TW201141212A - Video data processing system - Google Patents

Abstract

This invention relates to a video data processing system. The system includes a video interface supplying a serial of format of frames; and an embedded system receiving the serial of frames and performing the processes of (a) receiving a video including at least one frame; (b) defining a target video cube having a predetermined warping ratio, wherein the target video cube having the predetermined warping ratio includes a specific area; and (c) warping the at least one frame so that the processed at least one frame can conform to the aspect ratio of the target video cube.

Description

201141212 VI. Description of the Invention: [Technical Field] The present invention provides a film data processing system, which is a processing method implemented. In particular, the processing method is a processing method for a slice scaling and an application method thereof:应用 Application notes for related devices and circuits. It is the first technology of I. It is the development of pure technology, the change of the resolution of the picture and the zoom is an important topic, the two kinds of _ _ format, such as mobile phones, PDAs, wide-screen notebook computers and many more. For example, mobile phones, pDAs, etc., have a higher screen resolution due to the smaller screen and the different fish screens that are played. The current basis is based on the size of the target screen. Uniformly zoom or crop the left and right sides of the image and the movie to achieve the predetermined width and height ratio ^ ^ 疋 simple miscellaneous job into the image content distortion distortion, and cut and remove some important object information at the edge of the image. Therefore, in recent years, there has been learning | = content-based scaling technology, using the non-equal-scale distortion to = and the size format required by the shadow #, the scale of the face is more important than the shape of the inner valley, and will be distorted Distortion is hidden in less important areas. In recent years, content-aware images and video reconstruction have become more mature by using non-equal distortion or segmentation-based composition. Unfortunately, the former lacks image distortion. Hey, the latter's picture and the appearance of the clip's movie. However, re-scaling the movie is basically different from the still picture shrink, and can't completely solve the problem only rely on the present There are image scaling methods plus time limits. There are two problems with this problem: (1) object motion and time dynamics are the core elements of the film, and must be handled explicitly, before the method, re-scaling every frame in the movie Only along the time 201141212, it is impossible to process a movie with complex movements and there will be a lower face, (:) when the important object in the movie covers the entire picture, in this case 2 All of them are closed. In this regard, '疋 can't be achieved without removing or distorting the important content. For example, mobile phones, PDAs, this type of mobile farms are supported by small screens and pictures and videos. The scaling ratio may be no = no surface processing method and its application method 2 [Summary of the Invention] One of the ideas is to provide a film data processing system for realizing the processing method of the film zooming. The background information of the two t-two ratios is still under the requirements of different scales. The film is used in this case. The case is to discuss the fan brothers mentioned above. Technology (for example, this Lu's; seven, a case of shortcomings. U ^ shouted) often related to the scope of the access =:it device, system, computer order during the at least - region is subject to film recalibration, The program of the specific example can include ', the second part is based on the video data related to the frame. The receiving and at least one movie picture are configured by hardware, and the example is - the specific column (-and one) of 201141212 is: (1) and The content related to the information appearing in the - frame, and the next specific number of frames disappears, the next two = today = temple; the area related to 'and / or (9) contains the specific Ι Ϊ Ϊ = = oriented foreground The information of the object. The sample program can be protected and not protected. The specific area of the test area is more basic, and the sample program is associated with the parent-related pixel. The average traffic is dan, 任何 in any of the special boxes in the front box, whether the information appears, and can be maintained in any of the frames of the j-frames: ί: ί ::: . This information that failed the test can be marked. The entropy associated with the flow of the information of the previous 2:;::ΐ information is significantly shifted. The use of quantized traffic vectors and/or based on flow and progression includes specific rows and/or columns based on and over-predetermined gate insertions. School traffic entropy, to select the most likely _ function 敎fm can be the most relevant issue of the distribution of the plume distribution, in some embodiments, the sample program can further include a subroutine in the film cube, where the specific area This I(10)μ is performed in a target image. At least temporary - lang * equal scaling functions can be used to promote adjacent orders. The vertices of the ridges can be constrained, based on the least squares mosquito objective function and/or the vertices 7 including the recognizable grid 継 positions, wherein the grid 疋 pre-aligns the linear combination of the grid vertices of the adjacent regions. Changed to 201141212, including the unequal scale subroutine, a grid can be used, and the bean can be packed with a bite of a bit, wherein the particular squares are at least one further at least on the time axis adjacent to the grid = the same. Moreover, you can use - zoom = and / or slide two windows n example of the subroutine can be used - pixel level (four) paste can configuration ΐ, exemplify the process computer access media and system is also here - instant Process image # material to process the unequal scale in the film area. For example, the 2 stipulations include: 'receiving film material, which is generated by the map of Lianba to domain 1 and/or by = specific number = === information on moving foreground objects. In this example = step, it will be clearer that the dome is shaken to perform an unequal scale and the protected cube is modified. Tian Yuejing exemplifies the following examples: "The invention provides a film data processing system" to receive the series and execute (4) pure inclusion = 201141212 a film; (b) define the inclusion - the present a cube; (6) a predetermined scale of the map drag area, the target movie, etc., the frame conforms to the target, so that the processed image is displayed. The size ratio of the body to output another series system: ί ί: The present invention provides a kind of film data processing - format of the series to receive the three-dimensional image containing the - the first example and (c) the surface of the series

