201232475 六、發明說明： 【發明所屬之技術領域】 本發明是，-歸像放大方法，制是種適合在魏中執行數 位影像放大’而獲得清晰放大結果的影像放大方法。 【先前技術】 在數位化景彡像應用日益廣泛，深入人類生活的今天，數位化影像的放大 (up-sampling)已經成為一種重要的應用技術。無論是業界或學界，對於影像 鲁放大的技術研究，都不餘遺力。但是，數位化影像的放大，是將像素數量 較小的影像，放大成為像素數量較多的影像，在放大過程當中因可供參考 的樣本不足，會發生像素的流失，因而造成鋸齒邊緣、影像特徵模糊，亮 度衰減等結果。因此，如何能夠獲得清晰的放大結果，乃是業界所追求的 目標，並成為影像放大技術的重要課題。 在已經發表的技術中’ Freeman等人發表了以learning based為基礎的 影像放大方法，其作法可參見 Freeman et al·, “Learning low-level vision,” 鲁 IJCV，2000。Dai等人則發表了稱為soft edge prior的方法，可參見Dai et al.， “Soft edge smoothness prior for alpha channel super resolution，” 于4 於 Proceedings of IEEE Conference on Computer Vision, 2007 o Chang 等人貝ij 發 表過以neighbor embedding放大影像的方法，參見Chang et al.， “Super-resolution through neighbor embedding,” IEEE Conference on Computer Vision and Pattern Recognition, 2004。另外，Yang 等人則發表一種201232475 VI. Description of the Invention: [Technical Field] The present invention is an image-amplifying method which is suitable for performing digital image amplification in Wei and obtaining a clear magnification result. [Prior Art] Digital image enlargement (up-sampling) has become an important application technology in today's digital imaging applications, which are increasingly used in human life. Whether it is the industry or the academic world, there is no spare power for the technical research of image enlargement. However, the enlargement of the digital image is to enlarge the image with a small number of pixels into an image with a large number of pixels. In the process of amplification, due to insufficient sample available for reference, pixel loss occurs, thereby causing jagged edges and images. Features such as blurred features, brightness decay, etc. Therefore, how to obtain clear amplification results is the goal pursued by the industry and has become an important issue in image magnification technology. In the published technology, Freeman et al. published a magnification-based image magnification method, which can be found in Freeman et al., "Learning low-level vision," Lu IJCV, 2000. Dai et al. published a method called soft edge prior, see Dai et al., “Soft edge smoothness prior for alpha channel super resolution,” in Proceedings of IEEE Conference on Computer Vision, 2007 o Chang et al. Ij has published a method for magnifying images with neighbor embedding, see Chang et al., "Super-resolution through neighbor embedding," IEEE Conference on Computer Vision and Pattern Recognition, 2004. In addition, Yang et al. published a
以 sparse representation 放大影像的方法，見於 Yang et al.方法(Yang et al., “Image super-resolution as sparse representation of raw image patches,” IEEE 201232475The method of magnifying images with the sparse representation is found in Yang et al. (Yang et al., “Image super-resolution as sparse representation of raw image patches,” IEEE 201232475
Conference on Computer Vision and Pattern Rcognition，Vol· 9, pp 1-8, June 2008)〇 上述方法分別從不同角度提出該善影像放大品質的方法，獲得較為清晰 的影像。但是所獲得的影像品質，仍然不盡人意。同時，這些方法在影像 放大倍率較高的時候，無法獲得清晰的品質。此外，若干技術在放大影像 時’需要提供外界的訓練集(training set)影像，難以自動執行。 本案的共同發明人與Shawmin Lei曾經提出一種特殊的内插法，可以用 來放大數位化影像。參見 Sreedevi et al., “An examplar-based approach for texture compaction synthesis and retrieval，，’ IEEE Transactions on ImageConference on Computer Vision and Pattern Rcognition, Vol. 9, pp 1-8, June 2008) 〇 The above methods respectively propose the method of magnifying the image quality from different angles to obtain a clearer image. However, the quality of the images obtained is still unsatisfactory. At the same time, these methods do not achieve clear quality when the image magnification is high. In addition, several techniques are required to provide an external training set image when magnifying an image, which is difficult to perform automatically. The co-inventor of this case and Shawmin Lei have proposed a special interpolation method that can be used to magnify digital images. See Sreedevi et al., “An examplar-based approach for texture compaction synthesis and retrieval,,” IEEE Transactions on Image
