JP2017010564A - Device, method, and program for digital image resampling - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device, a method, and a program for digital image resampling that realize high-quality resampling for conducting flexible deformations including reduction in digital image processing, with a limited processing frequency and a limited storage capacity, without requiring a quality parameter.SOLUTION: A storage map is prepared, which has the same series as a deformation destination image has, the deformation destination image is scanned, and pixels of the storage map corresponding to the positions of pixels being scanned are updated based on the pixels being scanned. The deformation destination image is thereafter scanned and the pixels of the deformation destination image are updated on the basis of the pixels of the storage map corresponding to the positions of the pixels being scanned.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a digital image resampling apparatus, method, and program.

  Conventionally, in the resampling of a digital image, it is important to apply high-quality deformation to the deformation-target image as effectively as possible.

  In order to obtain a high-quality deformation result in image deformation, various resampling techniques have been conventionally employed. For example, in the enlargement method, there are bilinear interpolation, bicubic interpolation, and the like as methods for suppressing jaggy. In the reduction method, methods for suppressing aliasing include an area average method, a prefilter, a post filter, a trilinear filter, an anisotropic filter, and the like.

  Hereinafter, each of these methods will be described in detail.

  First, bilinear interpolation is an interpolation method that uses a weight according to a position by referring to a total of four pixels, that is, a pixel in a pre-deformation image and a neighboring pixel in order to determine each pixel to be deformed. Although the processing speed is inferior to the nearest neighbor method, it is faster than the average pixel method. Moreover, shaggy can be suppressed more than the nearest neighbor method.

  On the other hand, bicubic interpolation is an interpolation method similar to bilinear interpolation, but a total of 16 pixels of the deformation source pixel and the neighboring pixels are referred to for each deformation destination pixel. In other words, it is an advanced form of bilinear interpolation, and natural and smooth image quality can be obtained compared to bilinear interpolation.

  The deformation using bilinear interpolation and bicubic interpolation basically scans the deformation destination image without separation, so that not only enlargement and reduction, but also two-dimensional or more image deformation represented by projective transformation, rotation, and polar coordinate transformation. Also used for. However, when a method of scanning the deformation source image instead of scanning the deformation destination image is adopted, it is a conventional concern that not all the deformation destination images can be generated and some pixels may be missing.

  On the other hand, the area average method is a reduction method that uses an average of pixels of a pre-deformation image corresponding to the pixel in order to determine a pixel to be modified. This is also called an average pixel method, and is generally a method used exclusively for reduction.

  The pre-filter is a method of performing smoothing according to a reduction ratio before reducing an image.

  The post-filter is also called a supersampling method, and in order to determine the pixel to be transformed, the pixel to be transformed is further divided into a plurality of pixels, each of which is referred to the pixel group of the pre-transformation image, and the average is obtained. Is a reduction method using.

  The trilinear filter is mainly used for drawing a polygon in a three-dimensional graphic, and prepares a mipmap that has a reduced image using an average of pixels in a stepwise manner according to the reduction rate of the image. This is a method of applying and combining a single or a plurality of reduced images. The trilinear filter can correct common deterioration seen in a bilinear filter image using mip mapping.

  The anisotropic filter is mainly used for drawing polygons in three-dimensional graphics, and applies a mipmap using a reference method that takes into account the deformation of the texture in drawing polygons that are non-perpendicular to the camera. It is a method of synthesis. This can be said to be an improvement over the bilinear interpolation and trilinear filter described above.

  Here, in Patent Document 1, an input image having M pixels is acquired by photographing, and the input image is treated as a resampling target image and resampled, whereby each first reduced image having M / 4 pixels. And a second reduced image are generated, and the resampling is executed such that a sampling position shift occurs between the first and second reduced images.

JP 2010-251882 A

  In order to perform high-quality resampling for flexible deformation including reduction, it is necessary to have means for appropriately referencing each pixel of the transformation destination image from a single pixel or a plurality of pixels of the transformation source image. . Here, the flexible deformation is an image deformation of two or more dimensions capable of reverse coordinate transformation, represented by projective transformation, rotation, and polar coordinate transformation. However, the conventional deformation method cannot sufficiently realize high-quality resampling for performing flexible deformation including reduction in digital image processing with a limited number of processes and storage capacity without requiring quality parameters.

  The conventional method of averaging pixel groups in reduction has a problem that the center of gravity of an object in an image is shifted even if the pixel region is taken into consideration. For example, when attention is paid to a plurality of pixels in a certain area before deformation and all the pixels are within one pixel after deformation, the center of gravity cannot be specified. Such a shift in the center of gravity may cause a sense of incongruity in an image that places importance on the positional relationship between thin lines and fine parts (for example, face formation).

  Furthermore, in a two-dimensional or more deformation, enlargement in all directions or reduction in all directions is not always possible. That is, there is a deformation in which one direction is reduction but another direction is enlargement with respect to a certain point of the image. In this case, it is not easy to combine the conventional enlargement method and reduction method.

  In addition, the method of increasing the quality by increasing the number of reference points as in the post filter described above increases the number of pixels to be referred to as the quality is increased, so for users with limited computer resources, The quality is selected as a parameter and a trade-off is forced. That is, the upper limit of quality cannot be determined, and the number of processing times and the storage capacity cannot be limited.

  In view of the above technical problems, the present invention realizes high-quality resampling for performing flexible deformation including reduction in digital image processing with a finite number of processing times and storage capacity without requiring quality parameters. The purpose is to do.

  In order to solve the technical problem described above, the resampling method according to the first aspect of the present invention is a resampling method using an accumulation map having a sequence common to the transformation destination image, and A first step of scanning and updating the accumulation map based on the deformation source image, and scanning the deformation destination image until the extraction of all pixels included in a range to be deformed in the deformation destination image is completed. A second step of extracting the next pixel of the previous image and updating the deformation destination image based on the accumulation map, and in the process of updating the pixel of the deformation destination image based on the accumulation map, It is determined whether the influence value of the pixel information related to the pixel position of the accumulation map is a certain value or more. If the influence value is a certain value or more, the pixel information of the transformation destination image is generated from the accumulation map pixel information, Influence value is above a certain value Without it, and generates the pixel information modification destination image from pixel information modification original image by a predetermined interpolation method.

  The resampling apparatus according to the second aspect of the present invention is a resampling apparatus that uses an accumulation map having a sequence common to the transformation destination image, and is controlled according to display means for displaying various images and a resampling program. And a central control unit that scans the transformation source image, updates the accumulation map based on the transformation source image, scans the transformation destination image, and transforms the transformation destination image. A second step of extracting a next pixel of the deformation destination image and updating the deformation destination image based on the accumulation map until extraction of all pixels included in a range to be deformed in the image is completed. In the process of updating the pixel of the transformation destination image based on the accumulation map, it is determined whether or not the influence value of the pixel information related to the pixel position of the accumulation map is a certain value or more. If the influence value is not less than a certain value, the pixel information of the transformation destination image is generated from the accumulated map pixel information. If the influence value is not not less than the certain value, the transformation destination is transformed from the pixel information of the transformation source image by a predetermined interpolation method. The pixel information of the image is generated.

