JP4166646B2 - Digital image enlargement interpolation method, digital image compression method, digital image restoration method, recording medium on which digital image enlargement interpolation program is recorded, recording medium on which digital image compression program is recorded, and digital image restoration program Recorded recording medium - Google Patents

Description

  The present invention is an image processing method for a digital image composed of pixels arranged at a fixed interval, a method for interpolating a gap between pixels generated when a digital image is enlarged, a method for compressing digital image data, The present invention also relates to a method for restoring a compressed image.

  Digital images that are widely used at present, for example, digital images obtained by digital cameras and scanners, and digital images created by computer graphics are units of pixels (pixels) that include position data, luminance data, and other color data. It is managed by.

  And when enlarging this digital image, it is done by widening the distance between the pixels, but simply widening the distance between the pixels creates a gap between the pixels, resulting in a rough image as a whole. End up.

  In order to eliminate the roughness that occurs when enlarging such a digital image, conventionally, a method of interpolating the pixels obtained by calculation in the gap between pixels that occurs when enlarging the digital image has been used. Various methods for calculating the pixel by calculation have been proposed.

  For example, an algorithm called a bilinear method is adopted as one of the calculation methods. The bilinear method is based on a unit of a quadrangle having four adjacent pixels as vertices, and a gap in the quadrangle generated when the image is enlarged is obtained from position data and luminance data of the four pixels located at the vertices of the quadrangle. Interpolation is performed with the calculated pixels. This method is easy to calculate and is now widely used.

In addition, an application has been filed for an interpolation method that should be called a development system of the bilinear method. This is because the rectangular area, which is the minimum unit shown in the description of the bilinear method, is scanned over the entire digital image, and the luminance of one pixel among the pixels at the apex of the square is that of the other three pixels. It is determined whether or not it is greatly different from the luminance. If it is determined that the luminance is different, the triangular area formed by the three pixels is interpolated with the pixel having the luminance data obtained from the pixel located at the vertex of the triangle. The method of doing is applied (for example, refer patent document 1).
JP-A-9-102867

  However, when an interpolation method such as the Vignier method is used to calculate the pixels to be filled in the gap, the portion color-coded by the diagonal line or curve is interpolated. Staircase noise may be noticeably generated on the lighter side. This staircase-like noise has a drawback that the boundary portion in the image becomes jagged and rough, and the deterioration of the image is clearly impressed by the viewer.

  On the other hand, although it is possible to reduce the stepped noise (jaggy) generated at the boundary portion of the diagonal line or curve by the method described in Patent Document 1, in this method, the density is clearly reduced. It can be applied only to divided parts, for example, a special part where black characters are written on a white background, and is difficult to apply to a gradation part where the color gradually changes. In addition, every time the image is enlarged, all the pixels of the digital image must be scanned to determine the shading, so that considerable processing time is required.

  The present invention has been made in order to solve the above-described problem, and reduces digital noise that occurs when enlarging a digital image, and can cope with an image having a fine pattern or gradation. An object is to provide an enlargement interpolation method of an image.

In order to solve the above-mentioned problem, the first invention is a method of enlarging and interpolating a digital image expressed by a collection of pixels arranged at a constant interval by a digital image processing device , and each of three vertices. Is virtually divided according to the following conditions (1) to (3), and [(1) triangles do not overlap. (2) All the pixels within the enlargement target range are located inside or on the side of any triangle. (3) A triangular plane is obtained from the three dimensions of the position data and luminance data of the three pixels located at the three vertices of each triangle, and the actual luminance data of each pixel located inside or on the corresponding triangle; Comparing calculated luminance data obtained from the actual position data of the same pixel using the triangular plane, the difference between them is within a predetermined range, or the result of statistically processing the difference is a predetermined range. Be within. ] Further, the actual luminance data at the vertices on the overlapping sides of the smaller triangle plane and the position data of the vertices on the larger triangle plane, which are generated when the sizes of the adjacent triangular planes on which one side overlaps, are different. In order to eliminate the difference from the calculated luminance data obtained by substituting, the larger triangle plane is further divided into triangle planes, and the process of making adjacent triangle planes continuous is performed. A digital image enlargement interpolation method in a digital image processing apparatus , wherein a gap between generated pixels is interpolated with pixels having position data and luminance data obtained from the triangular plane .

