TW201141205A - System for transforming video outputting format - Google Patents

Abstract

This invention relates to a video output system. The system includes an external input instruction producing a video output format associated with an external instruction; and a graphic processing device which include at least one executing unit, wherein, the executing unit receives the external instruction and performs 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

201141205 VI. Description of the Invention: [Technical Field] The present invention provides a film data processing system, which is a method for processing unequal scale shirts. In particular, the processing method is a combination of cropping and unequal ratios. The processing method of the film zooming and the application description of the related device and circuit using the processing method. [Prior Art] Because of the excellent development of technology now, it will be important to display the device = wide S scaling to resolution, and there are various different types of display devices under different display devices, such as mobile phones, pDAs, and wide. TV, notebook type f brain. In the case of mobile devices such as mobile phones and PDAs, the resolution of the screen required by the smaller screen and the played picture is different from that of the PDA. At present, the location of the target is based on the size of the target county, and the picture and the film are both Β 缩放 或 或 或 裁 裁 裁 裁 裁 裁 裁 裁 裁 。 。 。 。 。 。 。 。 。 。 。 。 。 。 Once the linear test will cause the image to be distorted (4), and the crop will be an important object at the edge of the image. In recent years, there is a scaling technique for i, which uses non-equal distortion to achieve 4 The ulnar nucleus's turn human beings are more heavily weighted in the inner valley, and the distortion distortion is hidden in less important areas. In order to achieve content-aware images and video singularity by using non-proportional distortion or segmentation based on J^ntatum-based composition, the former has a lack of image distortion, and the latter contains ίί images and fragments of the film. Appearance. However, re-scaling the movie is basically a trick. = The problem is only limited by the time limit of the existing image zoom. There are two problems in 34 problems: On j-) object motion and time dynamics are film nuclide _; before the method, the new scaling operation for each age in the movie 201141205 fluctuates ίΐίί students? Can not process movies with complex actions and pictures There will be next, (the second internal volume is the main body of the body to cover the entire piece of j, in this case the zoom is, this is re-released in the second: planting shadow: square (four) ^ [invention content] The object in the - predetermined zoom 2 ^ = background image and the film in the different scaling = and the method of broadcasting is not distorted. The following is a purpose of the film - the film of the specific example of the disclosure can be: / or overcome at least a specific area of the video material with the prior art (such as 'this facet'), the at least - specific area is protected between the blocks P without cutting = the frame: S; the order can include, For example, 'received with at least - shadow ^ 201141205 ^ appears in the information in the -_, and the next specific number of frames disappear, the next === box with a specific flap, and / or (9) contains And the information that is expected to be used to guide the foreground object. Sample program It can further include a certain ί ί ί ί 虽 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 崎 特定 ' ' ' ' ' ' ' ' ' The frame remains visible, where f ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ The function of the second base of the packet is convinced that some of the embodiments relating to the uniform distribution of the flow are most in accordance with the embodiment of the present invention, the example ί = the unequal proportion of the subroutine, which = two: =. At least temporarily - not in the case of: to promote the adjacent order - fixed vertices can be limited, the process of the illustration can further include identifying the grid vertex i vertex position can be "predetermined neighboring top." The unequal scaling subroutine may use a grid comprising a group of squares ' and the instantiation process may further comprise determining at least one particular square having a flow vector extending outside of the particular frame, The particular squares may comprise a size that is the same as the size of at least one further square on at least the time axis adjacent to the particular square. And 'a scaling process may be used to limit the. The unequal scaling subroutine can use a pixel level grid and/or sliding window. In addition, the illustrated process is further included in a format accessible to the user and/or stored in a user readable format. </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> <RTIgt; </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> In the process of film re-production, the process of unequal scaling of at least one specific area of the movie content is promoted. For example, the process, according to some of the foregoing exemplary embodiments, may include receiving a video, a message relating to at least an image frame. With a hardware processing configuration? The two m may include information on at least the Tefang and/or L. mosquitoes. This is determined by the information of the foreground object that disappears within several frames of the ‘area. The instantiation process ίί ΐ, the image frame area is not equal to 该 ί : : : : : : : : : : : : : : 片 片 片 片 片 片 片 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不 不Only to avoid being cut off during the cutting process. And by the community to expose these and other objects, features and advantages, once the disk is illustrated, it will be clearer. The numb embodiment says that the 'think two-format system-external order Peijie 6 201141205 receives a film containing at least one frame - the scale of the film makes = 1 fine _ in line with the = format - with Lin Lai The film round "root (four) is again - conception" this hair day performance for the kind of video recording out of the second two; the exogenous and a film output format has a 'system, including an external input ^ for the shirt type monthly output 袼An external command α and -; 7 execution unit, associated === degree is sufficient to accommodate the requirement that each of the frames in the frame has a different size than the fire hand 2 of the embodiment: First define the important image objects. ~ The scale of the film will be cut. The method is as follows: Hall, object, how to avoid 201141205 1, calculate the visual flow of each frame, after analysis and quantification, get The flow vector of each element in the frame (fl0W vect〇r) 2. Import all flow vectors into a tfan chart to perform number statistics and import the probability of the numbers into the entropy formula (entr 〇py) to find at least one frame Each row of information entropy. 3, each row by using the frame entropy of all, do not allow to find a particular area of the cut (critical region) square

