TW200904206A - Improved RGB images scaling or resizing method - Google Patents

Abstract

An improved RGB images scaling or resizing method is used to solve low performance and high distortion problems during the scaling/resizing process. By transferring a RGB format image into a YUV format image, the method uses integer arithmetic instead of floating point operation to complete the transferring calculation. To reduce the distortion, a high level interpolation process is applied to the Y weight of interpolation points of the YUV image when scaling/resizing. Finally, the method reverts the scaling/resizing YUV image to a scaling/resizing RGB image so that a real and accurate scaling/resizing image is acquired with high performance.

Description

200904206 IX. Description of the Invention: [Technical Field] The present invention relates to a RGB image miscellaneous method, and a special image is a method for scaling a RGB image by converting an RGB image into a lent image. [Prior Art] In the technical field of image processing, 'the most commonly used image is the "image scaling" technology, the main purpose of which is to convert the acquired image to the actual desired image size after resampling. Such techniques are often applied to the processing of many medical, multimedia and digital materials. Among them, the most commonly used image scaling method is the RGB image scaling method provided for the RGB image with the highest usage rate. Please refer to the flowchart of the S-RGB picture scaling method in the "1st picture". It can be seen that the conventional rgb image scaling method is read after the RGB f-image and the zoom parameter value to be reduced (steps 10 and 20). ), directly calculating the re-sampling of the RGB image, that is, the side-calculation to calculate the new coordinate position of each pixel in the RGB-scaled image generated by the RGB image after scaling according to the scaling parameter value (step 30), and then directly Interpolation of each interpolation point is performed by the RGB components of each newly obtained coordinate point of the coordinate position (step 4〇), and the interpolation nose is usually directly interpolated by the R, G, and B components of the nearest pixel of the interpolation point. The RGB component of the point 'After all the interpolation points have been subjected to the interpolation operation, an RGB zoom image generated by scaling according to the scaling parameter value is obtained, and finally the RGB zoom image is output (step 5〇).

Although this conventional technique is commonly used, the RGB 200904206 scaled image generated by the image is often caused by a phenomenon that the edge is not smooth enough or the boundary is not clear enough to cause image distortion; and because the operation is full due to the entire zooming process The floating point operation (float p〇int 〇perati〇n) process also causes the computational efficiency of image scaling to be greatly affected, resulting in a problem of inefficient processing. Therefore, how to improve the existing RGB image scaling method, in particular, to improve the miscellaneous problems generated by the image and the large number of operations generated during the mis-scaling process is the subject of efforts. SUMMARY OF THE INVENTION In view of the image distortion caused by the conventional RGB image scaling method and a large number of computing techniques, the purpose of the present invention is to provide a RGB image scaling method that utilizes image format conversion, takes integer operations, and enhances partial interpolation points. The technical means of interpolation operation solves the technical problems existing in the prior art. The RGB 彡 彡 财 整个 改 , , , , , 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改 改(4) Then, according to the scaling parameter value, the new YUV scaling generated by the zooming of each pixel in the YUV image is calculated, and the pixel position of each pixel is frustrated, in particular, the entire operation is simplified to record the money; (d After the zoom image of the lining touch, the interpolation of the TOV component is performed for each interpolation point of the YUV zooming shawl. The main Wei is divided into the following two steps: (4) Y 关联 关联 细 细 细 细 细 细 细 细The unicorn value operation; (4) fetching the U and V components of each associated pixel to directly insert each interpolation point, and then convert the scaled image to the lion zoom image, and output the job zoom 200904206 image. Among them, the integer calculation method is mainly for calculating the coordinate position of each pixel in the newly zoomed image, because the scaling parameter value often generates a floating point number after substituting the arithmetic expression, so the method of the present invention is actually operated. When the available integers m and η are expressed as follows, the coordinate position (x〇, y〇) is first calculated by m and then divided by ιοη, so that the whole operation is performed in an integer manner to greatly reduce excessive floating point. Computation. Through the improved RGB county job method of the present invention, it is indeed possible to achieve the technical effect of scaling the image and improving the efficiency of the image, and can be closer to the real zoomed image. For the characteristics and implementation of this issue, the following is a detailed description of the implementation of the Hanghe (4). [Embodiment]