  In the first aspect of the invention, the triangle is a right triangle, and on both sides sandwiching the right angle, 2 pixels are added to the n-th power of 1 (n is an integer equal to or greater than 0), and the digital image If the number of pixels arranged on the side in the peripheral portion is less than the number obtained by the above formula, an arbitrary pixel may be arranged at the position of at least the apex of the triangle outside the peripheral edge of the digital image.

  Furthermore, in the said invention, it is good also as the number of pixels located on both sides on both sides of the right angle of the said triangle.

On the other hand, a fourth invention is a digital image compression method for compressing a digital image expressed by an aggregate of pixels arranged at regular intervals by a digital image processing device , wherein each of three vertices is on a pixel. The digital image is virtually divided according to the following conditions (1) to (3) by a plurality of triangles positioned at [3], and [(1) triangles do not overlap. (2) All the pixels within the compression target range are located inside or on the side of any triangle. (3) A triangular plane is obtained from the three dimensions of the position data and luminance data of the three pixels located at the three vertices of each triangle, and the actual luminance data of each pixel located inside or on the corresponding triangle; Comparing calculated luminance data obtained from the actual position data of the same pixel using the triangular plane, the difference between them is within a predetermined range, or the result of statistically processing the difference is a predetermined range. Be within. ] Further, the actual luminance data at the vertices on the overlapping sides of the smaller triangle plane and the position data of the vertices on the larger triangle plane, which are generated when the sizes of the adjacent triangular planes on which one side overlaps, are different. In order to eliminate the difference from the calculated luminance data obtained by substituting, the larger triangular plane is further divided into triangular planes, and the adjacent triangular planes are continuously processed. The above-mentioned problem is solved by adopting a digital image compression method in a digital image processing apparatus characterized by extracting only pixels located.

  In the fourth aspect of the invention, the triangle is a right triangle, and on both sides sandwiching the right angle, there are a number of pixels obtained by adding 1 to the nth power of 2 (n is an integer of 0 or more), and the digital image If the number of pixels arranged on the side in the peripheral portion is less than the number obtained by the above formula, an arbitrary pixel may be arranged at the position of at least the apex of the triangle outside the peripheral edge of the digital image.

  Furthermore, in the said invention, it is good also as the number of pixels located on both sides on both sides of the right angle of the said triangle.

In addition, for the image compressed according to the invention, a triangular plane is obtained from the three dimensions of the position data and luminance data of each of the three pixels constituting the triangle, and the position data and luminance data obtained from the triangular plane are provided. The image compressed by the compression method can be restored by the digital image processing apparatus by setting the obtained pixel as a pixel to be located inside or on the side other than the vertex of the triangle.

According to an eighth aspect of the invention, there is recorded an enlargement interpolation program for causing a computer of a digital image processing apparatus to execute a method of enlarging and interpolating a digital image expressed by a collection of pixels arranged at a constant interval. A recording medium, wherein the program virtually divides a digital image by a plurality of triangles each having three vertices positioned on a pixel according to the following conditions (1) to (3): [(1) Triangles do not overlap. (2) All the pixels within the enlargement target range are located inside or on the side of any triangle. (3) A triangular plane is obtained from the three dimensions of the position data and luminance data of the three pixels located at the three vertices of each triangle, and the actual luminance data of each pixel located inside or on the corresponding triangle; Comparing calculated luminance data obtained from the actual position data of the same pixel using the triangular plane, the difference between them is within a predetermined range, or the result of statistically processing the difference is a predetermined range. Be within. ] Substitute the actual luminance data at the vertex on the overlapping side of the smaller triangle plane and the position data of that vertex into the larger triangular plane, which occurs when the sizes of adjacent triangular planes with overlapping one side differ. In order to eliminate the difference from the calculated luminance data obtained in this step, the larger triangular plane is further divided into triangular planes, and the adjacent triangular planes are continued, and between the pixels generated when the digital image is enlarged The above problem is solved by a recording medium that causes a computer to execute the step of interpolating the gap with position data obtained from the triangular plane and pixels having luminance data.

Further, in the recording medium on which the digital image enlargement interpolation program is recorded, the triangle is a right triangle, and the number obtained by adding 1 to the nth power of 2 on both sides sandwiching the right angle (n is an integer of 0 or more). If the pixel is located and the program has less than the number of pixels obtained on the side in the peripheral part of the digital image obtained by the above formula, it is arbitrarily set at the position of at least the vertex of the triangle outside the peripheral part of the digital image. The step of arranging the pixels may be executed by a computer .