  A resampling program according to a third aspect of the present invention includes a first step in which a computer scans a deformation source image and updates a storage map having a sequence in common with a deformation destination image based on the deformation source image; , Scans the deformation destination image, extracts the next pixel of the deformation destination image, and updates the deformation destination image based on the accumulation map until extraction of all pixels included in the deformation range in the deformation destination image is completed. In the process of updating the pixel of the deformation destination image based on the accumulation map, the influence of the pixel information on the position of the pixel in the accumulation map It is determined whether or not the value is a certain value or more. If the influence value is a certain value or more, pixel information of the transformation destination image is generated from the accumulation map pixel information. Interpolation And generating the pixel information modification destination image from pixel information modification original image by law.

  According to the digital image resampling apparatus, method, and program according to the present invention, in digital image processing, high-quality resampling for performing flexible deformation including reduction can be performed in a finite manner without requiring quality parameters. It can be realized by the number of times and the storage capacity. Further, in the process of updating the pixel of the deformation destination image based on the accumulation map in the present invention, it is determined whether or not the influence value of the pixel information related to the pixel position of the accumulation map is equal to or greater than a certain value, and the influence value is If it is a certain value or more, pixel information of the transformation destination image is generated from the accumulated map pixel information. If the influence value is not a certain value or more, the pixel information of the transformation destination image is obtained from the pixel information of the transformation source image by a predetermined interpolation method. Therefore, the pixel of the deformation destination image can be updated easily.

It is a block diagram of the resampling apparatus which concerns on the 1st Embodiment of this invention. It is a figure which shows an example of the memory map of the main memory 12 used when performing resampling. It is a conceptual diagram for further explaining in detail a storage map and a coordinate map. It is a flowchart explaining the flow of the process by the resampling apparatus which concerns on the 1st Embodiment of this invention. It is a conceptual diagram explaining in detail the coordinate transformation used for resampling. It is a flowchart explaining the flow of the process which scans a deformation | transformation original image by the resampling apparatus which concerns on the 1st Embodiment of this invention, and updates an accumulation map based on a deformation | transformation original image. It is a conceptual diagram for further explaining in detail the process of scanning the deformation source image and updating the accumulation map based on the deformation source image. It is a conceptual diagram which shows the case where a texture map is lapping (repeat). It is a flowchart explaining the flow of the process which updates the accumulation map pixel group corresponding to the pixel of a texture map by the resampling apparatus which concerns on the 1st Embodiment of this invention. It is a conceptual diagram which further explains in full detail the flow of the process which updates the accumulation map pixel group corresponding to the pixel of a texture map. It is a flowchart explaining the flow of the process which scans a deformation | transformation destination image by the resampling apparatus which concerns on the 1st Embodiment of this invention, and updates a deformation | transformation destination image based on an accumulation map. It is a flowchart explaining the flow of the process which updates a target map pixel based on an accumulation map. It is a conceptual diagram explaining in detail the process which scans a deformation | transformation destination image with the resampling apparatus which concerns on the 1st Embodiment of this invention, and updates a deformation | transformation destination image based on an accumulation map. It is a conceptual diagram which shows the example of the interpolation based on the conventional texture mapping. As a modification, it is a conceptual diagram showing a concept of combining weighted values calculated from conventional interpolation and values calculated from an accumulation map. (A), (b) is a figure which shows an example of a resampling result. It is a flowchart explaining the flow of the process which updates a target map pixel based on the accumulation map by the resampling apparatus which concerns on the 2nd Embodiment of this invention. It is a conceptual diagram for further explaining in detail processing for updating a target map pixel based on an accumulation map. It is a figure which shows an example of the resampling result by 2nd Embodiment.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments according to a digital image resampling apparatus and method and a program (hereinafter, simply referred to as a resampling apparatus, a resampling method, and a resampling program) according to the invention will be described with reference to the drawings. The resampling apparatus, method, and program of the present invention are not limited to the following descriptions, and can be appropriately changed without departing from the gist of the present invention. Details will be described below.

(First embodiment)

  FIG. 1 shows the configuration of a resampling apparatus according to the first embodiment of the present invention.

  As shown in FIG. 1, the resampling device 1 includes a personal computer 10, a display device 20, a deformation parameter input device 30, and the like.

  The personal computer 10 includes a central control unit 11 including a multitasking processor, a main memory 12 including a RAM (Random Access Memory) as a temporary storage device, an image control unit 13 such as a graphic card, and an input / output control unit 14. A built-in nonvolatile storage medium 15 and a media read / write interface 16.

  The image control unit 13 also includes a video memory 13a. Similar to the main memory 12 of the personal computer 10 main body, the video memory 13a is a place for temporarily storing data, and the memory attached to the graphic card is also referred to as VRAM (Video RAM). When a 3D graphic image is displayed on the screen, the amount of data required at that time increases. The data that has been processed by the image controller 13 is stored in the video memory 13a and used as needed. As the capacity of the video memory 13a is increased, even a fine 3D graphic image can be displayed smoothly and without a defect. In recent years, the speed of VRAM has been increased, and a memory standard dedicated to high-speed processing called GDDR has appeared, and a huge data high-speed transfer in 3D graphics drawing has been realized.

  The display device 20 is a device that can display an image, such as a liquid crystal display. The deformation parameter input device 30 is a device that allows a user to arbitrarily specify parameters for deforming an image (for example, an enlargement / reduction ratio and a three-dimensional rotation angle) using deformation method definition data 42 described later. It is a device that can input coordinates and / or buttons such as a mouse, a touch panel, and a pen tablet, and a device that can input keys, such as a keyboard.

  Program data 50, deformation method definition data 42, and transformation source image input data 40 are input via the media read / write interface 16, and transformation destination image output data 41 is output via the media read / write interface 16.

  The program data 50 is software capable of operating the present invention. This corresponds to data of a resampling program to be described later.

  The deformation method definition data 42 is data such as a table or a mathematical expression that associates the coordinates of the deformation source image and the coordinates of the deformation destination image with each other uniquely based on the deformation parameters. When a polygon model is used, the polygon model data corresponds to this and becomes a three-dimensional object. The deformation method definition data 42 may accompany the program data 50 or may read externally defined data.

  The transformation source image input data 40 is an image handled by software. For example, when a polygon model is used, a texture image corresponds to this. The input texture image (transformation original image) and the like are temporarily stored in the main memory 12. The transformation destination image output data 41 is an image handled by software, like the transformation source image input data 40. A deformed image or the like temporarily stored in the main memory 12 is output as a deformation destination image.

