TWI240572B - Method of dynamic smoothing for de-interlacing images - Google Patents

Abstract

The present invention relates to a method of dynamic smoothing for de-interlacing images, which provides a de-interlacing image to perform the following steps: a dynamic image of the de-interlacing image is detected; the dynamic image is then compared with an image which adjoins the dynamic image in a spatial domain to find out a difference; and the smoothing is performed for the dynamic image based on the difference between the dynamic image and the adjacent image.

Description

1240572 P25-045 Order, description, invention description (Invention is called Ying Xu: Brief description of the technical field to which the invention belongs, prior technology, content, implementation, and drawings) [Technical Field of Hairdressing Institute 1¾ 丨] ^ This case is A method for dynamic smoothing of a de-interlaced image. In particular, a smoothing process is performed on the dynamic image of a de-interlaced image, so that the real field and the interpolation field of the de-interlaced image can obtain a smoother image. [Prior art] When de-interlacing the interlaced video, you can first distinguish the image into two parts: dynamic and static. For static scenes, you can use Weave to perform interpolation (ιηΐπρ〇). ι ^^ 〇η), to maintain the original image resolution for best results. For moving images, although there are deinterlacing methods such as johan Intcrpiation and Edge Enhancement, J ^^^ # (Sawtooth) > ^ / 4 (Serration) > MfKJudder) f Γ〇1Ι |, = 疋 may cause discontinuities in the interpolation points due to improper interpolation methods, causing the output image to flicker (hcker) or jitter. This kind of situation that the image is shaken due to the interpolation of the wrong material, is in Taiwan;:, ^ 516'310 "in the interlaced display, the field image display is smooth and the square two-dimensional image 2 is smoothed. 2. The method and device of the Taiwan Patent Announcement No.… to increase the resolution and increase the resolution "and Taiwan Patent Publication No. 00 516 31 ('^ Zhongda Yi field image display smoothing method, the solution disclosed, To solve the above problems with interlacing, conventional techniques such as Taiwan Patent Publication No. 00,228,584 "average at a single point (Plxel) and adjacent points in order to blur the image = slipping. However, Lang's technology for interlaced fields or deinterlacing Variation = Therefore, all images are processed in terms of processing efficiency. [丨 person m,], the purpose of the case is to create a kind of wrong image based on 1240572 P25-045. The method of this case. There are mainly two .M Δ, WM ^, t ^^ (Masking) 4 ^; ^^ 〇t; 〇n Adapt.ve «^ ^ ^ where one is Content Adaptive, for staggered 3 straight = paste , Or smooth the places where the interpolation effect is not ideal; and if _ 二 垂 = = For similar points, it will not be processed. In this way, the redundant two # can be avoided to avoid the problem of processing efficiency with technology. Law = The purpose of the case is to establish-the method of dynamic smoothing of bridging interlaced images. The processing point is the center, * 3x3 or more odd square matrix = mask (Mas lng) 'weight matrix corresponding to the mask (Weightir ^ MaM ^^ (Convolut.on Operat.on), 0 product [contents of the invention] In order to achieve the above purpose, the present case proposes a kind of dynamic smoothing processing of de-interlaced images ^ providing a de-interlaced image to perform the following steps: detecting the dynamic image of the de-interlaced image; comparing the dynamic image with the The difference between the spatial domain and the adjacent image of the moving image; and the difference domain corresponding to the moving image and the woman's image.