  Furthermore, in the said invention, it is good also as the number of pixels located on both sides on both sides of the right angle of the said triangle.

On the other hand, an eleventh aspect of the invention is a recording medium on which a compression program for causing a computer of a digital image processing apparatus to execute a method for compressing a digital image expressed by a collection of pixels arranged at regular intervals is recorded. The program includes a step of virtually dividing a digital image according to the following conditions (1) to (3) with a plurality of triangles each having three vertices located on a pixel, and [(1) triangles overlap. There is nothing. (2) All the pixels within the compression target range are located inside or on the side of any triangle. (3) A triangular plane is obtained from the three dimensions of the position data and luminance data of the three pixels located at the three vertices of each triangle, and the actual luminance data of each pixel located inside or on the corresponding triangle; Comparing calculated luminance data obtained from the actual position data of the same pixel using the triangular plane, the difference between them is within a predetermined range, or the result of statistically processing the difference is a predetermined range. Be within. ] Substitute the actual luminance data at the vertex on the overlapping side of the smaller triangle plane and the position data of that vertex into the larger triangular plane, which occurs when the sizes of adjacent triangular planes with overlapping one side differ. In order to eliminate the difference from the calculated luminance data obtained in this step, the larger triangular plane is further divided into triangular planes, and the adjacent triangular planes are continued, and only the pixels located at the three vertices of each triangle are obtained. The above problem is solved by a recording medium that causes a computer to execute the extracting step.

In addition, the triangle is a right triangle, and on both sides sandwiching the right angle, there are 2 pixels to which n is a power of 2 (n is an integer of 0 or more), and the program is executed at the peripheral portion of the digital image. If the number of pixels arranged on the side is less than the number obtained by the above equation, the computer may execute the step of arranging an arbitrary pixel at the position of at least the vertex of the triangle outside the periphery of the digital image. Absent.

  Furthermore, in the said invention, it is good also as the number of pixels located on both sides on both sides of the right angle of the said triangle.

Also, the restoration method for an image compressed by the invention, there is provided a recording medium on which restoration program is recorded for causing a computer to execute a digital image processing apparatus, the program, with respect to the compressed image, triangle Obtaining a triangular plane from the three dimensions of the position data and luminance data each of the three pixels constituting the pixel, and a pixel comprising the position data and luminance data obtained from the triangular plane inside the triangle other than the vertex of the triangle or The compressed image can be restored by a recording medium that causes the computer to execute the step of setting the pixel to be located on the side .

  According to the present invention, a digital image is virtually divided into triangles that satisfy certain conditions, a triangle plane is calculated based on the pixels located at the vertices of the triangles, and the pixels calculated from the triangle planes are used. Interpolates gaps that occur when enlarging digital images.

  That is, as shown in FIG. 2B, it is assumed that pixels are scattered in a three-dimensional space composed of two-dimensional position data and luminance data, and these pixels are approximated by a plurality of continuous triangular planes. Then, in order to interpolate the gaps between the pixels that occur when the image is enlarged using the triangular plane, the boundary portions that are color-coded by diagonal lines or curves in the digital image follow the boundary portions. If approximation is performed using a triangular plane, interpolation can be performed with an optimal approximate value. The enlarged image obtained by this interpolation method can suppress the generation of staircase noise even in a portion that is color-coded by diagonal lines, and an enlarged image that is difficult to feel roughness can be obtained.

  Also, if the triangle that divides the digital image is a right triangle, and each of the sides sandwiching the right angle has 2 n powers plus 1 (n is an integer of 0 or more), the number of pixels is located. A process for dividing a digital image into triangles is routine, and the process can be accelerated.

  That is, if a digital image is divided in advance with a triangle having a large value of n and the triangle does not satisfy a certain condition, the side sandwiching the right angle is divided into two equal parts to the power of 2−1. The number of pixels added by 1 is located on the side, and a line is virtually drawn in parallel to the side that sandwiches the right angle from the bisection point, and the resulting square region is divided by the diagonal line, Four triangles are created that are similar to the triangle before the division. If the same operation is repeated until n becomes 0, the process can be divided into the smallest triangles, so that the program for performing the process becomes shorter and the process becomes faster.