1 , .· 〇 σ 兴 兴 兴 寻 寻 寻 寻 寻 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳 最佳In other words, by defining a specific area in each frame, the content of each frame in this area must not be deleted. Visual traffic is used to formulate its criteria and to calculate a target film cube containing all specific regions, and content outside the critical region can be discarded. Specifically, when a slice 1 is to be narrowed, pixels of a particular line are not removed, and when a film is to be widened, a specific column is left. Take the practice of narrowing the film as an example: (1) When the content just appears in the frame or is about to disappear in the next frame, 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 sides of the area j as a specific line (criticalcolumn). Referring to Fig. 1 ', it is a flow chart 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; = 11: Find a moving foreground object I feature area associated with at least the frame. Η古ίΞ12: Defining the “predetermined zoom target stick movie cube” containing the area of the joy, where the predetermined zoom ratio is determined by humans. , - - 2 2: The quantified miscellaneous information includes introducing at least "(4) visual traffic into the public for numbering statistics, and importing the probability of distribution of the duties into the entropy &quot; seeking at least one row of entropy information, wherein The specific zone 201141205 domain is determined based on the entropy information. For the movie to be scaled, the first definition is included in a specific area of each frame. If the moving foreground object is important (4), the image object should not be cut by the method to calculate the visual flow of each frame (〇ptical fl〇w) , after / knife analysis! After the transformation, get the flow vector of each pixel in the frame (_ all flow vector leads _ fan type chart gae_ to compile the compilation statistics, and the distribution of the number is the frame of the mother H brother Find out which is not allowed to cut the ridge 4: combine cutting and unequal scaling = = = 'conform to the size ratio of the target image cube 2. In addition, in each frame through the absolute must not be deleted. Use view; flow The pixel is removed. In other words, the pixel is removed. In other words, the width is narrowed down. The narrowing of the message is not the case that it continues to appear on the timeline (ΐ ί ί ί ί ί ί Contains significant moving foreground objects and defines specific fields: ΐί Ϊ? Right side is a specific line (4) ical CO—. Whether the frame moves to enter or leave; == at the next ^ and test if it comes from the previous figure ==== to this line It will be marked in a specific area. It is judged that none of the pieces are established. The teacher's coffee (four) is used as the basis, and the movement of the object of the second and second lines of visual flow is different from that of the object that is moving, and the first thing is the flow of the object. (ΐΪ long vector will be quantified . More compartment cell region) derived from human noise table 'need to distinguish between more quantization z) table is not - flow vector / the quantized integer value via the following nit) = 2fc

With k=[〇.5C(ii)U is described as ^^ and Θ (/3 respectively represent the length and direction of ', respectively. The principle of this formula is that the main fan is composed of many concentric circles with the same spacing. The kth eight-th path is 2(k+1), and the k-ring is divided into 2k aliquots, and each r ρ ΐϋΐ is 2π/2. As shown in Fig. 2, all the spaces are from the innermost side. Υ, row consecutive numbering. Place a flow vector in the original of the pie chart: Use Eq.(l) to find the interval number in which the vector is located. In detail, L〇-5£(ft)” represents its flow rate. The vector is located in the first ring of the pie chart, which is the first interval in a ring of (6)/(~#). First, the entropy function (entr〇py) is introduced. In this patent, the method of entropy in information theory is used. Find the leftmost and rightmost boundary of a particular region; the entropy function, 疋 meaning is assumed, there is a system S with multiple events S={EhE2, E3,

} 'and the probability distribution of each event is ρ={Ρι,Ρ2,Ρ3,...,ρ is the entropy function' is as follows Eq.(2) H(C) = -YjP(fi)-\og〇 P(fi).

The i^C-entropy function has several important characteristics. One is that the function value must be greater than zero, and the second is that the hypothesis N is the total number of events in the system s, then the entropy function / 4PFPf·..," At this time, the entropy function of the system S is the largest. The main reason for using the entropy function is that when the probability of each event is equal, the function will reach the maximum, and the probability of the visual flow vector is taken as the probability. The more average distribution means that each flow vector is different, and there are important objects to move; in other words, when the probability distribution is concentrated in a range, the background of the traffic after 201141205 is the unimportant background. Area. Therefore, use the straightness of quantification; 2: Calculate the histogram and define the probability of flow distribution (the height is dominated by normalizing all integer histograms. This can be the parental one = entropy function C (four) ropey of col_ 〇. , set when the flow entropy function of a row exceeds 0.7 pieces max 2 ^ 仃 ^ is contained in a specific area; and I is the maximum entropy function value for each event. Figure 3 (8) and 3 _ for profit The leftmost and rightmost boundary of the fixed area shows _. The specific area (10) is bound to the specific area boundary that is required to be reserved. 'But it does not mean that all the contents of the so-called broadcast must be completely cut; the exact cut area must ensure that the secret Li scaling operation is timely and the limit is maintained. And "3 can be seen on the future time axis. Therefore, it is necessary for the detection before (4) ^7: because the system-based flow entropy is a proportional scaling, so that the processed Chu 4 hidden, L 3 standard The size ratio of the shirt piece cubes, as shown in Fig. 4 (4) and the public towel 'in step 14 , further includes at least one optimization frame' - j frame for unequal scaling, so that the processed picture films are processed. The size of the cube _, and the optimization formula is designed according to the spatial content and time-basedness. The proposed film re-scaling architecture is based on the variational optimization of the two-dimensional Ϊ lian, = _ _ zoom, and cut The cut operation includes the optimization of the tomb, a &lt;. The separation of the film cube area by using the grid structure, the calculation of the vertex position of the grid: under a few conditions, the minimum energy 4 function value Mixed with Lin Qing The new position produces a scaled movie for each (10) new position. The objective function consists of a piece of hardware that maintains time and visually important content, as well as the coherence of object content on the timeline. The symbol. The grid structure of each frame t in the movie Mt= {vt. 201141205 E,Q } 'The towel V= { VA;^··., VA } is the vector position collection, E and Q respectively represent Edges and quad faces, and all the frames are connected to each other. New vertices after scaling deformation = (xi, i, γί, these variables are used to optimize the calculation process. To simplify Complexity, the symbol t is reduced to V; when a single frame vertex is represented by . The reduced movie size (rx, ry, rz), 匕 and ry are the scaled film resolution. The total number of &amp; frames. Basically, the film cube of the input segment is compressed to - pre-scale = example target movie cube.