Three small, after outputting a second format, the second quarter: including ί 7 purpose 'The present invention provides a video data processing system, performing the following steps: (4) accepting a target image with a = ^ type a plurality of frame two-dimensional scales, the target image is imaged, and = two is a u scale, the single (four) is the time interval of each phase in the complex side frame; and (8) the target image is unequally retracted into the rectangle In the body, the movie has a second format for display. [Embodiment] First, there are different requirements for scaling a movie to be played on a mobile phone, a PDA, or a video having a smaller screen to a movie having at least one frame, and an important image object is defined first. For a movie to be scaled, first define the important image objects contained in a specific area of each frame, such as moving foreground objects, and how to avoid being cropped. The method is as follows: 1. Calculate the visual flow (optical fl〇w) of each frame, and after analysis and quantification, 'get the flow vector of each element in the frame (fl〇wvect() I^. 2. All traffic will be The vector is imported into a fan chart to perform 201141212 • number statistics, and the probability of distribution of the numbers is introduced into the entropy formula (entropy), to obtain the entropy information of each row of at least one frame. 3. Using each entropy of all the frames to find a critical region that does not allow cropping. 〇4, combining the cropping and unequal scaling to optimize the operation, so that the frame is rescaled, in accordance with Target image cube (target shirt cube) = size ratio. In other words, in each frame by defining a specific area, the content of each frame in this area can never be deleted. Use visual traffic to determine its criteria and calculate Content outside the critical area of the target movie cube table containing all specific areas can be discarded. Specifically, when you want to change a piece, the unique pixel of the line is removed, and you want to change the film to ^ Let the specific column be preserved. Take the practice of narrowing the movie as an example: (1) When the content just appears in the frame or is about to disappear in the next box, this content does not have the characteristics that continue to appear on the timeline (2) The specific area must contain significant moving foreground objects and define the leftmost and rightmost domain pillows of the fixed area _iealedumn). Please refer to Fig. 1, which is a view of the method of the first embodiment of the present invention. The flow of the method of the present invention is as follows: Step 10: Receive a movie containing at least one frame; Step 11: Find a specific area containing a moving foreground associated with at least the frame. Step 12: Defining a predetermined scale target slice cube containing one of the specific regions, wherein the predetermined zoom ratio is determined by an artificial one.

Step 13: The quantification process includes importing visual traffic of at least one frame into a statistical graph for number statistics, and introducing the probability of distribution of the numbers into the formula to obtain at least - although at least - the entropy information domain of the row It is determined based on entropy information. A special 疋S. For the movie to be scaled, it is necessary to define the image object that is important in the moving area object in one of the special areas of each frame. The method is: ι: calculate the visual flow (optical fl〇w) of each frame, after analysis and quantification, obtain the flow vector of each element in the frame (fl〇w vec (4). 2: Guide all flow to f The human-fan statistic _ ehart is used to perform the numbering statistics and to introduce the probability of distribution of these numbers into the entropy formula ^ntropy) ' to obtain the entropy information for each line of the frame. & uses every per line entropy of all frames to find a specific area (4) that does not allow cropping. 4·· Combine cropping with unequal scaling for optimization,

The box record is 5, which is in line with the target miscellaneous _arget vide. Cube) ϋ, proportion. The change &' in each frame defines a specific area, and each: ?= this content must not be deleted. _ Sense flow comes to the movie cube containing all the specific areas; the pixels representing the specific line are down by mi. Take the practice of narrowing the film as an example: Time; in the frame or the frame that is about to disappear on the lower side, then there is no temperament on the timeline. This =) special area must contain significant field = = the right side is a specific line (: =: frame tii amount into 7 = amount can be □ to know, the shadow content is in the next amount, and test whether it is The flow of each line of pixels is seen in the frame after the frame (using the k吟= box and will keep & and will keep the line marked in the specific "2 2 pieces are not established" then the movement of this object must be independent Second; the significant object, the table's per-line material is quantified (4), where the C generation (the micro-flow vector suppresses the negative 蕙 will be quantized to the regional lattice with more compartments). The quantity ^ message 'does not need to distinguish more quantization between a and 1 to 乂 via the quantized integer value as follows 201141212 %(1): /(fi)=2fc+ with fc= L〇.5£(f,)J. Willow and θ (/3 represent the length and direction of /, respectively, the principle of this formula is wide: this fan shape is composed of many concentric circles of the same pitch, its k ^ half (four) 2 (k + l) 'city k ring The area is divided into 2ic aliquots, and each