The improved RGB image zooming method proposed by the present invention provides a solution for the problem of image distortion caused by excessive operation and image scaling when the image is zoomed in. In the prior art, the improved image zooming method proposed by the present invention can be softened by a computer. The _form is implemented on any computer-executable image processing platform (eg digital TV, computer). "The so-called RGB image, which is widely used in multimedia display, electric surface non-image and television (10) image format, the principle of adding and mixing the three primary colors of the red, green, and blue The second primary color electron beam is emitted without intensity to generate the RGB image of the desired color; and the YUV image is displayed as the other _ brain display image and the television 35 is dry (four) ^ is lang in the evening media display, electricity The image format on the page is mainly based on the human eye 200904206. The visual sensitivity to brightness is greater than the sensitivity of the color neon. The image is changed to the redundancy component Y and the color difference components U, v to indicate 'the characteristic is γ The component system is separated from the U and V components independently. Since the RGB image is zoomed in, it is necessary to perform operations on RG and B in the image at the same time, and a large number of floating point operations are often performed during the operation. S is added in processing efficiency; plus RGB The composition and structure of the image itself are mixed, which causes the image distortion problem (color distortion, image blurring...) generated by the RGB shadow county to be particularly serious. Therefore, the present invention adopts the problem that the RGB image is used for image format conversion, integer and enhanced part of the value point interpolation operation to improve the problem of scaling the image in the biliary image, which is further explained below:

The conventional RGB image scaling method has been described in the prior art (refer to "FIG. 1"), and the reason for the main occurrence of the missing is not described here. Must be a problem to be solved in the prior art. Please refer to the section of Figure 2 for the improved RGB image scaling method. First, read the image to be fined (step 100). The method of reading can be done through the design of computer software. - 胄 胄 « « : 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 : :

In order to do the butterfly _GG), the miscellaneous change method is the Y, u, ") of the R and G images of each of the RGB recordings of the nose-type secrets.

σ 5] ο 0.299 587 0.114 -0.148 0.615 -0.289 -0.515 0.437 -0.100 After the image format conversion is completed, the scaling parameter value is received (step 3〇〇) 'The receiving of the scaling parameter value can be used by the computer software design. The user interface (UI) is implemented on the image processing platform for random input from the outside. Basically, the scaling parameter value is the multiple value of the image to be scaled, and the positive or negative integer can be used to indicate the magnification or zoom. Multiple (such as: 3 means 3 times magnification, -2 means 2 times reduction), but the input and performance are not limited to this. In addition, the scaling parameter value of step 300 in the process of computer software implementation can also be mentioned as input setting before step 100. The invention is not limited thereto.

After obtaining the scaling parameter value, 'the actual zooming process of the YUV image is started. Since the image scaling will inevitably cause the coordinate position f where the original pixel is located to shift, the _ pixel point coordinate position appears, and the county pixel position is located. Also due to the occurrence of offsets - some new interpolation points need to use the associated pixel points to interpolate (interp〇lati〇n) to get the erg component of their pixel points, so the _ first-domain number calculation is difficult to calculate After the parameter value is scaled, the position of the pixel point of each touch of the YUV occupational county towel is generated (step 400). The calculation method is also based on the original pixel coordinate position fe, yi and the zoomed record value (4) 0) Then calculate the new pixel point coordinate position (XG, y.) in the scaled image, and its expression can be expressed as follows: 200904206 Ο Value transmission / idea 疋 'Because 1 / in the algorithm will often appear floating point The numerical result of the number 'in the operation is often expected to cause a huge burden on the computing resources, especially in the series of image conversion operations, the image processing efficiency is very large, In particular, the present invention utilizes an integer arithmetic method to improve the efficiency of a large number of floating point operations (step 41). The integer arithmetic logic is as follows. When it is determined that the value of the arithmetic towel is a floating point number $' When the floating point number can be expressed as m*1〇n by the integers m and n, then the new coordinate position (x〇, y〇) is calculated by 111 to 3 in the tool, and then the obtained coordinates are obtained. By dividing the position (x0, y0) by 10n, you can still get the correct new pixel coordinate position (χ_, you can even reduce the conversion coding error probability through the recording operation method) to produce more accurate image capture results. When the parameter value calculates a new νυν zoom image where the new pixel is located, the _ can also be compared with the comparison of the (10) image and the YUV zoom image to obtain the coordinate of each interpolation point corresponding to the interpolation calculation. Position, this is a conventional technique. There is no more information here. With the monthly interpolation point, the next step is to calculate the YUV component for each interpolation point of the zoom image towel, in order to reduce the image distortion = calculation Volume, this _· scale The special part of the calculation is divided into two parts (ie, "Fig. 2" 500 and step 550), which are explained as follows: Step 500: It is easier to zoom in on the image. In order to interpret the knives for the human eye, in other words, the method of the present invention takes a more _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The distortion improvement effect of the 200904206 for the overall TOV zoom image is most significant. Therefore, the method of the present invention extracts the γ component of each associated pixel to perform high-order interpolation operations for each interpolation point. (2) Step 550: Scale the image for γυν It is less easy to interpolate the U and V components of the human eye. In other words, the method of the present invention has little effect on the distortion of the YUV-scaled image on the UV component, and is easy to cause a large amount of computation, so UV The component is calculated by the direct interpolation algorithm, which has little effect on the distortion problem of the overall YUV scaled image but can improve the overall computational efficiency. For this reason, the present financing method takes the U and V components of each associated pixel to perform direct interpolation calculation of each interpolation point. The high-order interpolation operation mentioned in the present invention can basically perform some γ-component calculations by using some conventional interpolation algorithms, but at least it must be an interpolation algorithm of a first-order or more. For example: bilinear interpolation algorithm (bmne batch interpolation), bicubic interpolation algorithm (bicubicinterp〇lati〇n); as for direct interpolation operation, the pixel points adjacent to the interpolation point are directly associated pixels, and The u and V components of the associated pixel are used as the U and V of each interpolation point, and the operation is usually performed with a single associated pixel closest to the interpolation point. 'This operation is most efficient (but not as limit). As for the associated pixel points, there are different associated pixel points for calculating the components of the interpolation point, depending on the interpolation algorithm used. Generally, the pixels adjacent to the interpolation point (which can be the pixels in the YUV image before zooming or the new pixels in the zoomed YUV image) are used as reference, although in the more advanced interpolation algorithm, Complex operations to determine the associated pixel points to be interpolated by the interpolation point; and in some high-order interpolation operations, the weighting coefficients of different points can be given according to the distance relationship between the associated pixel points and the interpolation points. It is possible to calculate the gamma split more accurately. There is no limit to the way the decision is made for the associated recorded points. 卞The weight coefficient is hereby treated only from the third map of the factory to the third figure. Zooming in on the image when you zoom in 2 times in the horizontal direction (the same as when the YUV image is zoomed out):