  On the other hand, according to another aspect of the present invention, a digital image is divided into triangles that satisfy certain conditions in the same manner as described above, and the interior of the triangle is based on a triangle plane calculated from pixels located at the vertices of the triangle. Alternatively, since it represents a pixel located on the side, digital image data can be compressed with high reproducibility even in an image having many portions that are color-coded by diagonal lines and curves.

  Furthermore, a program for compressing digital image data is simplified by the same operation as described above, and the compression process can be accelerated.

  In order to restore a digital image compressed by the method according to the present invention to a visible state, it is preferable to employ the restoration method according to the present invention.

  Next, the case of enlarging a color digital image using the interpolation method according to the present invention will be described with reference to the flowcharts of FIGS.

  First, as shown in the flowchart of FIG. 7, it is determined whether the digital image is color (S1). In this embodiment, since it is a color digital image, as shown in FIG. 1, the color digital image (a) is converted into three components of an R component (b), a B component (c), and a G component (d). Separate (S2). Thus, by separating the color digital image into digital images representing the luminance of the three color components, the pixels constituting the image for each composition sentence are pixels having three-dimensional values of luminance information and position information. Become.

  As a method for separating a color digital image into each component, any method such as dividing into each component of Y, Cb, and Cr can be adopted as long as the color can be expressed by further synthesis after the separation. sell.

  Next, each digital image for each component is virtually divided into triangles that satisfy certain conditions.

  In order to facilitate the operation of dividing into triangles, the pixels are virtually divided into a grid pattern (S3). It is assumed that the unit of the lattice is a square, and 2 ^ n + 1 pixels are located on each side. In the present specification, the term “square” means a virtual square, meaning that the same number of pixels are positioned on each side, and the length of the side is not considered.

  In the present embodiment, the size of the unit of the lattice to be divided first is a value obtained by substituting 6 for n, and this is referred to as the first lattice unit (S3). Further, the first grid unit starts from the upper left corner of the digital image, and is continuously arranged in the right direction and the lower direction while sharing one side.

  When pixels are virtually divided in a grid shape with a square of a certain size as a unit in this way, a first grid unit that does not include some pixels may be generated on the right side or the lower side of the digital image (S4). For a grid unit that does not include a pixel, a pixel having a certain luminance is replenished to fill the grid unit (S5).

  Next, as shown in FIG. 2A, each lattice unit is virtually divided by a diagonal line that rises to the right (FIG. 8: S6).

  By this operation, the pixels in each unit of the grid are virtually divided by two right isosceles triangles. The right isosceles triangle has vertices on the pixels and does not overlap with each other, and all the pixels are located inside or on the side of any right isosceles triangle. In the present specification, the isosceles triangle is a virtual isosceles triangle and the same number of pixels are located on the two sides of the triangle, as in the case of the square. The length of the side is not taken into account. In addition, the term “right angle” does not mean an exact geometric right angle.

  2 (a) to FIG. 5 (a) do not show the initial state by dividing the grid in which the 2 ^ 6 + 1 pixels are located on the side by diagonal lines, but to some extent. A part of the state in which the dividing operation has progressed is shown.

  Next, a triangular plane is obtained from the three-dimensional values of the position data and luminance data of the three pixels located at the three vertices of each isosceles right triangle (S7). Specifically, the triangular plane can be obtained by substituting the three-dimensional values of the three pixels into the formula aX + bY + cZ = 1 and obtaining the constants a, b, and c.

  The plane obtained from the above equation is an infinitely spreading plane, but the region necessary for the present invention is a region related to pixels located inside or on the side of the right-angled isosceles triangle. It is described as a triangular plane.

  A virtual representation of the relationship between the triangular plane and each pixel is the three-dimensional graph shown in FIG. Here, the X-axis and Y-axis in FIG. 2B indicate pixel positions in the digital image. Z represents luminance.