The angle of 3 i is 2π/2*; as shown in Fig. 2, all the compartments are consecutively numbered from the innermost =. Place a flow vector in the original of this pie chart! 1 Use Eq. (l) to find the interval number in which the vector is located. In detail, fc = L〇.5 is the first interval in the ring of a certain fiV (2?r/2fc) ring whose flow vector is located in the first ring of the fan graph. ^, the entropy function (entr〇py). In this patent, the leftmost and rightmost boundary of the miscellaneous domain is found in the information theory; the entropy function is as P={ Pl'P2'Ρ3,"·,ΡΠ } ' then the network function H(C) = _^p(f.).1〇g2p(f.) iec two=ίίΛ several important characteristics, which are - the value of the function must be greater than zero, and if the number of events in ί ^ ^ ^, the smoke function is like 0 secret. When the use of J ί ί is established, the system s is the largest. At this time, the main reason is that 'when the probability of each event is equal, the θ 翻 is the largest. According to the visual flow vector, when the probability distribution to the ratio is concentrated in the range, it represents the flow. The histogram of the statistics and the definition of the flow distribution probability (height is] the histogram reads all integer histograms for normalization), so that you can get the 201141212 parent function C (entr〇Py of column C). In the PiUM system, it is set that when the flow entropy function of a certain line exceeds 〇.7, the current probability is within a specific area; and I is the maximum entropy function value when each event is Ζί. Figure 3 (8) and 3 _ are in the ',' and the leftmost and rightmost boundary of the simplified area is indicated by ®. In this system, the specific F = the boundary of the specific area that must be reserved for reservation 'but does not mean that the content must be completely cut according to the content; the initial cropping area must be read and scaled and the time axis is maintained. Limit 'and the scene two to see with the axis S. Because;: For _ before the very clear guidelines. , because 'the flow entropy adopted by the system is a _ cage to scale the axis frame unequally, so that the processed Ϊ 4 L two f δ - the size ratio of the target film cube ' as shown in Figure 4 (8) and ° wherein 'the surface 14 further includes at least one optimized isi person and the like to perform unequal scaling, so that the processed size of the image cubes is proportional, and the optimization formula is based on Designed for the inner valley and time consistency. The film re-scaling architecture proposed by 瞀 is based on the unequal proportion of variation-optimized privilege, and the domain cut includes optimal optimization. Hunting uses the grid grid structure to separate the film cube area, set S3, the function calculates the vertex position of the grid: under a few conditions, the ϋ 到 、 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量 能量= Insert repositioning to produce a female movie. The objective function consists of four pieces, the content of the flipping time and the visually important content of the space, and the coherence of the object content on the time axis. Derogatory mark symbol. In each frame of the film, the grid frame Mt==other {v=tl, v=i2..., v=tn} is the vector position collection, and E and ^ knife represent the edge and square plane (_ Face), and all the frames are connected to each other. The new vertex after scaling deformation = (χί,ι,; 201141212 These variables use the optimization calculation process. To simplify the complexity, the symbol t is reduced to Vi 'when the previous frame vertex' is represented by v'i. Reduced movie size (rx'ry'rz) ' 1^ and ry are the resolution of the released movie. The total number of frames. Basically, compress an input movie cube to a predetermined scale target movie cube. The optimization formula includes a similar energy for maintaining spatial content (^^formal energy), a Temporal Coherence energy for temporal consistency on the time axis, and a frame for each frame after cutting. Smoothed secondary smoothing energy (second_order smo〇thing energy)

By using an iterative function (iterative minimization to solve a minimally flattened problem to obtain - group optimized silk.) 'In step 14 towel' at least - Figure (4) based on the exposure structure, not to scale 4 (the unequal ratio The zooming system uses the frame to slide in a long and smooth manner, and the (10) portion is less important. At least the geometric unit size i is lower than the following. This is a better description of the embodiment. Originally, the vertex coordinates of the corner Luo must be in the new resolution movie. In this example, all the specific areas are deformed, and then the vd square is outside the 'so--is cut off. i The slave left and the corner of the corner _. For the second = 丄 included here must meet the following conditions as Eq. (3 ^ M in the eye cube, must 201141212 to maintain the object, the time axis has a domain, the design is different Restriction., b) the object in the cutting area of the consistency of the axis, (conformal energy).

Let Vi丨, Vi2, Vi3, and vi4 respectively represent the similarity of the transformation. The parametric transformation is determined by the _number q of the four _points and the expected q and q, to achieve the best similarity...

