (j is 0 to 5 in FIG. 7). Are selected with the smallest difference (error) in pixel values, and the pixel with the smallest error is selected from the lower three pixels based on the selected pixel. This process is sequentially repeated until i becomes 0 (from 6 to 0 in FIG. 7), and the accumulated error is calculated and stored (the line passing through the valley of the error curved surface becomes a cut line). Then, the route with the smallest accumulated error is determined as the cut line. Here, the route with the smallest error is determined by sequentially selecting one of the coordinates calculated by the following equation (4).
E _{i, j} = e _{i, j} + min (E _{i−1, j−1} , E _{i−1, j} , E _{i−1, j + 1} ) (4)
The error is calculated by the following equation (5) for each pixel calculated by the above (4) to form an error curved surface.
e = (P1-P2) ² (5)

  The tiling unit 30 tiles (pastes) the patch selected by the patch extraction unit 20 on the outer peripheral part in the output image by the outer peripheral tiling part 31 (corresponding to the “first tiling part” described above). Then, after the tiling of the outer peripheral portion is completed, the inner tiling unit 32 performs tiling inside the output image. Here, the outer peripheral portion is a portion where a patch can be pasted at a position in contact with the edge portion of the output image. The inside is a remaining portion after the patch is attached to the outer peripheral portion of the output image, and is a portion where the patch can be attached at a position not in contact with the edge portion.

  Here, tiling will be further described. FIG. 8 is a diagram for explaining an example of a tiling order. This figure shows a case where an output image is generated by pasting patches in a 4 × 4 matrix. FIG. 8A shows a case where the outer peripheral portion is tiled sequentially from one patch pasted on the output image (symbol (a)) and tiled inside. In this case, the tiles are tiled in a spiral shape from the outside to the inside in preference to the patches at the top and bottom boundaries.

  FIG. 8B shows an outer peripheral portion of the edge in the other direction of the reference patch after tiling the outer peripheral portion with respect to two edges with reference to one patch pasted on the output image as indicated by reference symbol a. This is a case where tiling is performed on the outer peripheral portion of the remaining edge indicated by reference numeral b. FIG. 8C shows a case where the outer peripheral portion is tiled for each direction of the edge with respect to one patch pasted on the output image, then the remaining outer peripheral portion is tiled, and the interior is tiled. . 8A to 8B, the case where the position of the patch to be applied to the first sheet corresponds to any of the four corners of the output image has been described. Good.

  Conventionally, as shown in FIG. 8 (d), tiling has been sequentially performed only in a certain direction for each column indicated by reference numeral (a), for example.

Next, the outer peripheral tiling part 31 and the inner tiling part 32 will be described.
The outer periphery tiling unit 31 is a boundary between an outer periphery tiling function that tiles the patch extracted by the patch extraction unit 20 on the outer periphery of the output image, and a patch that is already tiled in the output image and the patch to be tiled. For the area, there is an outer peripheral image composition function for instructing the composition processing unit 40 to perform image composition processing using one or both of the feathering processing unit 41 and the image image quilting processing unit 43.

  The internal tiling unit 32 has a function of tiling the patch extracted by the patch extraction unit 20 inside the output image after the patch is tiled on the outer peripheral portion of the output image by the outer peripheral tiling unit 31, and the output image. For the boundary region between the patch already tiled and the patch to be tiled, any one of the feathering processing unit 41, the graph cut processing unit 42, and the image image quilting processing unit 43 in the synthesis processing unit 40, Alternatively, it has an internal image composition function for instructing the composition processing unit 40 to perform image composition processing using the two in combination.

Here, an instruction pattern in which the outer periphery tiling unit 31 and the inner tiling unit 32 give an image synthesis instruction to the image processing unit 40 will be described.
Pattern 1: Both the outer tiling unit 31 and the inner tiling unit 32 use the feathering processing unit 41 and the image quilting processing unit 43 together, or the feathering processing unit 41 and the graph cut processing unit 42 Combined.
Pattern 2: The outer periphery tiling unit 31 uses only the feathering processing unit 41, and the internal tiling unit 32 uses the feathering processing unit 41 and the image quilting processing unit 43 together, or the feathering processing unit 41 The graph cut processing unit 42 is used together.
Pattern 3: The outer periphery tiling unit 31 uses only the image quilting processing unit 43, and the internal tiling unit 32 uses the feathering processing unit 41 and the image quilting processing unit 43 together, or the feathering processing unit 41 The graph cut processing unit 42 is used together.