  Next, the difference between the actual luminance data of each pixel located inside or on the side excluding the three vertices of the corresponding right isosceles triangle and the calculated luminance data obtained from the actual position data of the same pixel using the triangle plane. Is statistically processed to determine whether the obtained result is less than a predetermined value. As one specific determination method, the present embodiment employs the following method. That is, the actual luminance data at the position r when the pixel position is represented by r is f (r), while the calculated luminance data obtained from the data at the position r using the triangular plane is g (r). The sum of squares of errors is calculated for two triangular planes according to the following equation. Formula: e ^ 2 = Σ (f (r) -g (r)) ^ 2. Then, e ^ 2 is substituted into the following equation to calculate the SNR value for judging the validity of the triangular plane (S8). Formula: SNR value = −10 × (log (e ^ 2/255 ^ 2)). In addition, since the difference between the actual luminance data and the calculated luminance data is 0 for the pixels located at the three vertices, it is not particularly necessary to be aware of excluding the three vertices.

Next, it is determined whether the SNR value is less than 100. When the SNR value of any of the two triangles included in the lattice unit is less than 100, the division operation is terminated for the lattice unit (S9: y).

In the case where the SNR value of at least one of the triangles in each lattice unit is 100 or more , as shown in FIG. 3A, each unit of the lattice is lowered to the right so that the division direction is different from the previous one. It divides again by the diagonal line which becomes (S10).

Then, a triangular plane is created for the changed triangle in the same manner as in the previous period (S11), the SNR value is obtained (S12), and if both of the SNR values of the two triangles in the lattice unit are less than 100, the division is performed. When at least one of the SNR values is 100 or more , the process proceeds to the next step and the division proceeds (S13).

  Next, it is determined whether the value of n is 0, that is, whether or not it is the minimum unit (S14). If n is 0, the division for the corresponding lattice unit is terminated.

Through the above process, for a lattice unit having a triangle with an SNR value of 100 or more , the midpoints of opposing sides of the square constituting the unit are virtually drawn to divide the lattice unit into four equal parts (S15 ). A square lattice unit having 2 ^ (n-1) +1 as one side is generated for the processed portion. In the present embodiment, after the processing for the first lattice unit with n = 6 is completed, the second lattice unit with n = 5 is created inside some or all of the first lattice units. . These are repeated sequentially.

With respect to the lattice unit newly generated by the above operation, the process returns to S6 to continue each process, and is repeated until the SNR value of the triangle included in the lattice unit becomes less than 100 or n becomes 0.

Next, crack processing as shown in FIGS. 9 to 11 is performed on the image virtually divided into triangles. The crack processing is an adjacent triangle with one side overlapping, and a triangle that occurs when the size of the triangle is different, that is, when the hierarchy to be divided is different (FIG. 9A: triangle ABC and triangle CDE). This is performed to eliminate the difference between the actual luminance data at the vertex D of the CDE and the calculated luminance data obtained by substituting the position data of the vertex D into the triangular plane obtained from the triangle ABC. By this crack processing, adjacent triangular planes are continuous (FIG. 9D).
This crack processing will be specifically described below.

  First, a vertex of a lattice unit in which 2 ^ n + 1 pixels are arranged on one side is referred to as n lattice points (P6 in FIG. 10A). That is, the vertex of the first lattice unit is 6 lattice points.

  In the case of the present embodiment, since the digital image is first divided into a grid shape with a size of 2 ^ 6 + 1, the crack processing starts by substituting 6 for n (S16).

  Next, the following processing is performed for each of the five lattice points (vertices of the second lattice unit) (P5 in FIG. 10B) (S17). In FIG. 10, for some of the lattice points, symbols are omitted for the sake of drawing.

  The process ends for the grid point 5 and the grid point 6 (S19). The grid point 5 which is not the grid point 6 is not a grid point existing at the left end of the image (S20), and one side of the grid unit does not extend to the left (S21). P5L), that is, when the lattice point to be determined is not the upper right corner and the lower right corner of the lattice unit, the triangle (E1 in FIG. 10D) that touches the lattice point from the left side is divided. (S22).

As a method for dividing the triangle, a line is drawn from the corresponding grid point (for example, P5L in FIG. 10D) to the other side or vertex of the triangle in parallel to the grid unit side to the left (for example, L1) in FIG. 10 (e), when the line reaches another side, the line is drawn from the reached point (for example, P5n in FIG. 10 (e)) to the vertex of the triangle not on the side. (For example, l2 in FIG. 10E), a method of virtually dividing the triangle is employed. Here, a point newly generated by the crack processing (point F in FIG. 9, P5n in FIG. 10) is not actual luminance data at that position, but calculated luminance data obtained from the position data by a triangular plane before division. Is adopted.