Smoothing. T As described in the method of dynamic smoothing of the wrong image, the motion image of the de-interlaced image can be de-interlaced by the motion detection of the system (with the addition of ㈣ to separate the dynamic and static images. The method for dynamic smoothing processing of the de-interlaced image, which includes ten adjacent images, may be an image adjacent to the vertical coordinate direction of the dynamic image. / ▲ The dynamic smoothing of the de-interlaced image as described The method of processing, wherein the difference between the dynamic image and the adjacent image may depend on the comparison result of the disc-threshold value of the dynamic image and the adjacent image. "As mentioned A method for dynamic smoothing of a de-interlaced image, wherein the smoothing process may be a mask smoothing process. As described in the method for a dynamic smoothing of a de-interlaced image, the $ mask of the dynamic image Smoothing can be performed by Convolution Kernel. 1240572 P25-045 As described in the method of dynamic smoothing of de-interlaced images, the mask of the kinetic energy image is flattened. Processing, a weighting matrix (Weighting fetrix) and a mask centered on the moving image (Masking) can be used to perform a convolution operation. ^ As described in the method for dynamic smoothing of de-interlaced images, Among them, the smoothing process of the dynamic image is performed, and a smoothing process is performed on the adjacent image. As described in the method for de-interlacing the dynamic smoothing process of the image, the smoothing process may be a Mask smoothing. As described in the method for dynamic smoothing of de-interlaced images, the smoothing of the masks of adjacent images can be performed using a convolution kernel (C〇nv〇luti〇n Kernei) method. As described in the method for dynamic smoothing of de-interlaced images, in which the smoothing of the masks of adjacent images can use a weighting matrix and the masking centered on the adjacent = image. Masking performs convolution operation (convolution 0perati addition). As described in the method of dynamic smoothing of de-interlaced images, the de-interlaced images can be decomposed into an interpolation field (Inte rpolation Field) and a Real Field, the interpolation field is a de-interlaced image, and the real field is an actual image. As described in the method for dynamic smoothing of a de-interlaced image, the detection Measuring the dynamic image of the de-interlaced image is to detect the dynamic image of the interpolation field of the de-interlaced image. The motion detection of the de-interlaced system can be used to separate the dynamic and static images. A method for dynamic smoothing of a de-interlaced image, wherein the adjacent image may be a real-field image adjacent to a vertical coordinate direction of the motion image of the interpolation field. The method of dynamic smoothing processing of the de-interlaced image as described above, wherein the motion image of the interpolation field and the image are smoothed. The degree of difference between adjacent real-field images may depend on the comparison between the difference between the motion image of the interpolated field and the true-lean field image adjacent to the edge and a threshold value (Threho 1 d). The method for dynamic smoothing of a de-interlaced image as described above, wherein the smoothing may be a mask smoothing. As described in the method for dynamic smoothing of de-interlaced images, the mask smoothing of the moving and evil images of the interpolation field can be performed by using the mirror of the Convolution Kernel. As described in the method for dynamic smoothing of de-interlaced images, the mask smoothing of the 1240572 P25-045 moving image of the interpolation field can use a weighting matrix and the interpolation Convolution operation (c〇nv〇luti〇n

Operation) As described in the method for dynamic smoothing of de-interlaced images, the smoothing processing of the adjacent images should be performed according to the mask smoothing processing of the% moving and evil images. The method for dynamic smoothing of a de-interlaced image as described above, wherein the smoothing may be a mask smoothing. The method for dynamic smoothing of de-interlaced images as described above, wherein the smoothing of the C mask of the adjacent true η 昜 heart image can be performed using the convolution kernel (c〇nv〇iuu〇n Kernei) method. . As described in the method for dynamic smoothing of de-interlaced images, the masking smoothing of adjacent real households, fields, and images can be performed using a weighting matrix and the adjacent The real-field image is centered by a masking (Masking) to perform the stitching operation ((: 〇 〇 1 ⑽ Operation) 〇α > See the schematic diagram of the de-interlaced image of the preferred embodiment of this case. De-interlaced image It is a complete picture generated after de-interlacing (de-inter lace) at the interlaced field of Miao Miao ^ (, iraniL). Therefore, it can be regarded as the interpolation field FI (Interpolation Field) and the real field FR (Reai hddW). The combination of fields, where FR is the actual image, ρ is the image obtained through de-interlaced interpolation, and the single-point (pixel) i ^ FI (x, y, t) of the interpolated field of the scene is represented, The single point of the real field adjacent to H x, ^ —for example, a single point below the vertical coordinate, is represented by FR (X y + it) = _, which is the flow of the de-interlaced image in the preferred embodiment of the present invention. Figure. There should be two non-existent aspects: the me and the mask are only used for masking (moving) of the moving image; the second is the complaint image and the adjacent image. The image with a small degree of difference will not be processed. The implementation steps of the preferred embodiment of this case in July are as follows: 1240572 Step 20: Detect the dynamic image of the de-interlaced image. P25 '° 45 wrong stage, _moving system (MQtiQn Deteet ^ wrong image is being disjointly judged. Since the de-interlacing system is used to obtain the second image of the interpolation field ^ and the result of the shift image is PI to perform dynamic and static separation. Therefore, in this case, _ For the interpolation field FlU, y, t), it is sufficient to generate a dynamic and static standard BMD (xyt). The judgment formula ^ measures the interpolation field early MIXx, y, t) fi; f / (x, #) is detected as dynamic Image l〇; / ύ (χ, detected as a static image Step 2: Compare the dynamic image with its neighbor \ 〇t MD (x, y, t) == l ^:! Ff '' Γί (. W, ^ _ Find the representative FI (X, y, t) as a moving image, and then further '> Compare HU, y, t) with the image adjacent to the straight coordinates, that is, the difference of true FR (x, y + l, t). The judgment formula is as follows: U, y ΙΛ) Ρ