  The transformation source image input data 40 and the program data 50 may be read from an external storage medium (not shown), received from an external computer (not shown) via a communication network by a communication control unit (not shown), and the input / output control unit 14 is received. Via the built-in nonvolatile storage medium 15. The output data may be written to an external storage medium (not shown) or transmitted to an external computer (not shown) via a communication network.

  Here, FIG. 2 shows an example of a memory map of the main memory 12 used when resampling is performed.

  As shown in the figure, when resampling is performed, the main memory 12 temporarily stores data of a resampling program, a deformation source image, a coordinate map, a deformation destination image, and an accumulation map.

  Hereinafter, the accumulation map and the coordinate map will be described in more detail with reference to the conceptual diagram of FIG.

  In resampling, the deformation source image indicates an input side image, and the deformation destination image indicates an output side image. The transformation source image and the transformation destination image have pixel information (a: alpha, r: red, g: green, b: blue) including opacity and color as a two-dimensional array. “Alpha” indicates opacity information. As an example, a personal computer records color information of one pixel in units of 24 bits (8 bits per color, 8 × 3 = 24 bits for three colors of red, green, and blue). Information values such as red, green, and blue are also referred to as “density values”, and the 8-bit type can record 256 levels. In the PNG with alpha (32-bit PNG) format, in addition to color information, the opacity of each pixel can be recorded in 256 steps of 8-bit type. An alpha value of zero means completely transparent, 255 means completely opaque. Note that, as an example of an application in which color information is handled with high accuracy in image processing, there is a case where a format in which the density value is expanded to 65536 levels of 16-bit type is used.

  The accumulation map is a two-dimensional array of pixel information (w: influence value, a: alpha, r: red, g: green, b: blue) including the influence value, opacity, and color, and the transformation destination. It is assumed that the image has a common series with the image, that is, the output side coordinate system. Although the details will be described later, the “influence value” is used to calculate the average for determining the color information of the pixel of the transformation destination image corresponding to the pixel of the accumulation map by counting the number of pixels or sequentially adding the pixel weight value. Indicates information to use.

  Since each value in the pixel information of the accumulation map is used for sequentially adding a plurality of pieces of pixel information from the deformation source image, the type of each value is a type that can handle a value sufficiently larger than the pixel information of the deformation source image. (Especially, when the weight value is a real number, a floating-point number type) is desirable.

  As an extreme example, the density value in the accumulation map is “the maximum number of density values of the transformation source image (eg, 255)” in “the number of pixels of the transformation source image (eg, 1000 × 1000)”, “the alpha value of the transformation source image. It is possible to handle a value (1000 × 1000 × 255 × 255 × 1.0) multiplied by “maximum of (example: 255)” and “maximum of weight value (example: 1.0)”.

  The coordinate map has coordinate information (x: horizontal coordinates, y: vertical coordinates) as a two-dimensional array, and has a common series with the series of deformation source images, that is, a coordinate system on the input side. “Coordinate information” is used to record the converted coordinates, that is, the coordinates on the output side, when performing coordinate conversion from the deformation source image to the deformation destination image. The type of each value of the coordinate information is preferably a type (for example, a floating point number type) according to the accuracy of the coordinates to be recorded. The coordinate map does not contribute to the quality of resampling, but is used for speeding up resampling according to the second embodiment of the present invention.

  Hereinafter, the flow of processing by the resampling apparatus according to the first embodiment of the present invention will be described with reference to the flowchart of FIG. This process corresponds to the resampling method according to the first embodiment. This also corresponds to the processing flow when the central control unit 11 executes the resampling program in the main memory 12.

  When the processing is started, first, the transformation source image is scanned, and the processing for updating the accumulation map is executed based on the transformation source image (step S1). Subsequently, the transformation destination image is scanned, and the transformation destination image is based on the accumulation map. Is executed (step S2), and the process is terminated.

  As described above, the resampling feature of the resampling apparatus according to the embodiment of the present invention includes not only scanning of the deformation destination image but also scanning of the deformation source image. Details of each process will be described later.

  Hereinafter, coordinate transformation used for resampling will be described in detail with reference to FIG.

Coordinate forward conversion (input → output) and reverse conversion (output → input) are defined by the following equations.
Coordinate forward conversion (input → output): Pos_OUT = 11 / (9-Pos_IN)-1.5
Inverse transformation of coordinates (output → input): Pos_IN = 9-11 / (1.5 + Pos_OUT)

  FIG. 5 illustrates a one-dimensional deformation in order to easily illustrate the coordinate transformation used for resampling according to the embodiment of the present invention.

  In the figure, the arrangement structure of input data is referred to as a transformation source map. This corresponds to the deformation source image in two dimensions. The array structure of the output data is referred to as a deformation destination map. This corresponds to a deformation destination image in two dimensions.

  A circle represents each element of the transformation source map, and here, as an example, assume a one-dimensional array of 9 elements (0 to 8). The double circle represents each element of the deformation destination map, and here is a one-dimensional array of 11 elements (0 to 10) as an example. The ellipse represents each element of the accumulation map. The double circle and ellipse contact “·” indicates that the accumulation map has a sequence that is common with the sequence of the deformation destination map, and is a one-dimensional array of 11 elements as an example.

  The horizontal line in contact with the circle represents an integer coordinate system connected to the position of the element of the transformation source map. The horizontal line below the circle represents the real coordinate system for the deformation source map. The horizontal line above the ellipse represents the real coordinate system for the accumulation map. The horizontal line in contact with the ellipse represents an integer coordinate system connected to the positions of the elements of the accumulation map.

  Regarding the polygonal line connecting the circle and the ellipse, the solid line represents an example of transformation that derives the position of the accumulation map corresponding to the element position of the transformation source map or the element of the transformation destination map, that is, forward transformation, and the broken line represents the accumulation map. Alternatively, an example of transformation for deriving the position of the element of the transformation source map corresponding to the position of the element of the transformation destination map, that is, inverse transformation is shown. The coordinates of the transformation source map and the transformation destination map are defined by mathematical expressions so as to be mutually and uniquely associated.

  In the example of FIG. 5, when only the position of the element of the transformation source map corresponding to the position of the element of the transformation destination map is viewed, there are positions (0, 1, 3, 4) that are not associated with each other as a result. ing. This is a portion corresponding to reduction in image deformation, and is a cause of aliasing due to leakage of reference.

  On the other hand, when the position of the element of the deformation destination map corresponding to the position of the element of the deformation source map is viewed in an overlapping manner, the part that has not been associated can be compensated. As a method for realizing this, the present invention uses means for scanning the deformation source map without any gap.