  Next, it is not the grid point existing at the right end of the image (S23), and one side of the grid unit does not extend rightward (S24), that is, the grid point to be judged is the one at the upper left corner of the grid unit. If it is not at the lower left corner, the triangle that touches the grid point from the right side is divided (S25).

  The method of dividing the triangle is substantially the same as the above except that the dividing direction is different, and draws a line from the corresponding grid point to the other side or vertex of the triangle in parallel to the grid unit side to the right. When reaching the other side, a method of virtually dividing the triangle is adopted by drawing a line from the reached point to the vertex of the triangle not on the side.

  Subsequently, it is not the grid point existing at the lower end of the image (S26), and one side of the grid unit does not extend downward (S27), that is, the grid point to be judged is the one at the upper left corner of the grid unit. If it is not in the upper right corner, the triangle that touches the grid point from the right side is divided (S28).

  The method of dividing the triangle is substantially the same as the above except that the dividing direction is different, and draws a line from the corresponding grid point downward to the other side or vertex of the triangle in parallel to the side of the grid unit, When reaching the other side, a method of virtually dividing the triangle is adopted by drawing a line from the reached point to the vertex of the triangle not on the side.

  The above operation is performed on all five grid points.

  Next, the same operation as described above is repeated for the four lattice points (P4 in FIG. 10F), and this is repeated until the second lattice point is reached. Thus, the crack process is completed. When crack processing is performed in this way, all triangle planes calculated from the position data and luminance data of each vertex of the divided triangles are continuous.

  Next, as shown in FIGS. 6A to 6B, the distance between the real pixels is increased to enlarge the digital image.

  By this enlargement, each triangular plane is also enlarged simultaneously. That is, since the position data of the pixels located at the three vertices of the triangle change, the constants a, b, and c that satisfy the expression aX + bY + cZ = 1 also change.

  Then, the pixel position data for interpolating the gap generated between the actual pixels is determined, and the calculated luminance data is calculated from the position data using the enlarged triangular plane as shown in FIG. 6C. Then, a pixel having the calculated luminance data and position data is created and interpolated.

  This completes the interpolation work. At this stage, a black and white enlarged digital image expressed only by the luminance of each color component is obtained.

  Finally, these processes are performed on an image obtained by dividing each process, and the digital image obtained can be recombined to obtain a color digital image.

  The above method is described as a program, and is executed by executing the program on a computer. The program is handled as one routine of the digital image processing software. The digital image processing software provided with the program is recorded on a CD-ROM and distributed.

  Next, a digital image compression method will be described. For the compression, the same operation as in steps S1 to S30 is performed to divide all digital images into triangles.

  Next, pixels that exist outside the three vertices of the triangle are deleted from the digital image. As a result, the amount of data constituting the digital image is reduced, and the compression is completed.

  In order to restore the digital image to a form that can be visually recognized from the compressed data, the constants a, b, c satisfying the expression aX + bY + cZ = 1 using the three-dimensional data of the pixels corresponding to the three vertices of the remaining triangle. To create a triangular plane. Then, as in the case of the interpolation, the position data of the pixel that should be inside or on the side of the triangle is applied to the triangle plane to calculate the calculated luminance data. A digital image can be displayed by arranging pixels having the position data and the calculated luminance data.

  In the present embodiment, the lattice unit is a square, but 2 ＾ n + 1 pixels may be arranged on one side as long as this pixel is arranged. In addition, although the size of the first lattice unit is n = 6, the initial size is arbitrary.

  As an application example of the present invention, there is a case where it is incorporated into a routine of digital image processing software as one of interpolation methods for enlarging a digital image.

  In addition, in order to compress data captured by a digital camera, there is a case where the processing method is installed in a ROM as a program and the image obtained by an image sensor is immediately compressed by being incorporated in the digital camera. It is done.