«I p [s, r. u. v]q^qi =: argmin s —r s,r,*u,v 7· S 4 r

• VL

JL This is a linear least squares problem, which can be written as [s'r'M9,9'=«49) Wb,' 'which is as follows. (5) ^ίχ -Vh 1 0 - "4 1 Vh xh 0 1 Yf Aq = I l ; ; , bgt = —yt4 1 0 -Vi4 ®»4 0 1 _ -vC . The matrix Aq element depends on the grid structure, and % is a matrix of unknowns. k CW,V]q , q', will be able to get wide 1 from the square q <-/) v can be written as follows Eq. (6;) A ((10)') = ΣΣΚ Intestine V), , is the visual importance of the t-th frame The weight value of the square, and the spatial importance map of each frame, combined with the color gradient measurement 'image visual feature map, and accurate face detection, the important information map 12 201141212 ^ normalized After the shed is [〇], (four), the important area of the shirt is excessively reduced to a significant distortion phenomenon; the unequal scaling method is approximately = (KRAHENBUHL, et al. 2009) and (WANQ et al. 2〇〇9). The grid line is strongly distorted. As follows, Eq.(7) " α = Σ“ Σ{ί, peach K)2+ (:Vi -Vj)2 )·

Ev*Eh represents the vertical and horizontal edges in the grid architecture, respectively. In order to achieve the time axis-induced shadow; = {Zoom, this special place β qi holds the object motion information, and minimizes the distortion of the film _, trembling, wave ϊ ί. By visual flow, etc., to the case of the lower-side frame, to! ^ indicates that the square of the waiter is taken, to the best fit, the linear conversion work ___ drag and drop the example does not need to calculate the ί change of the package I even inside, because only need to include the outer shape of the transform Stealing the position of J. The purpose of this patent is to zoom in with the film: J, the sex conversion, and the following energy function is formulated accordingly. For example, ^ maintain the above line (3⁄4) = 卜 tq. (8) ^ The energy function already contains the camera lens movement. The only thing left is the individual movement, and the unknown, like the vertices of the grid. For the correct definition of the vertices for the linear combination, the nearest neighbor is analyzed. In this shot, 3 is based on the linear combination of the vertices of the visual stream and the upper vertices of u, denoted by ~. As follows Eq. (9)

Vi' redefining 'financial can use 201141212 仏 (must=Σ 哝(<-<)-(11;+1/-〇||2, ' Ε(^) is a collection of squares q/. Some cases Next, the square Q% has moved to the outside of the movie frame. For these squares, use the following Eq.(ll) to simplify the time scaling, and the adjacent squares on the timeline can maintain similar restrictions as much as possible (must = Σ 丨丨«-<) - (V5+1' - 〇||2. 0';fc)6E(9*) ·

• All similarities on the time axis are Eq.(12) r 9:eQ; in Eq.(ll) this energy is used to localize the limit, to maintain the phase on the time axis, and the object has a -generational forest. The resulting error will accumulate between frames. In order to solve this problem, the method is proposed to both control the corresponding squares on the time axis and to reduce the degree of error. Clearly and anciently, the two qi and its corresponding square f replace the original λ=5. Because in the movie, the shape of the object: Π ίί obvious, especially when there is an object or camera moving Hi allows for slight inconsistency when zooming. The energy proposed is only discussed how to adjust the zoom. The position is changed between adjacent frames to maintain smoothness. Therefore, it is a second-order smoothing term. As follows: Hungary (1曰3) Baiqian

De = ||2v〇— (v〇~1/ + v〇+1/)||2. 〇t 'In the formula, n represents the number of grids. This equilibrium energy number 201141212 = function and other functions Different, it makes the ride a grid instead of a single top at 3 = wants an optimal solution, so that the deformation grid structure distortion reaches D ^ Dc + De + jDt + SDs, in this system, 7 = 10, The =1 s bound limit inequality is to protect the specific and non-negative side line length conversion to this inequality, requiring the grid edge line not to be a phenomenon. Straight boundary limits can be maintained straight by using the linear equation 2 boundary, which uses the top of each frame in the calculation: optimization process, in which some linear constraints and linear conditions are in the real, each time a new limit is generated Wait for the silk as the basis for the next entry into the lamp = generation. Please note that on the 14th side, whenever the conditional limit is changed, the system matrix will also change, depending on the opposite of the inequality. The system of this patent, according to [Buat〇isetal.2〇〇9] this research, GPU (GmPhiCS Process υώ) conjugate gradient solver (c〇njugate grafemsolver) and multiple parallel processing as the core architecture, compared to the general A single-time solver, which will require a lot of credit space, but also significantly reduces the time required for the operation. Once the deformed mesh _ has been calculated, the scaled movie is clipped from the specified target zoom cube. In addition, the picture of each frame is finely placed to the parent square, the system uses linear interpolation, or the EWA surface splatting of the solid imaging method proposed by Tao hen_ et al 2009] The specific example program disclosed herein performs a 15 201141212 test on a desktop PC with a graphics card of Duo 2.33 GHz #m GTX 285. Example 1 2 Processing ^ and Nvidia Reference π) The method described above = = ^^^ and _ (please become Xu Xi, the scene of the scene is the scene when the scene is converted to the frame of the frame ~ 曰Because of the sex. When the *4 should be considered when the _ is similar, j this dish slightly improved the performance and the noise of the shop is based on the number of unknown vertices. Zeng Yue _# Cheng Qianfang growth. After scaling, the σσ quality of the movie and the green 'in the touch +, usually made every square