  As will be described later, the resampling in the present invention can be performed even in an array of two or more dimensions, and can be used in the field of image processing.

  Hereinafter, with reference to the flowchart of FIG. 6, a flow of processing of scanning the deformation source image and updating the accumulation map based on the deformation source image by the resampling apparatus according to the first embodiment of the present invention will be described. . This corresponds to the details of the processing in step S1 in FIG. This corresponds to a part of the resampling method, and also corresponds to a part of the processing when the central control unit 11 executes the resampling program of the main memory 12.

  The process of this flowchart is an example of texture mapping using a polygon model, and scans a deformation source image and performs forward conversion of positions at each pixel. The texture map corresponds to the deformation source image. At the time of return, the accumulation map is completed and used for the process of updating the later transformation destination image.

  That is, when this process is started, the accumulation map is initialized (step S11), and it is determined whether or not the extraction of all the polygon surfaces is completed (step S12). If the extraction of all the polygon surfaces has not been completed (step S12 branches to No), the next polygon surface is extracted (step S13), and the extraction of all the pixels included in the polygon surface in the texture map is completed. Whether or not (step S14). If extraction of all pixels has not been completed (step S14 branches to No), the next pixel of the texture map is extracted (step S15), and the accumulated map pixel group corresponding to the texture map pixel is updated. (Step S16), and the process returns to step S14. Then, the processes in steps S14 to S16 are repeated until the extraction of all the pixels included in the polygon surface in the texture map is completed. When the extraction is completed (step S14 branches to Yes), the process returns to step S12. In this way, the processes in steps S12 to S16 are repeated until the extraction of all the polygon surfaces is completed. When the extraction is completed (step S12 branches to Yes), the process returns.

  However, the flowchart of FIG. 6 is an example assuming a polygon model (that is, a plurality of deformation methods) composed of a plurality of polygons. When there is a single polygon, or in the deformation method definition data 42, a single table or formula is used. When S is used, steps S12 to S13 can be omitted. That is, when this process is started, the processes of steps S14 to S16 may be repeated, and when extraction is completed (step S14 branches to Yes), the process may return.

  It should be noted that “texture map” is synonymous with “transformation original image”, and “polygon surface” is synonymous with “deformation range” in the deformation original image.

  Hereinafter, the process of scanning the deformation source image and updating the accumulation map based on the deformation source image will be described in more detail with reference to the conceptual diagram of FIG.

  Here, an example of texture mapping using a polygon model is shown as an image deformation capable of inverse transformation of two-dimensional coordinates in the present invention. This is an example of drawing a rectangular model in which an arbitrary posture and position are set in a three-dimensional space, that is, equivalent to deformation including projective transformation and rotation. The rectangular model is regarded as two triangular polygons, and polygon P and polygon Q are assumed to be respectively. The upper right of the figure represents the texture map of the polygon model, that is, it corresponds to the deformation source image. The upper left of the figure represents an accumulation map, which is a coordinate system common to a target map corresponding to a deformation destination image described later.

  It is assumed that the pixels of the texture map have pixel information (a_src: alpha, r_src: red, g_src: green, b_src: blue) including opacity and color. Assume that the pixels of the accumulation map have pixel information (w_sum: influence value, a_sum: alpha, r_sum: red, g_sum: green, b_sum: blue) including the influence value, opacity, and color. In each of the texture map and the accumulation map, the pixels are two-dimensionally arranged, and the unit of coordinates (horizontal direction and vertical direction) is 1 pixel, and for convenience, the boundaries of the pixel areas are represented by dashed grids.

  First, the accumulation map is initialized, that is, the influence value and the color information are cleared to 0 in all the pixels of the accumulation map. Next, the area surrounded by each polygon is scanned without being separated on the texture map side. Here, the position of the pixel to be scanned is (i, j), i is the horizontal position, and j is the vertical position. Scan target pixels related to the polygon P in the texture map are represented by black triangle marks, and scan target pixels related to the polygon Q are represented by black square marks. The input / output relationship of the pixel information is represented by broken-line arrows, which means transfer from the texture map to the accumulation map.

  Subsequently, with respect to the scanning target pixel, first, attention coordinates on the accumulation map side are derived from the position of the pixel according to a deformation method as shown in the figure. Here, the attention coordinate regarding the polygon P in the accumulation map is represented by a black triangle mark, and the attention coordinate regarding the polygon Q is represented by a black square mark. Note that the coordinates of interest derived at this stage are not necessarily pixel units (for example, real numbers).

  Next, the pixel of the accumulation map related to the target coordinate is determined as the update target pixel. In this example, following the conventional method of deriving a position in the nearest neighbor method, one pixel that is closest to the target coordinate, that is, one pixel in the region that includes the target coordinate is used as the update target pixel. To do. However, it should be noted that the pixel at the required position is not the input side but the output side as a difference from the conventional nearest neighbor interpolation. Here, the position of the pixel to be updated in the accumulation map corresponding to (i, j) is (x, y), x is a horizontal position, and y is a vertical position.

Next, a_src (i, j), r_src (i, j), g_src (i, j), and b_src (i, j), which are pixel information in the scanning target pixel of the texture map, are used as follows. Update w_sum (x, y), a_sum (x, y), r_sum (x, y), g_sum (x, y), and b_sum (x, y), respectively, in the pixel to be updated , Description of positions (i, j) and (x, y) is omitted).
That is,
For w_sum, a fixed value (for example, 1) is added.
For a_sum, only a_src is added.
For r_sum, only a value obtained by multiplying a_src by r_src is added.
For g_sum, only a value obtained by multiplying a_src by g_src is added.
For b_sum, only a value obtained by multiplying a_src by b_src is added.

When only a completely opaque image is handled, a_src can be replaced with a constant value (for example, 1), and a_sum can be substituted with w_sum.
That is, the update of the update target pixel of the accumulation map can be simplified as follows.
For w_sum, a fixed value (for example, 1) is added.
Only r_src is added to r_sum.
Only g_src is added to g_sum.
Only b_src is added to b_sum.

  Along with the processing as described above, the scanning of the texture map of each polygon is completed.

  The lower left of FIG. 7 represents the distribution of the influence value w_sum added in the accumulation map as a numerical value as a texture map scanning result. For convenience of explanation, the shaded area indicates pixels having an influence value of a certain value (for example, 2) or more. Details will be described later.

  The gray area is an example of a part of the correspondence between the texture map and the accumulation map, and the three pixels at the positions (7, 9) (8, 9) and (7, 10) in the texture map are positions in the accumulation map. An example in which pixel update of (9, 5) was involved is shown.

  The lower right of FIG. 7 shows an example of addition performed on the pixel at the position (9, 5) in the accumulation map as a result of scanning the texture map.