The optimization formula includes similar energy (jo^formal energy) for maintaining spatial content, Temporal Coherence energy for temporal consistency on the time axis, and smoothing each frame after cropping The secondary smoothing energy (second_order sm〇〇thing). And by using an iterative function (iterative minimization fhnction), the minimum level problem is solved for this algorithm to turn-group best (four) results. ^) Waiting for the moon, and then unequal scaling in the face of the towel, at least - the frame _ grid structure (this unequal scaling uses the frame to slide on the long and wide axes, cutting the outside Less important areas to reduce the accumulation of distortion) to maintain the shape of the object in the particular zone, and to make the grid structure of the grid structure to perform unequal ratio scaling. In order to turn the coffee, use the visual flow of at least the geometry of the frame to obtain the linear deformation of the geometric single 7G size and to unequal scale before and after. The linear deformation of the geometry unit size is as uniform as possible. The following is a description of the preferred embodiment. The method of not scaling the scale of the movie stipulates that after scaling and deformation, the vertex coordinates of the four corners of the original movie must be located in the new resolution movie. In this real e case, all the specific regions are deformed and then reduced. The size of the scene is 〇x, ry'rz). In some films, the non-specific areas on both sides of the film have been deformed. Let Vi and VY in the t frame, the closest to the upper left and lower right corners of the SAR 12 201141205 m must carefully select the vertices to ensure that the specific area - is included in this must be scaled in the target cube = &gt; 0. xl' &lt; Τχ

Ve ~ 0: ^ ry. for all 0 &lt; ί &lt; 7V Hewlett-Packard's unequal-scale film scaling has a time axis: and does not require a time-sensitive cutting area π田为^寺In the scene i, the visually important objects in each frame, (: 〇1^7 Vu, Vi2, vi3, and vi4 respectively represent the four squares of the square q = like ^ change parameterization is by four numbers Composition (eg, w U); the desired q and q ' to achieve the best similar solution: Eq.(4) '[a w=t5l[:], small]沁l * This is a linear least squares problem Can be written as [s''ML' = (4XriAibv, which is as follows. (3)

—2/ή 1 ο Γ χ( η Vh .Χίτ 0 ι y1 Aq = • • :bg/ 二*!/ll ^14 a m4 1 0 Xr • yu 0 1 . u:J will =Af all depends on the grid ·, Μ is—the matrix of unknown numbers will be k C w, v]qq, and will be able to get from the square q

Dc\q..q') = (Aq(A^Aq)-lA!;~I)bq/ can be written as follows Eq.(6) 〆)=ΣΣ^·〇ν)

t a1 J 13 201141205

Wq is the weight value of the visually important squares in the t-th frame, and the spatial importance map of each frame, combined with color gradient measurement, image visual feature map, and accurate face detection, The important information map is normalized to a range of [0.1, 1.0] to prevent undisturbed areas from being excessively scaled to cause significant distortion; the unequal scaling method is similar to (KRAHENBljHL, et al. 2009) and (WANQ, etc.) People 2009). In order to avoid strong distortion of the grid line. Eq.(7) — ^ = Σ* ( E{i,y}eEi; (χΐ ~ .xf)2+

Σ^0.3'}€ΕΛ (?/i -Vj')2 ).

The Ev and Eh cutters represent the vertical and horizontal edges of the grid architecture. In order to achieve the time-exposure shadow &gt;j scaling, this patent proposes a function to maintain the motion of the object and minimize the _, lin, and undulation. With visual traffic, you can get the situation when each square is changed from $/to the next frame. Find, quasi, linear conversion work _ her sfbrm, this example does not need to calculate the inner H is = to contain the square of the outer shape of the conversion τ. The location of the ground. The purpose of this patent is to maintain the image without scaling == shadow, and the unknown of the function, the correct definition means that these vertices are linear combinations closest to the _" vertices, and the fifth _, relative squares q (thickness, as shown in Fig. 5(a), the quantity is analyzed. 'In this case, the linear combination of the vertex of the grid of the visual flow thick grid is inverted, and the vertex of the upper left corner of the u symbol ^ is adopted by &lt;1 = Σ&gt ; ν: +1 paid silk table does not. The following % (9) 201141205 ten '% is the average weight of the square point νγι, now you can use Vi, heavy meaning. As follows Eq. (10). Team (6) = Σ -ι4 +1/)ΙΙ2, ,Ε(&lt;^) is a collection of squares#. In some cases, the square Qt〃 has been moved outside the film frame. For =Shige's use of TEq.(U), after simplification of time scaling, the adjacent squares on the timeline can maintain similar limits as much as possible. =Σ 11(&lt; - - (f - Oil2· b',fc)eE(々= 'The sum of similar energy on all timelines is as follows:

Dt=? Σ _ eQ« 1 ?|€Q^ The same amount uses the local limit to reach the upper phase of the hold time axis: Borrow: ί: gas f ' This represents the inconsistency error will continue to be between the frames, A method is proposed to keep the time axis and the square, and to reduce the error. When the movement of the film is very smooth, use the q/ and the pair 2 to take the =V and the square of the job f, instead of the original hair j change ' δ ^ λ = 5 ° because in the film, the shape of the object S = Ming's 'especially when there is an object or the camera is moving ϋ Allowing some scaling when zooming in - so far the energy of this special only selects a fixed point in this embodiment, choose The placement of the position in the phase _ frame can be smoothed. = second - order smoothing term. As follows Eq (1曰3) white 15 201141205 ' =ηΣ 丨丨W - (W + minimum solution 'synaptic distortion reaches D = DC + Dt + jDt + δϋΒ, in this system' r = ι〇, 5, Proud rabbit Dan / also:, the side is to protect the special === non-negative mf too m inequality, the grid edge can not be ί; = ΐ ί: limit mining, the boundary of the equation., Λ 1 ''. When using the top and bottom of each frame, you want to get the tank inspection and iteration in some linear constraints and the linear condition inequality I. In this embodiment, each time a new (four) formula is generated, it will be Please pay attention to the next entry. When the condition limit is changed, the system ® $ depends on the contradiction of the violation of the inequality.

r· &quot;i ^ I students. The system of this patent, according to [Buat〇isetai 2009] this study, uses GPU ((^hieS Process Unit) gradient solver) and multiple parallel processing as the core architecture, compared to the general single solver It will require a lot of memory space, but it will also significantly reduce the time required for the operation. Once the deformed mesh structure has been calculated, the scaled movie is clipped from the defined target zoom. In addition, the image of each frame is scaled to each square, the system uses linear interpolation, or the test is performed using the 201141205 妓 之 之 杨 杨 ΤΓΥ ΤΓΥ ΤΓΥ ΤΓΥ 中央 中央 中央 中央 中央Nvidia reference example;;= and return (please noisy performance m memorize consumption, because the calculated quality ί=ί put the film's network county structure, Houling ί=; ί=depends on the content of the film On average, you need y' inequality' in the process of mesh distortion. 接 ι ι ι ι 288 288 288 288 288 288 288 288 288 288 288 288 288 288 288 288 288 、 288 288 总数 总数 总数 总数 总数 总数 总数 总数 总数 总数 总数 总数 总数 总数 总数 总数 288 288 288 288 There are more types of frames in the room. The results of the examples are shown to be able to deal with various classes that are used before the parameter settings mentioned in the following = the weight of the decision = in (b) the visual traffic is fine (clear h_eut) 'and Through, for example, =====Qing and linear scaling, and by, for example: (Teng, 舆 ===== two =:=,: private order is the most advanced film scaling technology, so 17 201141205 • This result is compared with the MM and SVR programs. Previous studies [wolf et al. 2〇町], [Ru Binstein et al. 2008], [Zhang et al. 2008] did not consider the consistency of the action object on the time axis when scaling the film. Therefore, it is inevitable that the motion-based method is the most important, 2(8)9]. Based on the user survey, this conclusion is derived. Interestingly, the image rescaling method [Dong et al. 2009], [Rubinstein et al. 2009] combines cutting and other techniques to obtain and optimize The original picture is the most similar result, but these methods require a long calculation time, and only the secret image ' does not extend to the film, considering the problem of consistent objects on the axis. The problem as a comparative literature' Compared with [Wangetal 2〇〇9]; because in this study, it explicitly deals with the problem of consistency on the time axis. However, it needs to use the SIFT method to perform the alignment of each frame feature point, if - The film has a uniform monotonous background, which will cause the MAR to fail, as shown in Figure 7 (&amp;&gt; Figure 7), where Figure 7 (a) is the original frame 'Figure 7 (b) Linear scaling (linea) Rscaling frame, Fig. 7(c) is a picture processed by KrShenbtihl et al., and Fig. 7 ((1) is the present embodiment. ' • Further, ^^ exists when the feature points are aligned Parallax problem, the method can not consistently convert objects in different depth of field, in this case, MAR this method will degenerate into linear scaling, as shown in Figure 6 (8) - Figure 6 (8), Figure 6 (8) and Figure 6 (4) are the original frames, and Figures 6 (b) and 6 (f) are linear scaling (the frame of this postscaling, Figure 6 (c) and Figure 6) (g) is [picture processed by Wang et al.], and figures 6 (d) and (6) are the present embodiment. In contrast, the method of this case can process all action type movies 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 and the movement of the camera lens. film. μ [Krahenblihletal.2009] grid architecture is based on the size of the pixel, real 18 201141205 to such a fast performance, SVR threats to scale the ugly, and is simply _ time, the nature is "every five Sin Now the movie is rescaled. (Refer to the reference information such as the reference data, the important information in the space of these frames can be used to obtain higher important information. However, if the camera shakes or the dynamic object Zhao Da avoids the distortion of the face and Fluctuation