The grid structure, there will be more in the future - step by step, the new Wei pushes the case 2 use ί heavy grid (sense secluded strategy, ΐΐϋS can. 'Improve the system efficiency in the grid distortion to achieve an average of 6 sides per second The box 'can be more in the place = the shirt month. The type of it's fan marriage silk green 3 (10) rationality; SV film automatically generated according to the specific example of the road, the algorithm is used before the use Parameter setting, in some special cases, wood ^: system: to allow the user to manually determine the important objects in the film; in a certain frame 'to visually flow the object to the graphics #K h visual flow automatically extended To other frames. ^ Programs and linear scaling in accordance with the specific examples of the present disclosure, and motion-based/film scaling (MAR) programs, as described by, for example, Wang et al. (see Resources 26). For example, take a pool enb he 1 et al. (please refer to the 1 time edited movie scaling (SVR) program for comparison]. Since both MAR and SVR programs are currently the most advanced technology, this will be The results were compared with the MAR and SVR procedures. Previous studies [w〇if too乩2007], [Rubinstein et al. 2008], [Zhang et al. 2008] did not consider the consistency of the action object on the time axis when scaling the film. ^16] 201141212 This ^ read no silk beauty Taking action as the main method, [Kr^henbtihl et al. 2009] * ^pan, user survey as a basis to draw this conclusion. Interesting is the figure ^ et al. 2009] > [Rubinstein et al. 2009 ]^# 3 cropping and other techniques to get the most similar results to the original image in an optimized way, but these methods require a long calculation time and are only suitable for still images, and do not extend to the film, consider The problem of consistency on the time axis. - As a comparative literature, it is mainly compared with [Wang et al 2 (8) 9], because this research paper clearly deals with the problem of the one on the time axis. However, it is necessary to use the SIFT method to perform the alignment of the feature points of each frame. If the film has only a uniform monotonous background, it will cause the MAR to fail. For example, Figure 7 (8) and Figure 7 (9), Figure 7 (8), the original frame '7th just for the line The frame of linear scaling (Fig. 7 (c) is the picture processed by KrShenbtihl et al., and the figure 7 is the embodiment. In the case of the step, the MAR has parallax when the feature points are aligned (4) On the problem, the method cannot convert the objects in different depths of field consistently'. In this case, the MAR method will degenerate into a linear scaling, as shown in Figure 6 (8) and Figure 6 (h). Figure 6 (8) and Figure 6 (4) are the original frames, and Figures 6 (b) and 6 (ί) are the frames of linear scaling (lhear scaUn ), Figure 6 (c) and Figure 6 (g) is [picture processed by Wang et al.], and FIGS. 6(d) and 6(h) are the embodiments. Under the car, the method of this case can process all action type films without distortion, and does not need any alignment of the feature points of the frame. Therefore, compared with MAR, this case can successfully solve the change of the tone depth and the movement of the camera lens. film. "[Krahenbahletal.2009]'s grid architecture is pixel-sized and re-scaled in time. (See, for example, Reference 13). In order to achieve such fast performance, SVR optimizes each frame separately. The problem of non-scaling, and simply limits the consistency of 201141212 when the adjacent pixels are converted on the time axis. The processing time axis is the important information map in space for every five graphs, and Display = object, information; so - to 'visually significant and move to get more important information about the south. Then 'each frame to taste if there is a camera is the object of the wire object 2 to avoid face distortion and fluctuations. In addition to the most advanced film re-scaling method before, this case is also 盥 = = shadow = (this will be much better than the automatically generated results) than the car, you can find that 'there will be no sharp shaking of the lens _ The trend lies in the use of cropping and unequal scaling, especially when there is a significant change. In this experiment, the result of this ^ "visually important object or seriously change the camera path . Value = t: The use of finer, even pixel-level resolution of the grid frame information can be more freshly analyzed. Yes, in order to change the shadow ΓΪ 采用 'The use of finer mesh is limited, because every net Each square is ^-like. In this case, we experimented with various grids to analyze the grid structure and then calculate the operation time and the size of the required memory. Secondly, the system in this case is 40x40, 20x20, l〇xio, and 5 pain cells 3眷25. Miscellaneous uses a fine mesh ^ two fee-free ifr memory space 'but also makes the re-scaling film smoother to the best balance point. According to the first-expo-exemplary embodiment 2, the order is used to calculate the basis of the 96-bit different background and age layer. The result of the comparison between the film and the Pavid 196 rider is "n/ compare the results of the film of the road and the results of [Κτ池enbiihl et al. 2009] and 丹^ng et al. 2009]; the inventor will provide one The original shirt content and the results of either method allow the user to choose the re-scaled version of their favorite 2011·, 201141212. At the beginning of the test, no special technical guidance was provided to the user to ensure the experiment. Accuracy. In the experiment, the film used six different films 'Using the Exposure Algorithm, 2009】, [Wang et al. 2009] to reduce the automatic width by 50% of the film. Each film will do two times two The method is compared, so each user will be asked to do 3><6 = 18 tests. The six films selected in this case include different types of scenes and object motion patterns: like live shots, CG movies, Close-up, wide (four) shots, single- or multiple foreground objects, fast or slow moving lens faces, and the presence or absence of parallax effects; six shadows > {, we used ^ movie theme and - part CG昼 short film, And try to keep each clip film will not have too winged frame 'Because the silk short film will add 3 more comparisons and we can't expect each user to spend more than 2 〇 _3 () minutes to participate in the experiment. And ^ question is赖 咖 咖 丨 。 。 。 。 。 。 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本 本88 - 309 397 SVR 68 一一>—· 267 - 335 Form 1