First, the 3rd circle "H" shows that when the scene green is to be horizontally magnified 2 times, the original pixel points of the YUV image (P1 and _〇 will be laterally slanted to the left and right sides respectively. YUV position (4) (4) New pixel points (nPl and nP2) 810 The calculation method of the new coordinate position has been described above, and the original original pixel points (ρι and p2) are shifted laterally after the 'transmission of the YUV. The degree of distortion needs to be interpolated at the interpolation point between the new pixel point of the YUV zoom image (4) and the two new pixel points of the milk (ie, the first interpolation point 811 ιΡΙ and the second interpolation point 8l2 “iP2” shown in the figure. The coordinate position of the interpolation point determines the coordinate position of the YUV image and the YUV zoom image before and after the zoom, and the coordinates of the coordinates of the interpolation points corresponding to the interpolation calculation are obtained, and the description is not repeated. The method is to treat the γ component and the uv component in the interpolation point differently. Please refer to the section “3B” for a high-order interpolation operation on the γ component. The Y for the first interpolation point 811 is known from the figure. Component (iYl), taken The pixel points adjacent to the interpolation point are regarded as the associated pixel points (in this case, nPl and nP2), and their respective γ components are calculated, and the high-order interpolation operation adopted can be iYl=2*Yl/3. +Y2/3 indicates that the Y component of the associated pixel for different distances from the interpolation point is also suitable for 12200904206. When different weight coefficients are given, the Y component (iY2) of the second interpolation point 812 is also used, and interpolation points are also used. The pixels are calculated as the associated pixels (in this case, milk and nP2) by taking their respective γ components, but the difference is due to the difference in the associated pixel points of adjacent interpolation points, so the weight coefficient is given. It is also different, but instead is represented by iY2=Y1/3 + 2*Υ2/3. As for the “3C figure” section, it is a schematic diagram of direct interpolation of the UV component, basically due to the influence of UV component on image distortion. It is not high, so the direct interpolation method that does not require extra 〇 operation is used to process the υν component required for the interpolation point. The UV component (im, iV1) of the first interpolation point 811 and the UV component of the second interpolation point 812 are known from the figure. (iU2, iV2) are gambling points adjacent The pixel is used as the associated pixel. In the present embodiment, the direct interpolation is performed with the associated pixel closest to the interpolation point. Therefore, the UV component (iU1, iVl) of the first interpolation point 811 can be iU1=U1 and iVl= V1 is used to indicate 'the υ γ component of the second interpolation point 812 (iU2, iV2) can be represented by iU2=U2 and iV2=V2. Finally 'when each pixel in the YUV scaled image completes step 500 and ϋ step 550 After the program, a complete γυν zoom image can be generated. Finally, the YUV zoom image is converted to rgb format, a so-called RGB zoom image is generated (step 600), and then the RGB zoom image is output (step 700). The flow of the inventive method. In the step 600, the yuv zoom image is converted into an RGB zoom image, and the conversion manner is performed by converting the γ, u, and V components (/rt/(7) in each pixel of the YUV zoom image into corresponding ones through the following operation formula. R, G, B components of RGB scaled image: 200904206 1 1 1 [RG 5] = [Γ u V] 0 -0.39 2.03 1.14 -0.58 0 Due to the improved RGB image scaling method proposed by the present invention, particular emphasis is placed on the image zooming process. In the image format conversion, the use of integer arithmetic and the use of enhanced interpolation of interpolation techniques, the technical problems existing in the prior art can be fully solved and the expected technical effects can be achieved. The embodiments are disclosed above, but are not intended to limit the invention, and those skilled in the art can make some modifications and refinements without departing from the spirit and scope of the invention. The scope of the patent application attached to this specification shall prevail. [Simplified description of the drawing] Figure 1 is a flow chart of the conventional RGB image scaling method. A modified RGB image scaling method flow chart. Fig. 3A is a schematic diagram of an improved RGB image scaling method applied to an enlarged image embodiment. Fig. 3B is a third embodiment of the improved RGB image scaling method for the Y component. Schematic diagram of fine interpolation operation. Fig. 3C is a schematic diagram of direct interpolation of UV components by applying the improved RGB image scaling method of Fig. 3A. [Main component symbol description] 800 yuv image original pixel point 810 yuv zoom image New pixel point 811 first interpolation 812 second interpolation 200904206 step 10 read RGB image step 20 receive scaling parameter value step 30 calculate each pixel in the RGB image generated by the pixel parameter according to the scaling parameter value Point coordinate position step 40 extracts R, G, B components for each interpolation point direct interpolation operation Step 50 Output RGB zoom image Step 100 Read RGB image Step 200 Convert RGB image to yuv image Step 300 Receive zoom parameter value Step Calculate YUV f彡 各 各 各 骑 骑 骑 骑 - - - - - - - - Y Y Y Y Y Y Y Y Y Y Step 410 converts the integer arithmetic is set in step 500 _γ components for each higher order interpolation calculating step of interpolation points taken 55G _ U, V components for each interpolated point computation step direct interpolation 60G _ Medical spacious heteroaryl is zoomed image deletion step 700 the RGB video scaling