In addition to the first (four) re-scaling method, Lin also 2: shadow this: the automatic result is much better) than the ratio; Ϊ,, in the use of cropping and unequal scaling, especially when the change is significant. In this experiment, the knot = % ' in this case under this harsh condition' cutting may remove important objects or seriously change the camera path. It is worth mentioning that ί, ' uses a finer, even pixel-level resolution of the grid frame and achieve better results' because the visually significant object σ i moves the moon b to more accurately analyze the operation. However, in order to change the shadows, a finer mesh is used, because each of the two parts of the grid is the same. In this case, we experimented with various non-grid analysis and two-grid architectures, and the operation time and the size of the required memory. ^See that the system of this case lies in gamma and Shi. ,]G_, $合:眷 There are any obvious differences. Although the use of fine 'grid, smooth, H, S reminiscent space, but also makes the re-scaling film more smooth j 'experimental' found that the use of Mo pixel grid architecture, is the best in the product balance point. According to the previously disclosed one of the exemplified embodiments, the investigation of the method of using the 96 different backgrounds and age groups is carried out, and the basis of the method of estimating the case is evaluated. The comparison proposed by the case [David 1963] 201141205

Method 'Compare the results of the film of this disclosure with the results of [KrShenbiihl et al. 2009] and ^ang ejal· 2009]; the inventor will provide an original film content and the results of either method, allowing the user Choose the re-scaled version they like. At the beginning of the test, no special technical guidance was provided to the user to ensure the accuracy of the experiment. In the experiment, the film used six kinds of non-correction, mining forest revealing algorithm, [Krghen_e, t a=009], [W, et aL 2_] automatic width reduction article. The master will do a second two-two comparison, so each user will be asked = 3, χ 6 = 18 tests. The six films selected in this case, including different types of field ^ and object motion H, are scene shooting, CG movies, special I,: head 'wide-angle shooting, single- or multiple foreground objects, fast or slow moving ΐϊΐί Face, and whether there is parallax effect; in the six films, we used 3 electricity: theme and - CG animation short film, and try to keep each clip with more frames 'Because each movie will add 3 more comparisons and I Every day, it takes more than 2-30 minutes for each user to participate in the experiment. and