Next, a boundary region when the outer peripheral tiling part 31 and the inner tiling part 32 are tiling will be described.
FIG. 9 is a diagram for explaining a boundary region in a case where tiling is performed spirally from the outer peripheral portion. For example, in the case of tiling in the spiral shape shown in FIG. 8A, when the patch B, which is the second patch, is applied to the patch A, which is the first patch, as indicated by reference numeral a. The boundary area is one place. When the last patch C in the column is pasted, there are boundary areas indicated by the symbols b and c. The boundary area of the symbol c is a boundary area between one end of the patch A indicated by the symbol d and takes into account the periodicity when the textures are pasted together. When a plurality of images are pasted side by side, the joint between textures can be made inconspicuous.

  In the patch in the last column of the outer peripheral portion, for example, patch D has two boundary regions as indicated by symbol d, and patch E which is the last patch of the outer peripheral portion has three boundary regions indicated by reference symbol e. There is an area. In this way, by tiling from the outer periphery, the boundary area of the patch to be pasted last is the largest in four places, but since the patch with many comparison areas is inside, the seamlessness between textures is increased. Can be improved. As described above, with respect to patches that are tiled on the outer peripheral portion, it is possible to generate a texture with an emphasis on being seamless by reducing the number of boundary regions that are subjected to similarity comparison between the left and right patches. The tiling order is set in advance in the tiling unit 30 or a tiling order instructed from an input device such as a mouse or a keyboard is set. The tiling part 31 and the internal tiling part 32 perform tiling.

  On the other hand, in the conventional tiling method as shown in FIG. 8D, as shown in FIG. 10, the patches in the last row need to compare the similarity of three or four boundary regions. For this reason, the conditions for extracting patches become severe, and as a result, it becomes difficult to improve the seamlessness between textures.

  The texture output unit 50 outputs, as a texture, an output image in which all patches are tiled on the output image and image synthesis of the boundary region is completed. This output is output to, for example, another computer or the like, or output by a display device such as a CRT (Cathode Ray Tube) or a liquid crystal display device.

  The above-described image composition system 1 (image input unit 10, patch extraction unit 20, tiling unit 30, outer tiling unit 31, internal tiling unit 32, composition processing unit 40, feathering processing unit 41, graph cut processing unit 42) The image quilting processing unit 43 and the texture output unit 50) may be realized by dedicated hardware, and the image composition system 1 includes a memory and a CPU (central processing unit). The function may be realized by loading a program for realizing the function of each unit in the image composition system 1 into a memory and executing the program.

  In addition, when performing various operations of the image composition system 1, data is stored in a nonvolatile memory such as a flash memory, a volatile memory such as a RAM (Random Access Memory), or a combination of these. Is temporarily stored. In addition, an input device, a display device, and the like (none of which are shown) are connected to the image composition system 1 as peripheral devices. Here, the input device refers to an input device such as a keyboard and a mouse. The display device refers to a CRT (Cathode Ray Tube), a liquid crystal display device, or the like.

  Next, the operation of the image composition system 1 in FIG. 1 will be described using the flowchart of FIG. Here, the outer tiling unit 31 causes the feathering processing unit 41 to perform image composition, and the internal tiling unit 32 performs the image composition by using the feathering processing unit 41 and the image quilting processing unit 43 together. explain.

  First, when an input image serving as a basic image is input from the image input unit 10 (step S1), the patch extraction unit 20 extracts a patch similar to the boundary region from the input image (step S2). Here, since the first patch is pasted and the patch has not yet been pasted on the output image, the patch is extracted at random (step S2).

  Next, the tiling unit 30 tiles the patch extracted by the patch extraction unit 20 to the output image by the tiling unit 31 on the outer periphery. The outer periphery tiling unit 31 detects whether or not the patch has a boundary region by detecting whether or not the patch has already been tiled in the output image (step S4). Here, since the patch to be tiled is the first sheet, the outer peripheral tiling unit 31 proceeds to step S6, and then detects whether all the patches have been tiled on the outer peripheral unit (step S6). Here, since the tiling of the outer peripheral portion has not been completed, the outer peripheral tiling portion 31 instructs the patch extraction portion 20 to extract the patch together with information indicating the position of the next tiling patch in the output image. Then, the process proceeds to step S2.

  Upon receiving an instruction from the outer periphery tiling unit 31, the patch extraction unit 20 detects the boundary region of the patch already pasted on the output image based on the position of the patch to be tiled next, and the detected boundary region A patch similar to the image is extracted from the input image (step S2). Here, one boundary region is detected, a patch having an image edge similar to the one boundary is extracted as a candidate, and the similarity is calculated based on the above-described equation (1). (2) A patch that satisfies the threshold condition based on the expression is extracted as a candidate, and the patch with the highest similarity is determined as the patch to be attached next.