Table ^ shows the results of the paired comparison of % of the user study participants. The total ^ Fang Mu · / solid comparison was carried out. In this case, the element % in the table is listed in the table, and the number of the two is used. And it can be adjusted by the observers of the chart thirty-three, and the results clearly match the 201141212 results of the algorithm. In summary, the results of the present invention, after comparison, yielded 86.5% (996/1152) votes. Compared with SVR, there are more than 88.2% of the votes, and in addition to MAR, there are more than 847% of the votes. Conversely, SVR has only 29.1% (335/1152) and MAR only 34.5% (397/1152), which means that most participants tend to prefer the results of this paper. The measured Kendall's consistency coefficient is "=().356, which means statistical significance is p<o_oi. Kendall's consistency system' represents whether there is a contradiction between the two or two cycle comparisons, such as - this user Results survey, 78% of the results

User's individual user preference statistical inconsistency is ί = 1, meaning they are; and there are contradictory phenomena, and the average consistency coefficient in this user survey is? = 〇·94, the standard deviation is 〇·1, and only 3 user consistency scores are ο.5 ’, indicating that the survey results are extremely accurate. In this disclosure, it is not a good idea to conduct a user survey on all previous film rescaling techniques. (4) The bottle will focus on the two recent technologies [Kr. enbtihl et al. 2009], [Wang et Al 2009]. In [K^enbtim et al. 2_, the 'SVR method is clearly superior to linear scaling' and [Wang et al. 2009] research shows to prove the ratio [w〇lfet2 and [Rubinstein et al. 2008] The results are better' so the case is not repeated and rescaled, [Wolf et al. 2007], [Rubinstein et al. 2008] for comparison. According to the user survey, we can further help us to further study, and most of them want to see what kind of re-contracted film 'but also more and more complex experiments, such as this case can also be designed to let all the original film status not touched. Next, evaluate various methods. In the case where the user touches the towel, the core is also specially succinct. In the case that the cutting technique based on the time of appearance of the object on the time axis has the original film, the user only discusses whether it is often ignored. The information 'and in Yang Tong has a cut=& smoothing' from the chart three h« can be known as 'most users 20 i S' 201141212 accept this kind of cutting situation. Such as the "important object in the film" as described before the invention is a contradictory target; if in the film, the - segment J movement track covers all 4 faces, in other words, the π in the turn time is moved in (4) All backgrounds must be maintained in the same phase: Ϊ 2: ί ΐ , 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 用 用 用 用 用 用 用 用 用 如 如 如 如 如 如 如 如 如 如 如The lens over time—straight around the figure, the movement of the foreground object on the time of the singer, has covered all the space in the t' [Wang et al. 2009] and the linear scaling has no difference. Can still retain the important objects of the film (4) large shape, ·, where 8 (8) is the county _ box, the 8th is the linear zoom (lu ^arscahng) frame, the 8th (6) is the image processed by Wang et al. And Fig. 8(d) is the embodiment. Compared with the pure use of cutting, the method of this case can retain the most important things _ specific areas, the social important information is removed; cut the unequal scale scaling, can greatly reduce the virtual lens utilization rate; In the case where the cropping is impossible or not (4), the scale is not scaled, the space position is effectively utilized, and the unimportant background is deformed and dusted. The cropping is equal to the unequal ratio of the four (4) _, (4) sub-optimal ties. The method of this case is to increase the area that can be deformed and removed in time and space, but in the shadows; (4) the content towel's presence of significant features and halls, objects, once rescaled, in time and Space distortion distortion may occur, for example, Figure 9 (8) _ 9 (6), where the 9th _ is the original _ box '9th _ is a linear scali diagram, '9th figure ( i) is a picture processed by Wang et al., and Figure 9(d) is the embodiment. In this extreme case, it is necessary to analyze from the artistic point of view. 'Determining which of the key frames in the film can be permanently present. 21 201141212 Cutting and I method is a sneak peek. ===35=Can be tilted by S such a lens, its: (2) =: For noise and 昼 flash: Make the best detection technology possible because of nothing: or not, the special second : ί: Mobile =:=:3z S3: When allowed, 'use a pixel-level grid architecture to do two: b device L1. - Unified block diagram, _ system includes a processing body 23, a 4 = and 24 such as storage medium 23 can store a plurality of execution fingers - internal / external network resources: material selection J network, - wireless network Road, display = and ====_ interest. In addition, the viewer's access to the singular stalk can be obtained by using the stream (10) and the at least one frame. First, the receiving packet has an associated foreground object information_2) ==== 22 a specific area of Si 201141212 (step 33). The foregoing description of the present invention can be seen from the embodiments of the present invention. The present invention is known from the above-described plurality of system 'devices, and: different examples and implementation concepts . Further, a description of the method of the present invention will be made by referring to the description of the method. Incorporating the invention into the memory, the read-only memory, the hard disk, the random storage