MaskConstraintI (x, y, t) ^ ((l ^^

MaskCCDsmuntUx (port) is the difference between the interpolation field and the adjacent real field, and Thcl is the threshold (Threshold). Μ_〇 on ainti (x, y t ㈣ table HUy, t) is interpolation error or poor interpolation effect. In addition, in order to avoid the extraordinary situation of the image picture, the real field single point FR (x, y-1, adjacent to FIU, y, t) must be FR (x, y, where-'must be equal to FI (x, y, t) are processed synchronously. In the preferred embodiment of this case, FRU'y + l't) # FI (X'y, t) is selected for processing together, so the real field judgment rule MaskConstraintR (x, y + 1, t) is rainUU, y, t)

MmsIU ons I ra i π l R (x, y U, l.)-. MaskConst Step 22 · According to the degree of difference between the M image and the adjacent image, it is determined that the frequency dynamic image 1240572 P25-045 image is smoothed by $ mask deal with. When (MD (x, y, t) == l & MaskConstraintI (x, y, t) == l) is established, it means FI (x, y, t) interpolation error or poor interpolation effect, Mask smoothing is needed ^ · 篆 篆 + slipping processing 'can take the method of Convolution Kernel. The method is to select the convolution kernel ... a weighting matrix w (Weighting Matrix) sums up to 1 Kx'y, 1 ) As the center of the interpolation field mask MaskFI (x, y, do convolution operation (c0nv〇lutiOn Operation). Where MaskFI (x, y, t) is an odd square matrix, generally taken as shown in the third figure The 3X3 mask of the formula can obtain a good smoothing effect; the mask can also obtain a larger one, such as 5 × 5 or 7X7, but the required calculation time will be lengthened. The calculation formula of the mask smoothing processing is as follows w2 h3 η4 ννδ v ^ 6

vt7 mS —

MaskPK x ^ y, t) FRix-ly ^ U) FR (xKy ^ U) FB (x + ly ^ U) F '/ ix-lyj) Ff (w) F! (X + \-y ^ / TiU -l , V + U) state force ton +1, state force Xl (x, y, t): w * MaskFI (x, y, t)

^ 'One interpolation ^ field order care about the output of the mask smoothing process' results such as T l nr X, y, t) == 1 MaskC〇nstraintl (x, y, t) == l ^ t Ribbey j XI (X, y, t) = FI (x, y, t). Perform mask smoothing. Go and pluck the similar image ~ ",. Tian (MI) (x, y't) 22 1) (Masku) When this LrainU (x, y, t) = i) holds, (MDi, n (MaskC〇nstraintR (^ + ^ -ίΓιΪγ 4ίΤ '; FR (x, y + li mask MaskPR (x, y + l, t) as shown in the fifth figure, the calculation formula is as follows: 10 1240572 P25-045 Let MaskFR (x , Y + 1, t) 'FI (XW) FI (x, y, t) FI (x + l, y, t) F / 3〇-1, less + U) F take factory / 3 ^ + 1, Less + i, 〇FI (xl, y + 2, t) F ^ y + lt) household 7 (jc + 1, less + 2 ^) then XR (x, y + l, t) = w * MaskFR (x , Y + l, t)

XlKx, y 11, and 0 are the output of masking smoothing of the real field single point fr (x, y + i, t). The result is shown in the sixth figure. Of course, if MD (x, y, t) = l Or MaskConstm.intR (x, y + 1, t > -l, if one is not true, then XRl ^ y + i, t) = FWxj + i, t). · The above-mentioned preferred embodiment of the present case is a mask smoothing process for a video field that generates a de-interlaced image after de-interlacing. Therefore, it is necessary to retain the movement of image dynamic and static separation during the de-interlacing phase. The detection result is used as the basis for detecting the moving image of the interpolation field of the deinterlaced image. In fact, other implementations in this case can also perform mask smoothing on dynamic images at the same time as the deinterlacing of the video cafe, so that when the hardware is implemented, the video can be deinterlaced and masked. The smoothing function is integrated in a single integrated circuit. As mentioned above, this case is based on the lack of conventional technology, and proposes a method for dynamic smoothing of de-interlaced images. It is based on the _ image for smoothing, which can make the real field and interpolation of interlaced images. Complement the field to obtain a smoothed image. The step in this case is that