  When the patch is determined, the outer periphery tiling unit 31 tiles the determined patch at the next tiling position of the output image (step S3), and detects whether the patch has a boundary region (step S3). Step S4). Here, since there is one boundary region, the process proceeds to step S5, and the boundary region is synthesized by the feathering processing unit 41 by the feathering method (step S5). And the outer periphery tiling part 31 detects whether all the tiling was performed to the outer peripheral part (step S6). Steps S2 to S6 are repeatedly executed until all the outer peripheral parts are tiled. When all the outer peripheral parts are tiled, the process proceeds to step S7.

  After step S6, the tiling unit 30 instructs the patch extraction unit 20 on the position of the next tiling to be tiling inside based on the set tiling order. Based on this instruction, the patch extraction unit 20 detects a boundary region, and extracts a patch similar to the image of the detected boundary region (step S7). Here, since there are a plurality of boundary regions, a patch that shakes an edge similar to each boundary region image is selected and extracted from the input image.

  When the patch is extracted by the patch extraction unit 20, the internal tiling unit 32 tiles the patch at the next tiling position of the output image (step S8), and performs image synthesis on the boundary region (step S8). S9). Here, the internal tiling unit 32 causes the feathering processing unit 41 or the image quilting processing unit 43 to perform image synthesis within the boundary region. As for whether the feathering processing unit 41 or the image quilting processing unit 43 performs image synthesis, a method in which joints are not conspicuous in the boundary region is applied. For this application, the image is synthesized by one of the methods, and the image is synthesized by the other image synthesizing method based on an instruction inputted from the input device to synthesize the joint by a different method. May be.

  Next, when image composition for the boundary region is performed, the internal tiling unit 32 detects whether or not all the insides of the output image have been tiled. The process proceeds to S7. When the patches are tiled on all of the output images by sequentially repeating from step S7 to step S10, the texture output unit 50 outputs the texture to the display device (step S11).

  Next, the texture output by the texture output unit 50 will be described. FIG. 12 is a diagram illustrating an example of overlapping of patches when patches are pasted on an output image. As shown in this figure, when the patch size is 2.5 in length and 2.5 in width and is tiled in a spiral order, when the width of the boundary region is 0.5, The size of the output image is 8.5 vertical and 8.5 horizontal (FIG. 12A). Then, among the output images, as shown in FIG. 12B, for the edge that does not have the boundary area in the patch A attached to the first sheet, the patch B corresponding to the boundary area corresponding to the edge. And paste it with the left and right reversed. Then, two columns (columns a and b) of the boundary region on the edge side of the patch B are deleted. Here, since the boundary conditions are considered in advance as a periodic condition and are combined as a boundary region, when the opposing edges such as the edge c and the edge d are bonded together, they can be seamlessly bonded. Similarly, the upper and lower edges can be seamlessly bonded together. Then, the texture after being deleted for two columns is output. By arranging a plurality of textures and pasting them together, a seamless image can be obtained without repeating the pattern based on the input image.

  Furthermore, a seamless image can be printed by forming an image with this texture as one unit and forming a printing plate on a cylinder as a printing plate, attaching this cylinder to a printing device, and printing it on a printing medium. Printed products can be produced. Here, since the edges of the texture are subjected to the image composition processing at the edges of the texture, the seamlessness is good even when formed on the circumferential surface of a cylinder like a drum. It is desirable to set the texture so that it matches the circumference of the cylinder when the unit size of the texture is multiplied by an integer. Here, when a decorative sheet is used as a printing medium, a decorative sheet on which a texture image is printed is obtained, and can be used as an interior such as a ceiling, a wall, or a floor.

  In the above-described embodiment, the generated texture may be corrected based on an instruction from the input device. For example, as shown in FIG. 13, the patch extraction unit 20 sets a boundary region based on the correction range based on the correction range in which a part of the texture is input using a mouse or the like, and the boundary region is set in the boundary region. Similar patches are extracted from the input image. Then, the graph cut processing unit 42 performs a graph cut process on the extracted patch and pastes it on the correction range, so that it is possible to partially correct an arbitrary image range of the user, considering the user's request. A texture can be generated.

In the above-described embodiment, a case has been described in which tiling patches that are tiled on the outermost peripheral portion of the output image are joined by a feathering method, but the patch that is tiled on the outermost peripheral portion is imaged. You may make it tiling by connecting by a quilting method.
In the above-described embodiment, the case where the patches tiled on the outermost peripheral portion of the output image are connected by the feathering method and tiled is described. However, the patches tiled on the outermost peripheral portion are connected together. In this case, not only the feathering method but also one of the feathering method and the image quilting method is selected and connected each time one patch is connected, or one of the feathering method and the graph cut method is selected. And may be joined together. That is, each time one patch is connected from the outermost part to the inside, either the feathering method or the image quilting method is selected and connected, or the feathering method or the graph cut method is selected. And may be joined together.