The method was rated a total of 1152 times (2x6x96). MVR —---- SVR Total 488 508 996 ------ 309 397 267 - - 335 Form - Shows the results of the paired comparison of % of the study participants. Total 20 201141205 Execution. In this example, in the middle of the table, the method is better than the method 7. And it can be observed by the chart thirty-three, so that it is clearly preferred to the 8f% (996/1152) times of the algorithm. Compared with the SVR, there are more than 8 seconds. The Ji SVR has more than 84.7% of the votes. The opposite ^ -i ϋ? f 52), the card only 34 '5 〇 / 〇 (397 / 1152) Ϊ t t decided that most participants tend to better the results of this paper. The consistency of = is representative of the two-two cycle comparison. Is there a spear? I-; This user results in the survey, 78% of the 3 spears are not statistically good - the sex is 1 meaning that they are , the phenomenon of contradiction occurs 'and in this user survey, the average - is? = _, the difference is (U' and 3 users - the score is "=0·5, s children have a very high accuracy. - In this $露巾,完|对It is unlikely that all face-to-face techniques will be investigated by users. Therefore, the focus of this case will be on the two recent technologies [KrShenbtihl et al. 2009], [Wang et al 2〇〇91. ^^^^. The study towel 'Caf method (4) is linearly scaled, and the [Wang et al. 2009] research shows that the results are better than Chen F 2 and [Rubinstein et al. 2008], so the case is not repeated and then linear. Scaling, [Wolf et al. 2007], [Rubinstein et al. 2008] for comparison. Based on user surveys, we can further help us to further study and understand what re-contracted films most people want to see' but also need more People assist in the investigation and design more complex experiments. For example, this case can also be designed to allow all methods to be evaluated without touching the original film. In the user survey experiment, this case is also specifically for the purpose of this disclosure: on the time axis Cutting based on the time of appearance of the object Whether it will cause annoyance to watch the movie. However, according to the results of this experiment, regardless of whether or not 2) 201141205 has the original video, the user only focuses on the center of the film, often ignoring the information on both sides of the film, and The system has a smoothing and smoothing for the cutting. The following table shows that the 'reporter can accept this kind of cutting situation. As described in the present invention, keep the important objects in the film in time and space. The upper form, this is a contradictory target; if in the film, the moving track of a heavy object covers all the faces, in other words, to pile up the pants-g, all the backgrounds of the object moving in space are the same Scaling' and in this case, unequal scaling will exhibit a linear scaling effect. In this case, the method will look for an optimal balance point, using the crop to remove portions that have visually appeared or are future Seeing the scene. As shown in Figure 8 (a) - Figure 8 (d), the photographic lens has been around the girl in the picture for a long time, thinking and smelling the movement of the foreground object in time, already covering the space. Above all = scene area, [Wang et al. 2009] method and linear scaling ^ no difference, but the method of this embodiment can still retain the film size, which is the original picture frame, Figure 8 (8) For the frame of line = (linearscaling), Fig. 8 (6) is the picture processed by Wang et al., and Fig. 8 (d) is the embodiment. Compared with the pure use of cutting, the method of this case must be retained the most important. A specific area of the object prevents important information from being removed; in addition, the cropping combines unequal _scaling, which can reduce the virtual lens's usage rate; and if the cropping cannot be _ or not significant, the unequal ratio , put, effectively use the position of the line, the important background is expected to become shrinking. The cropping / unequal scaling uses the ratio, and the variational optimization function balances the two. The method of this case is to increase the area that can be deformed and removed in time and space, but there are still many visually significant features and other objects in the content of the film space. After re-scaling, it may lead to time and inhibition. The phenomenon of clever distortion occurs, for example, Figure 9 (4) - 9th _), the 9th figure (4) in JL is the original frame, the 9th _) is the linear scaling gamma (4) (4) 22 201141205 '9th figure (6) is The map processed by WanS et al. &gt;1, and the 9th figure (4) ^this is the example. In the case of this _ end, it must be divided from the artistic point of view = '= in the film which key frame range can be permanently present when the ^ axis shooting method is presented 'if the tilting automatic cutting standard may cause some Significant objects are always used! · Face = but this #县县结构 is less flexible, but (4) J is the limit of action-oriented, so as long as this type of lens is completely defined. Secondly, the method of this case must be:, and the _ _ leads to analysis errors, so that = d: to be included in the specific area that can not be cropped 'the impact of the scaled after the adoption of the grid architecture' when one of the parties The grid contains the conversion data of the movement of the ί, which does not represent the strictness of the grid. The grid structure of the pixel level can completely avoid the device 1 containing a processing 眺-position 25 and -tf=22, - Computer storage media _ Λ / ... The device 22 includes a wired network, a helmet, a wiring and a sensor, etc.: = a gas connection and a touch message outputting the processing device = two devices 26 and / or the storage device μ transmission two (four) the explicit access format / user-readable 的 资 / 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或 或A 23 201141205 flowchart proposed by the embodiment is executed by the processing device 20. First, a movie containing at least - Chen is received (step 31). Next, the foreground object information associated with at least one of the frames is included (step 32). Then, it is decided to be exempt from the specific area being cropped (step 33). The foregoing describes various embodiments and principles of the present invention, and those skilled in the art will be able to know from the actual closures a different system and implementation description of the invention' and from the above-described systems, devices, and methods. The concept of the invention is understood. In addition, all documents and references presented may be incorporated into the present invention. It can be understood that the computer storage medium includes a hard disk, a random memory, a read only memory, a flash drive, a compact disc, a memory stick, etc. This type of device can be processed by the processing device/computing device 2G. Benefits, - microprocessor, mini ' macr〇, m ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ According to the above, in the film reconstruction, the method, the process, the storage medium and the method mentioned in the embodiment can make the film _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The position of the object in the video can still maintain the scale of the image. According to the preface, the time dimension of the object is solved. And according to the visual sense, the space is applied to the reconstruction process in time according to the specific demonstration disclosed herein. The sexual embodiment, from the lens - until the analysis and optimization of the scene segment is the program? View, 'the cost of its calculation is relatively higher than the real point that only uses the pre-frame optimization. Because the iis technology should be able to The results of an exemplary embodiment of the ancient processing results provided by the disclosure of the present disclosure may be as follows: 24 201141205 Modifications and alterations to the description are expressly disclosed or disclosed by those skilled in the art, but still include the scope of this == and: Jin Jie Jie. In addition, all publications related to the foregoing. U and reference Wei _In this whole read enough to test the existence of any computer can be like the bribe media = such processing device is more suitable for the job. ml _*Q' _fr_ ^ then condensed for the film output format system. Details, Du lLtiV 〇V^^ ;〇° ^ Include an execution unit. When the shadow/wheel is used to invent the method of the invention, usually the system is used, and in the application, the external command is performed as shown in the 12th The steps may be respectively: receiving at least one 145-100 dans and - the predetermined zoom ratio of the three-step specific area _ £ =, defining the inclusion-cube object, the upper-scale movie, defining the package, including At least one of the frames, the image coordinates, and The ^ image ^ target film zoom ratio of the three-dimensional, in line with the film has Ke Ruhong - 俾 俾 该 41 201141205 = unequal scale reduction; the second sign of the 俾 该 该 该 该 该 该The three occasions of the ceremony can also be: accept the inclusion-target image