Set. The right execution command can be -hard f, a microprocessor, *, _〇, mainfr_ and U 荨. 7 persons understand that the synonymous terms of the language can be included in the summary of the present embodiment. Look for the name of the sub-system, the system mentioned in this embodiment in the film reconstruction. The moving object plays an important role in this embodiment, and the scale of the image can still be preserved in the object. According to the introduction, the object of n can solve the problem of the time dimension of the reconstructed object. And according to the present stimuli flow, it is suitable for spatially timely for rebuilding the lining towel. According to the specific exemplary embodiment disclosed herein, since the analysis and optimization of the entire lens-up to the scene segment is a very 'view' in the program, The cost of the calculation is higher than the time system that is only optimized. ϋ ϋ 人士 人士 人士 人士 人士 人士 人士 人士 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县 县problem. The description merely describes the principles of the disclosure. Various modifications and alterations to the implementation of the description can be readily accomplished by those skilled in the art from the teachings herein. Those skilled in the art 23 201141212: to design more systems, devices and methods, but even if the above is not explicitly shown or disclosed herein, the principles of the present invention are included and are within the scope of the disclosure. . In addition, all of the above-mentioned contributions are hereby incorporated by reference in their entirety. The program described here can be stored in any computer-usable storage medium; =; random access memory, read-only memory, flash drive, CD, ίϊΪ, and can be used in microprocessors, mhi, Processing devices such as m_ and mainframe should all be known. Figure ^ below the imitation of the needles ^ film data processing system, the general convergence. In more detail, please refer to U =) </ br /> according to the invention - a film f material processing system includes: Ϊ Ϊ two series 昼 之 ,, and - embedded system 220: implementation. The video material processing system 200, when applying itf, usually receives a movie by the embedded system 22G, as shown in the 12th step 24G, 25G and 260. And: the step can be / the knife is: receiving at least a frame - the film area, a predetermined zoom target film cube, and zooming on the target case to make the processed frame meet the ' Another proportion. After the completion of the three steps, the content that can be outputted can be the above one, and the three steps can be respectively: receiving the inclusion of: - the picture - the film, defining the inclusion - one of the specific areas: ϊ ϊ; Small, _ in the round of the three-dimensional image of the first, the two steps can also be: the acceptance of - the first two cases of the boundary 2 for each of the plurality of frames - the specific area i, the mother - the frame Unequal scaling, so that the processing two: 24 201141212 = in line with the financial scale, so that the film has a second format for display s ^, the two steps can also be: accept the inclusion and a $ pull the third dimension The target image is measured and the unit time is 1 time interval. Each adjacent two frames in the frame makes the film have a ι ΐ 例 该 该 该 该 该 该 该