In accordance with the rules described above, addition is performed as follows.
w_sum (9,5) = 1 + 1 + 1 = 3
a_sum (9,5) = 1 × a_src (7,9)
+ 1 × a_src (8, 9)
+ 1 × a_src (7, 10)
r_sum (9,5) = 1 × a_src (7,9) × r_src (7,9)
+ 1 * a_src (8,9) * r_src (8,9)
+ 1 × a_src (7, 10) × r_src (7, 10)
g_sum (9,5) and b_sum (9,5) can be obtained in the same manner as r_sum (9,5).

  As in the example of FIG. 8, if the texture map is wrapping (repeated), it should be noted that scanning is performed in consideration of repetition when scanning the range surrounded by the polygons on the texture map side. To do. For example, the pixel at position (2, 2) is to be scanned a total of four times, three times for polygon P and once for polygon Q.

  Next, a flow of processing for updating the accumulation map pixel group corresponding to the texture map pixel by the resampling apparatus according to the first embodiment of the present invention will be described with reference to the flowchart of FIG. 9. This corresponds to the details of the processing in step S16 in FIG. This corresponds to a part of the resampling method, and also corresponds to a part of the processing when the central control unit 11 executes the resampling program of the main memory 12.

This flowchart shows a part of means for suppressing the shift of the center of gravity due to the reduction.
Upon entering this process, “accumulation map positions (B1 to B4)” and “weight values (KB1 to KB4)” are created from the accumulation map attention coordinates (B) corresponding to the pixel positions (A) of the texture map. (Step S21). Subsequently, the accumulated map pixel information group (SB1 to SB4) is updated from the texture map pixel information (DA) and the weight values (KB1 to KB4) (step S22), and the process returns.

  Note that “texture map” is synonymous with “transformation source image”.

  Hereinafter, the flow of processing for updating the accumulation map pixel group corresponding to the texture map pixel will be described in more detail with reference to the conceptual diagram of FIG.

  In the above example, following the conventional method of deriving the position in the nearest neighbor method, one pixel that is closest to the target coordinate is set as the update target pixel.

  In the example of FIG. 10, as a method of suppressing the shift of the center of gravity in deformation, following the position deriving method and weighting method in the conventional bilinear method, a plurality of (for example, two horizontal and two vertical) positions close to the target coordinate are used. 2 × 2 = 4) pixels are set as update target pixels. However, it should be noted that, unlike the conventional bilinear interpolation, a plurality of pixels to be weighted are not on the input side but on the output side. Details will be described below.

  First, the attention coordinate B of the accumulation map corresponding to the position A of the scanning target pixel in the texture map is obtained by forward conversion. A broken-line arrow from A to B represents forward conversion of coordinates. Here, when the coordinate map is used, the horizontal coordinate and the vertical coordinate of B are stored in the element of the position A of the coordinate map. This eliminates the need to obtain the attention coordinate B corresponding to the position A again later.

  Next, the positions of the four pixels close to the target coordinate are B1, B2, B3, and B4, respectively. At each of the four positions, weighting is performed according to the distance from B, and weight values KB1, KB2, KB3, and KB4 are derived. The sum of the four weight values is a constant value (for example, 1).

  Here, the pixel information related to the position A is DA, and the pixel information SB1 related to the position B1 is incremented by “DA considering KB1” from the current value.

That is, a_src (A), r_src (A), g_src (A), and b_src (A), which are pixel information in the scanning target pixel of the texture map, are used as follows. Certain w_sum (B1), a_sum (B1), r_sum (B1), g_sum (B1), and b_sum (B1) are respectively updated (the description of the positions (A) and (B) is omitted hereinafter).
That is,
Only KB1 is added to w_sum.
For a_sum, only the value obtained by multiplying KB1 by a_src is added.
For r_sum, only the value obtained by multiplying KB1 by a_src and r_src is added.
For g_sum, only the value obtained by multiplying KB1 by a_src and g_src is added.
For b_sum, only the value obtained by multiplying KB1 by a_src and b_src is added.
This calculation is expressed as “SB1 + = KB1 · DA” as a simple expression.
“+ =” Represents addition to the left side itself, that is, indicates that the result of “SB1 + KB1 · DA” is substituted into SB1.

When only a completely opaque image is handled, a_src can be replaced with a constant value (for example, 1), and a_sum can be substituted with w_sum.
That is, the update of the update target pixel of the accumulation map can be simplified as follows.
Only KB1 is added to w_sum.
For r_sum, only the value obtained by multiplying KB1 by r_src is added.
Only g_sum multiplied by g_src is added to g_sum.
For b_sum, only the value obtained by multiplying KB1 by b_src is added.

  The above addition is similarly performed on the pixel information SB2, SB3, and SB4 related to the positions B2, B3, and B4.

  Thus, while the purpose of position derivation and weighting is addition of pixel information in the update target pixel, it is possible to apply a conventional pixel position derivation method and weighting method for reference purposes. As a similar example, a bicubic method or the like can be applied.

  Hereinafter, with reference to the flowchart of FIG. 11, a flow of processing of scanning the deformation destination image and updating the deformation destination image based on the accumulation map by the resampling apparatus according to the first embodiment of the present invention will be described. . This process corresponds to step S2 in FIG. This corresponds to a part of the resampling method, and also corresponds to a part of the processing when the central control unit 11 executes the resampling program of the main memory 12.

  This flowchart is an example of texture mapping using a polygon model, and is an explanation for scanning a deformation destination image. The target map corresponds to a deformation destination image.

  That is, when entering this process, it is determined whether or not the extraction of all polygon faces has been completed (step S31). If the extraction has not been completed (step S31 branches to No), the next polygon face is extracted. (Step S32) Subsequently, it is determined whether or not the extraction of all the pixels included in the polygon surface in the target map has been completed (Step S33). If not completed (Step S33 branches to No), the next of the target map is determined. A pixel is extracted (step S34), the target map pixel is updated based on the accumulation map (step S35), and the process returns to step S33. In this way, the processing of steps S33 to S35 is repeated until extraction of all the pixels included in the polygon plane in the target map is completed. When extraction of all pixels is completed (step S33 branches to Yes), the process returns to step S31. In this way, the processing of steps S31 to S35 is repeated until the extraction of all the polygon surfaces is completed.

However, the flowchart of FIG. 11 is an example assuming a polygon model (that is, a plurality of deformation methods) composed of a plurality of polygons. In the case of a single polygon or a single table or formula in the deformation method definition data 42. When S is used, steps S31 to S32 can be omitted.
That is, when this process is started, the process of steps S33 to S35 may be repeated, and when extraction is completed (step S33 is branched to Yes), the process may be returned.

  Note that “target map” is synonymous with “deformation destination image”, and “polygon surface” is synonymous with “deformation range” in the transformation source image.