y accommodating the target image, and for a unit time of 2 units of time, the plurality of side frame towels are separated every time, and the target image is unequally scaled into the rectangular body, 俾Same as the shadow λ _ format - miscellaneous carving display. ~ There is no reference 1 · AVIDAN, S., AND SHAMIR, A. 2007. Seam carving for contentaware image resizing. ACM Trans. Graph. 26, 3,10. 2. BARNES, C., SHECHTMAN, E., FINKELSTEIN , A., AND GOLDMAN, DB 2009. PatchMatch: A randomized correspondence procedure for structural image editing. ACM Trans. Graph. 28,3. 3. BUATOIS, L., CAUMON, G., AND LE VY, B. 2009. Concurrent number cruncher: a GPU implementation of a general sparse linear solver. Int. J. Parallel Emerg. Distrib. Syst. 24,3,205-223. 4. CHEN, LQ, ΧΙΕ, X., FAN, X., MA, WY , ZHANG, HJ, AND ZHOU, HQ 2003. A visual attention model for adapting images on small displays. ACM Multimedia Systems Journal 9,4,353-364. 5. CHO, TS, BUTMAN, M., AVIDAN, S., AND FREEMAN WT 2008. The patch transform and its applications to image editing. In CVPR Ό 8. 6. DAVID, HA 1963. The Method of Paired Comparisons. Charles Griffin &amp; Company. 26 201141205 7. DESELAERS, T.9 DREUW5 P.5 AND NEY, H. 2008. Pan, zoom, scan: Time-cohere Int, trained automatic video cropping. In CVPR. 8. DONG, W., ZHOU, N., PAUL, J.-C.5 AND ZHANG, X. 2009. Optimized image resizing using seam carving and scaling. ACM Trans. 28,5,1-10. 9. FAN,X·,ΧΙΕ,X·,ZHOU,AND MA,W.-Y. 2003· Looking into video frames on small displays. In Multimedia 503,247-250. 10. GAL , R.3 SORKINE, 0., AND COHEN-OR, D. 2006. Featureaware texturing* In EGSR,06, 297-303.

11. ITTI, L., KOCH, C.? AND NIEBUR, E. 1998. A model of saliencybased visual attention for rapid scene analysis. IEEE Trans. Pattern Anal. Mach. Intell. 20,11,1254-1259. KARNI, Z.? FREEDMAN, D, AND GOTSMAN, C. 2009. Energy based image deformation. Comput. Graph. Forum 28, 5, 1257-1268. 13. KRAHENBUHL, P” LANG, M” HORNUNG, A” AND GROSS ACM Trans. Graph. 28, 5. 14. LIU, F.? AND GLEICHER, M. 2006. Video retargeting: automating pan and scan. In Multimedia Ό6, 241- 250. 15. LIU, H., XEE, Xs MA, W.-Y., AND ZHANG, H.-J. 2003. Automatic browsing of large pictures on mobile devices. In Proceedings of ACM International Conference on Multimedia, 148- 155. 16. PRITCH, Y., KAV-VENAKt E·, AND PELEG, S. 2009. Shift-map image editing· In ICCV'09· 17. RASHEED, Z.3 AND SHAH, M. 2003. Scene detection in Hollywood movies and TV shows. In CVPR '03, vol. 2,11-343-8. 27 201141205 18. RUBINSTEIN, M., SHAMIR, A., AND AVIDAN, S. 2 008. Improved seam carving for video retargeting. ACM Trans. Graph. 27,3. 19. RUBINSTEIN, M., SHAMIR, A., AND AVIDAN, S. 2009. Multioperator media retargeting. ACM Trans. Graph. 28, 3, 23. 20. SANTELLA, A.? AGRAWALA, M.? DECARLO, D., SALESIN, Ds AND COHEN, M. 2006. Gaze-based interaction for semiautomatic photo cropping. In Proceedings of CHI, 771-780.

21. SHAMIR, As AND SORKINE, O. 2009. Visual media retargeting. In ACM SIGGRAPH Asia Courses. 22. SIMAKOV, D·, CASPI, Y” SHECHTMAN, E·, AND IRANI, M. 2008· Summarizing visual data using bidirectional Similarity. 23. In CVPR, 08. SUH, B., LING, H., BEDERSON, B. B., AND JACOBS, D. W. 2003.

Automatic thumbnail cropping and its effectiveness. In Proceedings of UIST, 95-104. 24. VIOLA, P.s AND JONES, M. J. 2004. Robust real-time face detection. Int. J. Comput.

Vision 57, 2, 137-154. 25. WANGS Y, S·, TAI, CL., SORKINE, 0., AND LEE, Τ·-Υ. 2008. Optimized scale-and-stretch for image resizing. ACM Trans. Graph. 27, 5, 118. 26. WANG, Y.-S·, FU5 H·, SORKINE, 0·, LEE, T, Y·, AND SEIDEL, H, P· 2009.

Motion-aware temporal coherence for video resizing. ACM Trans. Graph. 28? 5. 27. WERLBERGER, M.? TROBINs Ws POCK5 T, WEDEL, As CREMERS, Ds AND BISCHOF, H. 2009. Anisotropic Huber-Ll optical flow. In Proceedings of the British Machine Vision Conference (BMVC). 28. WOLF, L.5 GUTTMANNs Ms AND COHEN-OR, D. 2007. Nonhomogeneous content-driven video-retargeting. In ICCV 507. r Λ iii .28 201141205 29. ZHANG, Y,-F., HUS S.-M., AND MARTIN, RR 2008. Shrinkability maps for content-aware video resizing. In PG Ό8. 30. ZHANG, G.-X., CHENG, M.-M ., HU, S.-M., AND MARTIN, RR 2009. A shape-preserving approach to image resizing. Computer Graphics Forum 28, 7, 1897-1906. [Main component symbol description] 20 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 400 movie output format system 410 external input command 420 drawing processing device 430 execution unit 440~460 step [simple drawing Description] Fig. 1 is a flow chart used in the present invention. Figure 2 is a pie chart. 2 Figure Guansha's financial method and the rightmost noun are not intended to be shown in Fig. 4 (8). The frames are scaled unequally, and the frames conform to the size ratio of the target movie cube. Figure () and Figure 5 (9), corresponding to the grid diagram. 29 201141205 Figure 6 (8) - Figure 6 (h), where Figure 6 (8) and Figure 6 (e) are the original frame 'Figure 6 (b) and Figure 6 (f) for linear scaling The frames of the linear scaling 'Fig. 6(c) and Fig. 6(g) are pictures processed by Wang et al., and Figs. 6(d) and 6(h) are the present embodiment. Figure 7 (8) - Figure 7 (d), where Figure 7 (4) is the original frame, Figure 7 (b) is the linear zoom (this coffee scaiing) frame, Figure 7 (c) is The picture processed by Krahenb Uhl et al., and Fig. 7 (4) is the embodiment.