It is a figure which shows the image which isolate | separated the original image (color) and the said image by 3 components of color. (A) is a figure which shows a part of image after the said isolation | separation. (B) is the three-dimensional graph which showed the said part virtually. It is the figure which shows the case where the method of a division | segmentation is changed, and the graph which shows the said part in three dimensions. It is the figure which shows the state which the division | segmentation advanced to some extent, and the graph which shows the said part in three dimensions. Furthermore, it is the figure which shows the state which the division | segmentation advanced, and the graph which shows the said part in three dimensions. It is a figure which shows the interpolation state at the time of expansion. It is a flowchart which shows a part of process which divides | segments a digital image into a triangle. It is a flowchart following the flowchart shown in FIG. It is the conceptual diagram which expanded a part in order to explain crack processing It is a conceptual diagram for demonstrating a crack process. It is a flowchart showing a crack process.

Explanation of symbols

1, 2, 3, 4: Pixel a located at the apex of a triangle b) Triangle plane
X, Y: Position Z: Luminance P6, P5 ...: Grid point

Claims (14)

In a method of enlarging and interpolating a digital image represented by a collection of pixels arranged at regular intervals by a digital image processing apparatus ,
A digital image is virtually divided according to the following conditions (1) to (3) with a plurality of triangles each having three vertices located on a pixel,
(1) Triangles do not overlap.
(2) All the pixels within the enlargement target range are located inside or on the side of any triangle.
(3) A triangular plane is obtained from the three dimensions of the position data and luminance data of the three pixels located at the three vertices of each triangle, and the actual luminance data of each pixel located inside or on the corresponding triangle; Comparing calculated luminance data obtained from the actual position data of the same pixel using the triangular plane, the difference between them is within a predetermined range, or the result of statistically processing the difference is a predetermined range. Be within.
In addition, the actual luminance data at the vertex on the overlapping side of the smaller triangle plane, and the position data of that vertex are substituted into the larger triangle plane, which occurs when the sizes of adjacent triangular planes with overlapping one side differ. In order to eliminate the difference from the calculated luminance data obtained, the larger triangle plane is further divided into triangle planes, and the process of making adjacent triangle planes continuous is performed.
A digital image enlargement interpolation method in a digital image processing apparatus , wherein a gap between pixels generated when enlarging a digital image is interpolated with pixels having position data and luminance data obtained from the triangular plane. The triangle is a right triangle;
A number of pixels (where n is an integer equal to or greater than 0) are located on both sides of the right angle, with 1 added to the nth power of 2.
The arbitrary pixel is arranged at least at the position of the vertex of the triangle outside the periphery of the digital image when the number of pixels arranged on the side in the peripheral portion of the digital image is less than the number obtained by the above formula. An enlargement interpolation method of the digital image described.   The digital image enlargement interpolation method according to claim 2, wherein the number of pixels located on both sides sandwiching a right angle of the triangle is the same. A digital image compression method for compressing a digital image represented by a collection of pixels arranged at a fixed interval by a digital image processing apparatus,
A digital image is virtually divided according to the following conditions (1) to (3) with a plurality of triangles each having three vertices located on a pixel,
(1) Triangles do not overlap.
(2) All the pixels within the compression target range are located inside or on the side of any triangle.
(3) A triangular plane is obtained from the three dimensions of the position data and luminance data of the three pixels located at the three vertices of each triangle, and the actual luminance data of each pixel located inside or on the corresponding triangle; Comparing calculated luminance data obtained from the actual position data of the same pixel using the triangular plane, the difference between them is within a predetermined range, or the result of statistically processing the difference is a predetermined range. Be within.
In addition, the actual luminance data at the vertex on the overlapping side of the smaller triangle plane, and the position data of that vertex are substituted into the larger triangle plane, which occurs when the sizes of adjacent triangular planes with overlapping one side differ. In order to eliminate the difference from the calculated luminance data obtained, the larger triangle plane is further divided into triangle planes, and the process of making adjacent triangle planes continuous is performed.
A method for compressing a digital image in a digital image processing apparatus, wherein only pixels located at the three vertices of each triangle are extracted. The triangle is a right triangle;
A number of pixels (where n is an integer equal to or greater than 0) are located on both sides of the right angle, with 1 added to the nth power of 2.
The arbitrary pixel is arranged at least at the position of the vertex of the triangle outside the periphery of the digital image when the number of pixels arranged on the edge in the peripheral portion of the digital image is less than the number obtained by the above formula. A method for compressing a digital image as described.   The digital image compression method according to claim 5, wherein the number of pixels located on both sides sandwiching the right angle of the triangle is the same. A digital image restoration method for restoring an image compressed by the method according to claim 4 using a digital image processing apparatus,