  Here, a flow of processing for updating the target map pixel based on the accumulation map will be described with reference to the flowchart of FIG. This corresponds to the details of the process in step S35 of FIG. This corresponds to a part of the resampling method, and also corresponds to a part of the processing when the central control unit 11 executes the resampling program of the main memory 12. This flowchart is an example of simply updating the target map pixel, and shows means for suppressing the shift of the center of gravity in the case of reduction. Note that means suitable for deformation in which reduction and enlargement are combined at the reference point of the image will not be described here and will be described later.

  In this process, first, it is determined whether or not the influence value (w_sum) of the pixel information (SP) related to the position (P) of the accumulation map pixel is a certain value or more (step S41). If the influence value (w_sum) is equal to or greater than a certain value (step S41 is branched to Yes), target map pixel information (DP) is generated from the accumulated map pixel information (SP) (step S42), and the process returns. On the other hand, if the influence value (w_sum) is not equal to or greater than a certain value (step S41 is branched to No), target map pixel information (DP) is generated from the texture map pixel information DA1 to DA4 by the conventional interpolation method (step S43). To return.

  It should be noted that “target map” is synonymous with “deformation destination image”, and “texture map” is synonymous with “deformation source image”.

  Hereinafter, with reference to the conceptual diagram of FIG. 13, the process of scanning the deformation destination image and updating the deformation destination image based on the accumulation map by the resampling apparatus according to the first embodiment of the present invention will be described in detail. To do.

  The right figure of FIG. 13 is the above-described texture map corresponding to the deformation source image. The upper left figure of FIG. 13 is the above-described accumulation map, and it is assumed that the above-described scanning on the texture map side has been completed. The lower left diagram in FIG. 13 represents a target map corresponding to the deformation destination image.

  It is assumed that the pixels of the target map have pixel information (a_dst: alpha, r_dst: red, g_dst: green, b_dst: blue) including opacity and color.

  In each of the texture map, the target map, and the accumulation map, the pixels are two-dimensionally arranged, and the unit of coordinates (horizontal direction and vertical direction) is one pixel, and for convenience, the boundary of the pixel area is a dotted grid. Represents.

  On the target map side, the range surrounded by each polygon is scanned without being separated. Here, the position of the pixel to be scanned is (i, j), i is the horizontal position, and j is the vertical position. Incidentally, as described above, since the accumulation map has a common coordinate system with the target map, that is, (i, j) can also be expressed as the pixel position of the accumulation map corresponding to the position of the pixel to be scanned. The scanning target pixel related to the polygon P in the target map is represented as a black triangle mark or white triangle mark, and the scanning target pixel related to the polygon Q is represented as a black square mark or white square mark.

  The input / output relationship of the pixel information is represented by broken-line arrows, which means transfer from the accumulation map to the target map, or transfer from the texture map to the target map. Here, a white triangle mark or a white square mark indicates that the influence value w_sum (i, j) is a certain value (for example, 2) or more, and a black triangle or a black square mark is an influence value. This indicates that w_sum (i, j) is less than a certain value (for example, 2). Different update target pixel update methods are used when the influence value w_sum (i, j) is 2 or more and less than 2.

Regarding the scanning target pixel, first, the influence value w_sum (i, j) of the pixel in the accumulation map corresponding to the position of the pixel is read out. When the influence value w_sum (i, j) in the accumulation map is 2 or more, it is considered that the reduction is locally equivalent, the number of pixels in the texture map involved in the pixel is considered to be sufficient, and the average of the pixels involved is used. To do.
That is, using the pixel information w_sum (i, j), a_sum (i, j), r_sum (i, j), g_sum (i, j), and b_sum (i, j) in the accumulation map as follows: The pixel information a_dst (i, j), r_dst (i, j), g_dst (i, j), b_dst (i, j) in the update target pixel of the target map is updated (hereinafter, the position (i, j)). Description is omitted).
That is,
A value obtained by dividing w_sum from a_sum is substituted for a_dst.
A value obtained by dividing a_sum from r_sum is substituted for r_dst.
A value obtained by dividing a_sum from g_sum is substituted for g_dst.
A value obtained by dividing a_sum from b_sum is substituted for b_dst.
The pixel information of the accumulation map is S (i, j), and this calculation is expressed as “S (i, j) / w_sum (i, j)” as a simple expression.

As described above, a_sum can be replaced by w_sum when an opaque image is always handled. That is, the update of the update target pixel of the target map can be replaced as follows.
That is,
A value obtained by dividing w_sum from r_sum is substituted for r_dst.
A value obtained by dividing w_sum from g_sum is substituted for g_dst.
A value obtained by dividing w_sum from b_sum is substituted for b_dst.

  If the influence value w_sum (i, j) of the pixel of the accumulation map corresponding to the position of the pixel to be scanned is less than 2, it is considered that it corresponds to enlargement locally, and the number of pixels of the texture map engaged with the pixel is insufficient. Consider. In this case, the pixel information in the update target pixel of the target map is updated using a value calculated by interpolation (for example, bilinear interpolation) based on conventional texture mapping without using the pixel information of the pixel.

  That is, for the pixel to be scanned, first, the target coordinate on the texture map side is derived from the position related to the pixel according to the deformation method as shown in the figure. The attention coordinates regarding the polygon P in the texture map are represented as black triangle marks, and the attention coordinates regarding the polygon Q are represented as black square marks. Note that the coordinates of interest derived at this stage are not necessarily pixel units (eg, real numbers).

  Next, the pixel of the texture map related to the target coordinate is determined as the referenced pixel. Here, 2 × 2 pixels close to the target coordinate, that is, a pixel in a region including the target coordinate and a neighboring pixel are set as a reference pixel. Details will be described below.

FIG. 14 shows an example of interpolation based on conventional texture mapping.
A generation example of the pixel information DP according to the position P regarding the target map is expressed by the following equation.
DP = (KA1, DA1 + KA2, DA2 + KA3, DA3 + KA4, DA4)
/ (KA1 + KA2 + KA3 + KA4)

  That is, first, the attention coordinate A of the texture map corresponding to the position P of the target map is obtained by inverse transformation. The dashed arrow from P to A represents the inverse transformation of coordinates. At the positions A1, A2, A3, and A4 of four pixels close to A, weighting is performed according to the distance from A, and weight values KA1, KA2, KA3, and KA4 are derived. The sum of the four weight values is a constant value (for example, 1). The pixel information related to the positions A1 to A4 is DA1 to DA4, and the pixel information DP related to the position P is generated by combining “DA1 to DA4 considering KA1 to KA4”.

  FIG. 15 shows, as a modification, a concept of combining weighted values calculated from conventional interpolation and values calculated from an accumulation map.