The characteristics of Maptlve and Content Adaptive are aimed at moving images and are interpolated *.丨 Performing mask smoothing at 5 points can avoid unnecessary calculation steps and avoid the problem of processing efficiency of conventional techniques. The & technique concealed in this case can be implemented by the technical personnel of this case, and its secret work ^ is patentable '爰 according to the law * patented towel please ^ but the above embodiments are not enough to cover this case The patented brain is desired to be protected, therefore, the scope of patent is requested to be attached. [IHi system (say)] This case can get a deeper angle with the following icons and detailed descriptions ^:, a picture: the schematic of the de-interlaced image of this Yu Jiashiguan ^ second picture: the case The implementation flow chart of the method for smoothing the mask of the de-interlaced image in the preferred embodiment. 1240572 P25-045 Third figure: Interpolation field mask MaskFKx, y, t). The fourth function: the mask smoothing output XI (x, y, t) of the interpolation field. Figure 5: MaskFR (x, y + l, t) of the real field mask. Figure 6: Mask smoothing output XR (x, y + 1, t) of the real field. The figure numbers of the main components shown are as follows: FI: Interpolation Field FR: Real Field MI: Dynamic and Static Marking

MaskConstraintI: Judgment rule for interpolation field MaskC () nsira i ntR • Judgment rule for real field w: Weighting Matrix

MaskF]: Interpolation field mask

MaskFR: Real-field mask XI · Mask smoothing output XR · Real-field mask smoothing output X, y: Space-domain coordinates t: Time-domain coordinates

12

Claims (1)

1240572 P25-045 Patent Application Scope 1 Method for dynamic smoothing of parental malformed images, which provides-deinterlaced images to perform the following steps: "Don't interpret the dynamic images of the parental malformed images; compare the dynamic images The difference between the image in the spatial domain and the image adjacent to the dynamic image, and the degree of difference between the flat image and the adjacent image, decided to perform the solution on the dynamic image 2. The method of dynamic smoothing of interlaced images = dynamic images of wrong images' can use motion detection of the de-interlacing system to separate dynamic and static images. The two-phase reading can be the method of dynamic smoothing of the de-interlaced image with the vertical coordinate of the tactile image. The method of deleting the image and the phase planting is deleted. The method of smoothing the secret of the image, method = smoothing of the mask of the moving image, which can be taken as the core of the convolution. Edge) 7. As explained in item 5 of the scope of patent application, the method of smoothing the mask of the moving image, the processing method of the heart image, and the masking method centered on the moving image-weight matrix (We Coffee-Matrix) Operation). Mask (MasklnS) performs convolution operations (Convolution 9, such as the patent application, the 7th § flattening of the month, processing-smoothing the adjacent image. Where the flat: two = _ item: Slip ^^ 1 10, as described in the application of the 9 items of the Lai Zhiling materials mixed with the reading method, 13 1240572 1240572 P25-045 === tender objects turn, hurt the city nuclear — 1 + she described in 9 items A method of error-checking hybrid smoothing, Mat. The mask smoothing of the shirt image of the mesh ^ can use a weighting matrix (Weighting tnx ^ (LonvoluUon Operation). ($ =====:?: Ξ5Η =: ==: The kinetic energy of the field "1. The dynamic initial image is the detection of the interpolation of the deinterlaced image), = the real field image of the dynamic smoothing process of the image. _Image of the method of "Coordinate Direction Adjacent Method :: m: 2" described in the "De-interlaced Image Dynamic Smoothing Process" ^ Image Subtraction_ Real Subtraction «Subtraction, may depend on === with this phase The difference between the neighboring real-field images and -_ mask-, can _ maneuver the core ^ such as the 16-position erroneous type of Pu smooth smoothing method of the Chinese material, its cap "weak dynamic video materials === : The image is a 1240572 P25-045 19, such as the method of dynamic smoothing of the de-interlaced image as described in the 12th item of the patent, including the "1.4 Smoothing of the dynamic image of the smooth image, read it The image is subjected to a smoothing process. 20. The method of de-interlacing the image & smoothing process as described in item 19 of the patent application system, wherein the smoothing process may be a mask smoothing process. The method for smoothing the secret of the wrong silk image described in item 2G is detailed, in which the smoothing of the mask of the adjacent real-field image can be performed by the Convolution Kernel method. 22. If applying for a patent The method of dynamic smoothing of de-interlaced images as described in the scope of item 20, which covers the masking theory of real-field images, can be calculated using the weight matrix = plus π-aMx). The concealment of the center of the neighboring real-field image is a convolution operation. 15