In the factory. Solid, the camera lens has been around the girl in the picture, thinking about the movement of the foreground object in time, has covered all the background areas in the space. The 8th picture (8) is the original picture, Figure 8 (9) For the linear scaling frame, Fig. 8 is a picture processed by Wang et al., and Fig. 8 (φ is the embodiment. Fig. 9 (4), wherein Fig. 9(a) is The original frame, the 9th „linear scaling frame, the 9th figure (6) is the picture processed by the face, and the 9th _) is the embodiment. FIG. 10 is a system of the present invention. A block diagram of a film output format system of the present invention. Fig. 12 (8) Another flow chart proposed by the practice of the present invention.

Claims (1)

201141205 VII. Scope of application for patents: 1. A film output format system, including the command of the Ministry of Foreign Affairs, which produces a foreign language associated with a film output format, in which the target film is established. A drawing processing apparatus comprising: at least one execution unit row unit receiving the external command to perform (a) receiving a movie including at least one frame; (b) defining a predetermined scaling cube including a specific region; (=) The frames are unequal-scaled so that the size ratio of the processed target movie cubes in the round-out format is predetermined. 2. If you apply for the secret described in Item 1 of the Special Na, the towel step (8) is followed by a step-by-step screaming for a specific area containing one of the moving foreground objects associated with at least the frame. 3* The system of claim 2, further comprising the step of: finding an optical flow of at least one frame and performing quantization to correspondingly obtain a flow vector of at least one result ( f|〇w vector). 4. The system of claim 3, wherein the quanting process comprises importing an optical flow to the frame into a statistical chart for numbering statistics and importing the numbered distribution probabilities into the entropy ( Entr〇py)&amp; is used to obtain entropy (pyr) information of at least one line of at least one frame. 5. The system of claim 4, wherein the particular region is determined based on entropy information. 6. The system of claim 1, wherein the predetermined scaling is determined by an artificial one. 7. The system described in claim 1 further includes at least one optimal 31 201141205 formula for unequal scaling of the frames to match the size ratio of the target movie cube. _Materials 8. The system described in Shen Qing, her H, item 7, in which the material optimization is designed according to the spatial content and time consistency. 9. The system of claim 7, wherein the optimization comprises the energy used to transfer (four) mail (eGnf_al, similar energy for time-based content on the time axis (Temp〇 Ral c〇herence energy) and a second-order smoothing energy for each frame after cropping 〇=system as described in claim 9, wherein an iterative letter is used The ICitemtive minimization function) obtains a set of optimized results for the energy solution. 11. The system of claim i, wherein the step of performing the step is based on the grid_ and is scaled to maintain the shape of the object in the feature area. The special fiber 11th 11th system "the towel system uses the geometric single 7L size to scale the grid structure unequally. =0 The system described in item n of the patent application scope, wherein in order to achieve time-, at least - at least - the visual flow of the geometric unit size = 1 cal flow) bribe at least _ rib unit ruler line (four) change and Vija The system described in the section, wherein the unequal scale S slides smoothly on the long and wide axes, and cuts the outer smaller area to reduce the accumulation of distortion. 15. A film output format system comprising: _ an eight-person external input command, generating an external command associated with a - movie output format; and an execution unit 'receiving the external command to execute. 32 201141205 (a) receiving a film comprising at least one frame; (b) defining a predetermined three-dimensional image coordinate including one of the specific regions; and (c) arranging the frames in an unequal scale, the frame conforming to the The size of the 3D image coordinates of the target movie is different from that of the new format of the movie output format for display. The film has a 16.-type movie output format system, including:, •” No. An external input command generates an operation unit associated with a movie wheel command; and/or a call-out type, and receives the external execution unit. External command and (4) accept a plurality of frames; (b) define a scale for each of the plurality of frames and a predetermined frame of the B-region to perform the non-axis mismatch, so that the per-out format The second new grid 2 is for display 〇 = 17. - a video output format system, comprising: a command to generate an external execution unit associated with a movie rotation format, which is executed by receiving the external command. a) accepting a plurality of frames containing a target image; , the object ΐί I f rectangular body, the two-dimensional scale of the rectangular body is sufficient to accommodate the unit time as its third dimension, the unit time is the plurality of side frame towels per _ The time interval between the two frames; and the ticket ^ is unequally retracted into the rectangle, so that the movie - there is a new format for the display of the format of the image.