  In the above-described embodiment, the case where the feathering method and the image quilting method are used together when tiling the four internal patches has been described. However, any one of the internal patches or the last tiling is performed. Only the innermost patch may be used in combination with the feathering method and the image quilting method, or may be used in combination with the feathering method and the graph cut method.

In the above-described embodiment, the case where the patch shape is a rectangle (square) has been described, but a shape other than a square may be used. For example, it may be any one of a regular triangle, a regular square, a regular hexagon, and a rhombus, or may be a shape constituted by a curve.
Further, the shape of the patch may be the same topologically and may be a shape in which a part of the patch overlaps a part of another patch. Thereby, it can arrange | position so that a crossword puzzle may be accumulated and stuck, for example.

  In addition, the image input unit 10, the patch extraction unit 20, the tiling unit 30, the outer periphery tiling unit 31, the internal tiling unit 32, the synthesis processing unit 40, the feathering processing unit 41, the graph cut processing unit 42, the image in FIG. 1. A program for realizing the functions of the quilting processing unit 43 and the texture output unit 50 is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read into a computer system and executed to synthesize an image. Processing may be performed. The “computer system” here includes an OS and hardware such as peripheral devices.

Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.
The “computer-readable recording medium” refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, and a CD-ROM, and a storage device such as a hard disk built in the computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system serving as a server or a client in that case, and a program that holds a program for a certain period of time are also included. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes design and the like within a scope not departing from the gist of the present invention.

1 is a schematic block diagram showing a configuration of an image composition system 1 according to an embodiment of the present invention. It is drawing which shows an example of an input image and a texture. It is drawing which shows an example of a boundary region. It is drawing for demonstrating similarity. It is a figure for demonstrating the image-synthesis process which the feathering process part 41 performs. It is a figure for demonstrating the image-synthesis process which the graph cut process part 42 performs. 6 is a diagram for explaining image composition processing performed by an image quilting processing unit 43; It is drawing for demonstrating an example of the order which performs tiling. It is drawing for demonstrating the boundary area | region in the case of tiling helically from an outer peripheral part. It is a figure for demonstrating the conventional tiling method. 3 is a flowchart for explaining the operation of the image composition system 1 in FIG. 1. It is drawing which shows an example of the overlap of patches when a patch is affixed on an output image. It is drawing for demonstrating the case where a texture is corrected based on the instruction | indication from an input device.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Image composition system 10 Image input part 20 Patch extraction part 30 Tiling part 31 Outer peripheral tiling part 32 Internal tiling part 40 Composition processing part 41 Feathering processing part 42 Graph cut processing part 43 Image quilting processing part 50 Texture output part

Claims (6)

In the texture synthesis method that selects and tiles patches with high similarity randomly from the input image to generate the output image,
A seamless texture synthesis method characterized by tiling patches from the outer periphery of an output image. In the texture synthesis method that selects and tiles patches with high similarity randomly from the input image to generate the output image,
A method for synthesizing a seamless texture, characterized in that patches are tiled by a feathering method and an image quilting method, or a feathering method and a graph cut method, from the outer periphery to the inside of an output image. In the texture synthesis method for generating an output image by randomly selecting patches with high similarity from the input image and tiling from the outer periphery,
A method of synthesizing a seamless texture characterized in that the outermost patch is pasted by a feathering method, and the internal patch is tiled by a feathering method and an image quilting method, or a feathering method and a graph cut method. In the texture synthesis method for generating an output image by randomly selecting patches with high similarity from the input image and tiling from the outer periphery,
A method of synthesizing a seamless texture, characterized in that an outermost patch is pasted by an image quilting method, and an internal patch is tiled by a feathering method and an image quilting method, or a feathering method and a graph cut method. An image composition program used in an image composition system that randomly selects a patch with high similarity from an input image and generates an output image by tiling,
A first tiling step of tiling the patch on the outer periphery of the output image;
A second tiling step of tiling the patch inside the output image after the outer periphery of the output image has been tiled by the first tiling step;
Composition program for causing a computer to execute. In an image composition system that randomly selects a patch with high similarity from an input image and generates an output image by tiling,
A first tiling unit for tiling the patch on the outer periphery of the output image;
A second tiling unit that tiles the patch inside the output image after an outer periphery of the output image is tiled by the first tiling unit;
An image synthesis system comprising:

Images