Claims (1)

200904206 X. Patent application scope: h - an improved RGB image scaling method, comprising the following steps: (8) reading in an RGB image, converting the RGB image into an image; 0) receiving a scaling parameter value; (c) taking one The integer operation method calculates a coordinate position of each pixel in the YXJY zoom image generated by each pixel in the γυν image according to the scaling parameter value; (d) interpolating each pixel except the pixel in the YUV zoom image For the YUV component of the point, perform the following steps: (dl) extracting the Y component of each associated pixel from each pixel in the YUV scaled image for high-order interpolation operation of each interpolation point; and (d2) scaling the image from the YUV Each pixel captures the U and V components of each associated pixel to perform a direct interpolation of each of the interpolation points, and (e) converts the YUV scaled image into an RGB scaled image, and outputs the RGB scaled image. 2. The improved RGB image scaling method according to claim 1, wherein the step (a) is an R, G, and B component of each pixel in the RGB image by the following expression ([/? G5] Converted to the Y, U, V component of the YUV image ([Ft/P]): 200904206 -0.148 0.615 -0.289 -0.515 0.437 -0.100 3. The improved rgb image scaling method as described in the scope of the patent application , wherein the step (c) is based on the position of each pixel in the YUV image by the following operation formula (Xi, y tree computing _ scaling the coordinates of each pixel point in the image (know, y〇): {\! ratio 〇〇〇\iratio 0 〇0 1· I where is the scaling parameter value; and the integer operation is when the value of 1 / raizo is a floating point number and can be expressed as m* by the integers m and η L〇-n 'First calculate the coordinate position (Xg, y()) with m and divide by ι〇η. 0.299 〇 Β] 0.587 0.114 4. The improved RGB image scaling method according to claim 1, wherein the high-order interpolation operation refers to each pixel point adjacent to each interpolation point as an associated pixel point, and Y of the associated pixel points. The component performs at least one or more interpolation operations as the γ component of each interpolation point. 5. The improved RGB image scaling method of claim 4, wherein the at least one-order interpolation operation assigns weight coefficients to the associated pixels at a degree close to the interpolation point, and the weighted correlation The Y component of the pixel is interpolated. 6. The improved RGB image scaling method according to claim 5, wherein the at least one order interpolation operation is bilinear interpolation and bicubic interpolation (bicubic 17 200904206 interpolation) ). 7. The improved RGB image scaling method according to claim 1, wherein the direct interpolation operation refers to using adjacent pixel points as associated pixel points, and using the U and v components of the associated pixel points as U and V components of each interpolation point. 8. The improved RGB image scaling method of claim 7, wherein the direct interpolation operation interpolates with U and V components of the associated pixel closest to the interpolation point. 9. The improved RGB image scaling method as described in claim 1 wherein the step (e) is to scale the gamma, U, and V components of each pixel in the YUV by the following expression ([F ¢ / η) is converted to the R, 〇, and B components of the RGB scaled image ([and G5]): Γ 1 1 1 G = t/ Π 0 -0.39 2.03 . 1.14-0.58 0 18