[Main component symbol description] 2 〇 processing device 21 processor 22 input/output device 23 computer storage medium 24 execution command 25 storage device 26 display device 10~14 steps 31 to 33 step 2 〇〇 video material processing system 210 film interface 220 Embedded System 240 to 260 Step [Simplified Schematic Description] Flowchart Fig. 1 is a genus diagram used in the second drawing of the present invention. Figure 3 (a) and Figure 3 are schematic diagrams of the right boundary. () In order to utilize the above method, the left side of the side defeat area and the most recent projections of the item surface are unequal scaled, and after the flute t, the frames conform to the size ratio of the target image slice cube. Figures (a) and 5 (b) show corresponding grid diagrams. 6 (8), in which the sixth figure (8) and the sixth _ are the original figure J '6th _ and the sixth figure _ linear sea (linear seaHng) Figure 6 it ΐ (6) and 6 (g) is Wang # The pictures processed by the person, and the figures (d) and 6 (h) are the embodiments. As shown in Fig. 6, the seventh figure (8) is the original frame, and the first one is processed, and the seventh is just the embodiment. 3 picture (=)·”_”The photographic lens has been around the picture for a long time. It is a movement of the foreground object in time. It has covered the space area. The 8th picture (8) is the original frame. 'Fig. 8 (8) is the frame of line $, (h^scaling), the 8th _ is the picture of Wang et al., and the 8th figure (d) is the embodiment. Fig. 9 (8) Fig. 9 (d) In the figure, Fig. 9(a) is 9 (9) is a linear scaling frame, the '9th figure (6). is processed by the person, and the 9th_ is the implementation = _ is ~ ton, the 10th figure is BRIEF DESCRIPTION OF THE DRAWINGS FIG. 12 is a block diagram of a film material processing system according to an embodiment of the present invention. FIG. 12(b) shows another flow of an embodiment of the present invention. Fig. 26

Claims (1)

201141212 Lian; VII, the scope of application for patents: ' ^ A video data processing system, including: ; a video interface, providing a series of formats; and an access system, receiving the series and performing (a) receiving a movie containing at least one frame; (b) defining a predetermined scale of the target movie containing one of the specific regions. ^ (c) unequal scaling of the frames so that after processing The box conforms to the size ratio of the target movie cube to output the format of the other system. 2. The system of claim 1, wherein the step (a) of the step (a) is to find a specific region containing one of at least one of the components. Lai Jingjing 3*, as described in the second paragraph of the patent application scope, further includes a step-by-step finding of a visual flow (〇ptical fl〇w) and quantifying to obtain at least one set of results accordingly. Flow vector. 4. The secret described in claim 3, wherein the quantization process imports at least one frame of visual flow (optical fl0W) into a chart for compilation, statistics, and the probability of distribution of the numbers. An entropy formula is used to find information on at least one line of at least one frame (entr〇Py). = = the system described] wherein the specific area is as described in item 1 of the scope of the patent application, wherein the predetermined scaling is determined by man. 7. The system of claim 1 of the patent application scope further includes at least one optimization formula that scales the frames unequally such that the processed frame conforms to the size ratio of the target movie cube. 8. The system of claim 7, wherein the optimization is based on spatial content and temporal consistency. 27 201141212 9. If the system described in claim 7 of the patent scope, sr Yuwei; the similar energy of the spatial content "=== time-to-sex tree, the content of the content (Temp (10) 丨 ^ gy) and the right After cutting, the second-order smoothing energy is performed for each frame. β This is the system 'j-square problem described in the patent application, and a set of optimized ones is taken. The result is as follows: 11. The system described in the application patent (4)!, the 1-dimensional size is the same as the system described in the scope of claim U, using at least one of the at least one frame ( Optical flow) to obtain at least a few dimensions of the visual flow I unequal _ at least before and after the grading - the dimension as much as possible to maintain the symmetry. 幻 · ^ size linear deformation energy mm1 item ^ money, its data scaling area Loss =: Sliding smoothly 'cutting the money described in item 1 of the less important outside range, and (4) compiling it as a film data processing system, including: an embedded system, performing the following steps: (a Receiving includes a first format (b) defining one of a specific area, Dimensional image; and the proportion of the pre-emptive-shirts (0) unequal scaling of the series, so that the processed system 201141212 A 9 The size of the 3D image of the standard film is issued by the wheel - the second grid 17*: a kind of video data processing system, including: embedded system, the following steps are performed: it is connected to a plurality of frames in a first format; Each of the plurality of specific side frames - a predetermined one of the 11 fields is subjected to a non-z_scaling, so that each of the processed ones is combined with the predetermined scaling ratio, so that the movie has a second format for 18·one movie A data processing system, comprising: an embedded system, performing the following steps: comprising a plurality of frames of a target image having a first format; (b) defining a target rectangular body, the two-dimensional scale of the rectangular body being sufficient to accommodate the The target image 'and its unit time is the third dimension, the unit is the time interval between each frame of the plurality of side frames; and (c) the target image is unequally retracted into the rectangle, Make a second format for display. ~month 29