  In the above-described example (the process of FIG. 12), different pixel update methods are used depending on whether the pixel influence value w_sum in the accumulation map is greater than or equal to a certain value. Each pixel value may be calculated, and pixel information synthesized using a ratio corresponding to the value of the influence value w_sum may be used as the pixel information of the update target pixel. That is, when the influence value w_sum is large, the ratio of using the pixel information SP at the position P of the accumulation map is also increased. As a result, the boundary between the enlarged portion and the reduced portion becomes smooth.

  FIG. 16 illustrates an example of the resampling result.

  FIG. 16 is an example in which texture mapping when projecting the inside of a rectangular parallelepiped with a very long depth is compared with the anisotropic filter of the conventional means. FIG. 16A is a conventional anisotropic filter (16 × sample setting), and FIG. 16B is a means of the present invention, each emphasizing a far portion, that is, a deformation including reduction. Focusing on the checkered pattern and the straight line pattern, the results of the present invention show that the quality of the distant part is kept high particularly from the viewpoint of the average and the center of gravity of the pixels.

(Second Embodiment)

  In the processing by the resampling apparatus according to the second embodiment, the processing of FIG. 12 by the resampling apparatus according to the first embodiment is changed as shown in FIG.

Hereinafter, with reference to the flowchart of FIG. 17, the flow of processing for updating the target map pixel based on the accumulation map by the resampling apparatus according to the second embodiment of the present invention will be described. This flowchart shows means suitable for deformation in which reduction and enlargement are combined at the reference point of an image, in addition to means for suppressing the shift of the center of gravity due to reduction. This corresponds to a part of the resampling method according to the second embodiment, and also corresponds to a part of processing when the central control unit 11 executes the resampling program of the main memory 12.
Details will be described below.

  That is, when entering this process, from the texture map attention coordinate (A) corresponding to the pixel position (P) of the accumulation map, “texture map position (A1 to A4)” and “weight value (KA1 to KA4)” (Step S51), and from the accumulated map attention coordinates (B1 to B4) corresponding to the texture map positions (A1 to A4), “accumulated map positions (B11 to B44)” and “weight values (KB11 to KB44)”. (Step S52), and pixel information (DP) of the position (P) of the target map is generated from the pixel information (DB11 to DB44) and weight values (KA1 to KA4, KB11 to KB44) of the accumulation map (step S52). In step S53, the process returns.

  It should be noted that “target map” is synonymous with “deformation destination image”, and “texture map” is synonymous with “deformation source image”.

  Here, the process of updating the target map pixel based on the accumulation map will be described in more detail with reference to the conceptual diagram of FIG.

  First, the attention coordinate A of the texture map corresponding to the position P of the accumulation map is obtained by inverse transformation. The dashed arrow from P to A represents the inverse transformation of coordinates. Next, weighting is performed according to the distance to A at the positions A1, A2, A3, and A4 of the four pixels close to the target coordinate A, and the weight values KA1, KA2, KA3, and KA4 are derived. The sum of the four weight values is a constant value (for example, 1).

  Further, the attention coordinate B1 of the accumulation map corresponding to the position A1 is obtained by forward conversion. A broken-line arrow from A1 to B1 represents a forward conversion of coordinates. Next, weights KB11, KB12, KB13, and KB14 are derived by weighting according to the distance from B1 at the four pixel positions B11, B12, B13, and B14 close to the target coordinate B1. The sum of the four weight values is a constant value (for example, 1).

  In the case where a coordinate map is used, if coordinate information that is an element of the position A1 of the coordinate map exists, the coordinate information may be B1.

  Although coordinates, positions, and the like corresponding to A2, A3, and A4 are not shown, these are similarly derived.

  Then, the pixel information regarding the positions B11 to B44 is set as DB11 to DB44, and the pixel information DP regarding the position P is generated by synthesizing “DB11 to DB44 taking into account KB11 to KB44 and KA1 to KA4”.

  By the way, in the state where the above-described scanning on the texture map side is completed, there is a gap in the update of the pixels of the accumulation map in the enlarged portion, and the pixels may not be updated at all depending on the position. However, the coordinate conversion for referring to the accumulation map in the method described here has already been performed at the time of scanning on the texture map side, and the pixels of the associated accumulation map have been updated. That is, it may be considered that the position of a pixel that has not been updated at all is not derived. Therefore, there is no need to separate processing for reduction and enlargement, so this method can be said to be resampling suitable for deformation in which reduction and enlargement are combined at the reference point of an image.

  FIG. 19 shows an example of the resampling result according to the second embodiment.

The forward transformation of coordinates (input → output) and the inverse transformation of coordinates (output → input) are expressed by the following equations.
Coordinate forward conversion (input → output):
X_OUT = (w / 2) + cos (X_IN / w * 2 * PI) * Y_IN / h * (w / 2)
Y_OUT = (h / 2) + sin (X_IN / w * 2 * PI) * Y_IN / h * (h / 2)
Inverse transformation of coordinates (output → input):
X_IN = atan2 ((Y_OUT-h / 2) / h, (X_OUT-w / 2) / w) / (2 * PI) * w
Y_IN = sqrt (((X_OUT-w / 2) / (w / 2)) ^ 2 + ((Y_OUT-h / 2) / (h / 2)) ^ 2) * h

  FIG. 19 is an example of a result obtained by performing transformation by polar coordinate transformation by means of the resampling apparatus according to the second embodiment of the present invention.

  FIG. 19B shows a deformation original image in which squares are arranged in a grid pattern as a pattern. The uppermost part of the transformation source image corresponds to the center of the transformation destination image, and the line segment from the lower left to the lower right of the transformation source image corresponds to a circle that makes one round clockwise from the right end of the transformation destination image. FIG. 19A shows coordinates obtained by forward conversion from the position of each pixel in the transformation source image in the accumulation map. Black circles are used for black pixels and white coordinates are used for white pixels. Represented by open circles. FIG. 19D shows the coordinates obtained by inverse transformation from the position of each pixel serving as the transformation destination image in the transformation source image with white circles.

  In the example of FIG. 19 (a), near the end of a circle, the coordinate group is reduced in the radial direction because the coordinate group is dense, but in the circumferential direction, the coordinate group is sparse, so that the deformation is enlarged. It is. As a result of the deformation, as shown in FIG. 19C, jaggies are suppressed by interpolation in the enlargement direction, and aliasing is suppressed by averaging in the reduction direction.

  The first and second embodiments of the present invention have been described in detail above. However, the present invention is not limited to these, and various improvements and modifications can be made without departing from the spirit of the present invention.

  For example, the present invention may be applied to the drawing processing described in the omnidirectional image editing apparatus described in the specification and drawings after Japanese Patent Application No. 2011-096940, which is a prior application of the same inventor. In this case, resampling that suppresses the shift of the center of gravity due to the reduction can be expected in a fine drawing in a state where the viewing angle is extremely small, that is, in a state where the display is enlarged. Also, resampling suitable for deformation in which reduction and enlargement are combined at the reference point of an image can be expected in drawing that expresses a sense of perspective in a state where the viewing angle is extremely large, that is, in a state where the display is wide.