A triangular plane is obtained from the three dimensions of the position data and luminance data of the three pixels constituting the triangle,
A method for restoring a digital image in a digital image processing apparatus , characterized in that a pixel having position data and luminance data obtained from the triangle plane is a pixel that should be located inside or on a side other than the vertex of the triangle. A recording medium on which an enlargement interpolation program for causing a computer of a digital image processing apparatus to execute a method of enlarging and interpolating a digital image expressed by a collection of pixels arranged at a constant interval is recorded,
The program is
Virtually dividing a digital image according to the following conditions (1) to (3) with a plurality of triangles each having three vertices located on a pixel;
(1) Triangles do not overlap.
(2) All the pixels within the enlargement target range are located inside or on the side of any triangle.
(3) A triangular plane is obtained from the three dimensions of the position data and luminance data of the three pixels located at the three vertices of each triangle, and the actual luminance data of each pixel located inside or on the corresponding triangle; Comparing calculated luminance data obtained from the actual position data of the same pixel using the triangular plane, the difference between them is within a predetermined range, or the result of statistically processing the difference is a predetermined range. Be within.
Substitute the actual luminance data at the vertex on the overlapping side of the smaller triangle plane and the position data of that vertex into the larger triangular plane, which occurs when the sizes of adjacent triangular planes with overlapping one side are different Dividing the larger triangular plane further into triangular planes to eliminate the difference from the calculated luminance data obtained, and continuing adjacent triangular planes;
A recording medium for causing a computer to execute a step of interpolating a gap between pixels generated when enlarging a digital image with pixels having position data and luminance data obtained from the triangular plane. The triangle is a right triangle;
A number of pixels (where n is an integer equal to or greater than 0) are located on both sides of the right angle, with 1 added to the nth power of 2.
The program, when pixels are arranged on the sides in the peripheral portion of the digital image is less than the number obtained by the equation, to place an arbitrary pixel in at least a triangle position of the vertex of the periphery outside of the digital image The recording medium according to claim 8, wherein the step is executed by a computer . The recording medium according to claim 9, wherein the number of pixels located on both sides sandwiching the right angle of the triangle is the same. A recording medium on which a compression program for causing a computer of a digital image processing apparatus to execute a compression method of a digital image expressed by a set of pixels arranged at a constant interval is recorded,
The program is
Virtually dividing a digital image according to the following conditions (1) to (3) with a plurality of triangles each having three vertices located on a pixel;
(1) Triangles do not overlap.
(2) All the pixels within the compression target range are located inside or on the side of any triangle.
(3) A triangular plane is obtained from the three dimensions of the position data and luminance data of the three pixels located at the three vertices of each triangle, and the actual luminance data of each pixel located inside or on the corresponding triangle; Comparing calculated luminance data obtained from the actual position data of the same pixel using the triangular plane, the difference between them is within a predetermined range, or the result of statistically processing the difference is a predetermined range. Be within.
Substitute the actual luminance data at the vertex on the overlapping side of the smaller triangle plane and the position data of that vertex into the larger triangular plane, which occurs when the sizes of adjacent triangular planes with overlapping one side are different Dividing the larger triangular plane further into triangular planes to eliminate the difference from the calculated luminance data obtained, and continuing adjacent triangular planes;
A recording medium that causes a computer to execute only the step of extracting pixels located at the three vertices of each triangle. The triangle is a right triangle;
A number of pixels (where n is an integer equal to or greater than 0) are located on both sides of the right angle, with 1 added to the nth power of 2.
When the number of pixels arranged on the side in the peripheral part of the digital image is less than the number obtained by the above formula, the program arranges an arbitrary pixel at least at the position of the vertex of the triangle outside the peripheral part of the digital image. The recording medium according to claim 11, which causes a computer to execute the steps. The recording medium according to claim 12, wherein the number of pixels located on both sides sandwiching the right angle of the triangle is the same. A recording medium on which a restoration program for causing a computer of a digital image processing apparatus to execute the restoration method of an image compressed by the method according to claim 4,
The program is
For the compressed image,
Obtaining a triangular plane from the three dimensions of position data and luminance data each of the three pixels constituting the triangle;
A recording medium for causing a computer to execute a step of setting a pixel having position data and luminance data obtained from the triangle plane to be located inside or on a side other than the vertex of the triangle.