  Also, the present invention does not use means limited to two dimensions. Therefore, any processing can be extended to three dimensions (for example, transformation of three-dimensional data using voxels).

DESCRIPTION OF SYMBOLS 1 Resampling apparatus 10 Personal computer 11 Central control part 12 Main memory 13 Image control part 13a Video memory 14 Input / output control part 15 Built-in non-volatile storage medium 16 Media read / write interface 20 Display apparatus 30 Deformation parameter input apparatus 40 Deformation original image input data 41 Deformation destination image output data 42 Deformation method definition data 50 Program data

Claims (12)

A resampling method using an accumulation map having a sequence in common with a transformation destination image,
A first step of scanning a deformation source image and updating the accumulation map based on the deformation source image;
Scan the transformation destination image, extract the next pixel of the transformation destination image, and update the transformation destination image based on the accumulation map until extraction of all pixels included in the transformation range in the transformation destination image is completed. A second step,
In the process of updating the pixel of the deformation destination image based on the accumulation map,
It is determined whether or not the influence value of the pixel information related to the pixel position of the accumulation map is greater than or equal to a certain value. If the influence value is greater than or equal to the certain value, the pixel information of the transformation destination image is generated from the accumulation map pixel information. If the influence value is not equal to or greater than a certain value, the pixel information of the transformation destination image is generated from the pixel information of the transformation source image by a predetermined interpolation method. A sampling method using an accumulation map having a common sequence with a transformation destination image,
A first step of scanning a deformation source image and updating the accumulation map based on the deformation source image;
Scan the transformation destination image, extract the next pixel of the transformation destination image, and update the transformation destination image based on the accumulation map until extraction of all pixels included in the transformation range in the transformation destination image is completed. A second step,
In the process of updating the pixel of the deformation destination image based on the accumulation map,
The pixel information of the transformation destination image is generated by combining the value calculated from the accumulation map pixel information and the value calculated from the pixel information of the transformation source image with a predetermined complement method using a ratio corresponding to the influence value. A resampling method characterized by the above. In the first step,
The accumulation map is initialized, and extraction of the next pixel of the deformation source image is continued until extraction of all pixels included in the range to be deformed in the deformation source image is completed, and the accumulation map corresponding to the pixel of the deformation source image The resampling method according to claim 1, wherein the pixel group is updated. The range to be deformed in the deformation source image is a polygon surface,
4. The resampling method according to claim 3, wherein in the first and second steps, extraction of the next polygon surface is continued until extraction of all polygon surfaces is completed. A resampling device that uses an accumulation map having a common sequence with a transformation destination image,
Display means for displaying various images;
A central control means for controlling according to a resampling program,
The central control means is
A first step of scanning the transformation source image and updating the accumulation map based on the transformation source image, and scanning the transformation destination image to complete extraction of all pixels included in the range to be transformed in the transformation destination image. The second step of extracting the next pixel of the deformation destination image and updating the deformation destination image based on the accumulation map,
In the process of updating the pixel of the deformation destination image based on the accumulation map,
It is determined whether or not the influence value of the pixel information related to the pixel position of the accumulation map is greater than or equal to a certain value. If the influence value is greater than or equal to the certain value, the pixel information of the transformation destination image is generated from the accumulation map pixel information. If the influence value is not equal to or greater than a certain value, the pixel information of the transformation destination image is generated from the pixel information of the transformation source image by a predetermined interpolation method. A resampling device that uses an accumulation map having a common sequence with a transformation destination image,
Display means for displaying various images;
A central control means for controlling according to a resampling program,
The central control means is
A first step of scanning the transformation source image and updating the accumulation map based on the transformation source image, and scanning the transformation destination image to complete extraction of all pixels included in the range to be transformed in the transformation destination image. The second step of extracting the next pixel of the deformation destination image and updating the deformation destination image based on the accumulation map,
In the process of updating the pixel of the deformation destination image based on the accumulation map,
The pixel information of the transformation destination image is generated by combining the value calculated from the accumulation map pixel information and the value calculated from the pixel information of the transformation source image with a predetermined complement method using a ratio corresponding to the influence value. A resampling device characterized by the above. In the first step, the central control means initializes the accumulation map and continues to extract the next pixel of the original image until extraction of all pixels included in the range to be deformed in the original image is completed. The resampling apparatus according to claim 5 or 6, wherein a process of updating a pixel group of an accumulation map corresponding to the pixel of the deformation source image is performed. The range to be deformed in the deformation source image is a polygon surface,
8. The resampling apparatus according to claim 7, wherein in the first and second steps, extraction of the next polygon surface is continued until extraction of all polygon surfaces is completed. Computer
A first step of scanning the deformation source image and updating an accumulation map having a common series with the deformation destination image based on the deformation source image, and within a range where the deformation destination image is scanned and deformed in the deformation destination image A second control unit that performs a second step of extracting a next pixel of the transformation destination image and updating the transformation destination image based on the accumulation map until extraction of all included pixels is completed. Because
In the process of updating the pixel of the transformation destination image based on the accumulation map, it is determined whether or not the influence value of the pixel information related to the pixel position of the accumulation map is a certain value or more, and the influence value is a certain value or more. If there is, the pixel information of the transformation destination image is generated from the accumulated map pixel information, and if the influence value is not a certain value or more, the pixel information of the transformation destination image is generated from the pixel information of the transformation source image by a predetermined interpolation method. A resampling program characterized by Computer
A first step of scanning the deformation source image and updating an accumulation map having a common series with the deformation destination image based on the deformation source image, and within a range where the deformation destination image is scanned and deformed in the deformation destination image A second control unit that performs a second step of extracting a next pixel of the transformation destination image and updating the transformation destination image based on the accumulation map until extraction of all included pixels is completed. Because
In the process of updating the pixel of the deformation destination image based on the accumulation map,
The pixel information of the transformation destination image is generated by combining the value calculated from the accumulation map pixel information and the value calculated from the pixel information of the transformation source image with a predetermined complement method using a ratio corresponding to the influence value. A resampling program characterized by In the first step, the accumulation map is initialized, and extraction of the next pixel of the deformation source image is continued until extraction of all pixels included in a range to be deformed in the deformation source image is completed. The resampling program according to claim 9 or 10, wherein a process of updating a pixel group of an accumulation map corresponding to the pixel is performed. The range to be deformed in the deformation source image is a polygon surface,
12. The resampling program according to claim 11, wherein in the first and second steps, extraction of the next polygon surface is continued until extraction of all polygon surfaces is completed.