Processing，Vol_ 19, PP. 1307-1318, May 2010。該文提出一種雙立方内插法， 用來放大影像。於該雙立方内插法處理中，另包括以最佳匹配臨界值、平 滑化控制臨界值、本地動態(亮度)調整值、鄰近像素標準偏差值等參數，調 整β亥雙立方内插法計算。該方法雖然可以提供相當清晰的放大影像但其 適用範為偏重在同質性較高或有規律性的影像，對於非同質性的影像，處 理效果仍不理想。 因此目則確有必要提供一種新穎的影像放大方法，可以獲得高度清晰的 放大影像。 同時也有必要提供一種不需額外資訊，可以自主執行影像放大的影像放 大方法。 同時也有必要提供—種可關來放大非同質性影像，獲得清晰的放大結 果的影像放大方法。 201232475 本發明的目 放大影像。 勺p在提卩種新賴的影像放大方法，可以獲得高度清晰的 資訊’可以自主執行影像放大的影 本發明的目的也在提供—種不需額外 像放大方法。 本發明的目的也在提供—種可來放大賴質性影像，獲得清晰的放 大結果的影像放大方法。 【發明内容】 #本發明提出一種三階段的影像放大方法，用以在-電腦設備中將-數位 化影像放大(UP sampling)。在第_p嫩將做理影像作縮小取樣，在第二 階段從該縮小取樣的影像中，取得—誤差影像。在第三階段麟該待處理 p像逐步放大’直峨骑需放大倍率並_錯誤影雜正該放大的待 处〜像其中。亥為差衫像為該待處理影像與該縮小取樣影像相減的結 果其中，心差影像也可為該待處理影像與該縮小取樣影像放大到相同 倍树’減的絲。該修正贿理影像之步驟，職大過程中或放 大完成後進行。 在本發明的實施例中，該縮小取樣的倍率較好為1/1〇_謂之間，最好 為 1/2 〇 在本發明的實施例中’該誤差影像放大時，可包括以内插法放大。在其 中某些實施例中’減大方法為_雙立方内插法(bieubiein卿。丨ati〇n)。 在本發明的實施例中，尚可包括一修正該誤差影像的步驟，該步驟包括 一銳化(sharpening)該誤差影像的處理。其中，該銳化處理可包括以至少之 201232475 醜化程度，航輸絲像，續肢少2财同·程狀誤差影像 以及將該至少2種不同銳化程度之誤差影像之特徵，加人該誤差影像之步 驟。其中’駐少2種柯航域之誤絲像之特徵，加入該誤差 影像之步驟，可包括一最佳匹配處理(bestmatching)。 在本發明的某些實例中，該待處理影像進行放大處理前，可另包括一將 該待處理影像銳化之步驟。其中，該銳化處理可包括—將該影像作低度銳 化處理之步驟。 在本發明的某些實例中，該待處理影像的放大處理，可包括一内插法處 理。其中，該内插法可包括-雙立方内插法處理。其中，該雙立方内插法 處理中，可另包細最健配臨界值、平滑化控繼界值、本地峨亮度） 調正值、鄰近像素標準偏差值等參數之至少一種，調整該雙立方内插法計 算的步驟。 本發月的方法尚可包括在完成修正及放大後，將所得影像銳化的步驟。 上述本發明的目的與優點，可從以下詳細說明並參照圖式，而更形清楚。 【實施方式】 本發明的影像放大方法主要包括三大步驟階段4先，取得—待處理影 像於第步驟p白丰又從s亥待處理影像縮小取樣(d〇wn幼师此幻，獲得一像素 數置較少之縮小取樣影像n從該縮小取樣影像巾抽取—誤差影像， 代表趟像放大時會產生的錯誤部分。於最細段以_法將該待處理影 像放大’並去_放大影像的錯誤部分，得到效果良好的放大結果。 第1圖表示本發明衫像放大方法之流程圖。如圖戶斤示，於⑼電腦設備 201232475 先取得-待處理影像！。於1〇2將該待處理影像t作取樣，以獲得—降低像 素數里之取樣純⑽小取樣，丨㈣。在進行此步啊，可以任和 I式對該贿·像取樣，只要《得像素數量較少的影像。在作法上壬通可 常可以任—比例’在該待處理影像中擷取其像素，該比例通常可為丨/1〇 = 〇視衫像大小(像素數量）、系統資源成本以及所需之解析度等條件而 異。局於1/1G或低於1Λ㈣比例，也可適用在本發明。Processing, Vol_ 19, PP. 1307-1318, May 2010. This paper proposes a double-cube interpolation method to magnify the image. In the double-cube interpolation process, the parameters including the optimal matching threshold, the smoothing control threshold, the local dynamic (brightness) adjustment value, and the adjacent pixel standard deviation value are adjusted to adjust the β-Huicubic interpolation method. . Although this method can provide a fairly clear magnified image, its application is biased towards images with higher homogeneity or regularity. For non-homogeneous images, the processing effect is still not satisfactory. Therefore, it is indeed necessary to provide a novel image enlargement method to obtain a highly clear magnified image. At the same time, it is also necessary to provide an image enlargement method that can perform image enlargement independently without additional information. It is also necessary to provide an image magnification method that can be used to amplify non-homogeneous images and obtain clear magnification results. 201232475 The enlarged image of the present invention. Spoon p is a new method for image enlargement, which can obtain highly clear information. The purpose of the invention can be autonomously performed. The purpose of the invention is also to provide an additional image amplification method. It is also an object of the present invention to provide an image enlargement method which can magnify a gradation image and obtain a clear amplification result. SUMMARY OF THE INVENTION The present invention proposes a three-stage image enlargement method for amplifying a digital image in a computer device. In the first _p, the image will be reduced for sampling, and in the second stage, the image will be taken from the reduced sampled image. In the third stage, the lining should be processed. The p-image is gradually enlarged. The 峨 峨 需 需 放大 放大 并 并 并 并 并 _ _ _ _ _ _ 〜 〜 〜 〜 〜 〜 〜 The result of subtracting the image from the image to be processed is subtracted from the image to be processed, and the image of the heartbeat may be enlarged to the same tree by the subtracted image. The step of correcting the image of the bribe is carried out during the course of the university or after the completion of the enlargement. In the embodiment of the present invention, the magnification of the downsampling is preferably between 1/1 〇 _, preferably 1/2 〇. In the embodiment of the present invention, when the error image is enlarged, the interpolation may be included. The method is enlarged. In some of these embodiments, the method of reduction is _double-cube interpolation (bieubieinqing.丨ati〇n). In an embodiment of the invention, a step of correcting the error image may be included, the step comprising a process of sharpening the error image. Wherein, the sharpening process may include at least a 201232475 degree of ugliness, a flight image, a continuation of less than 2 different modes, and an image of the error image of the at least 2 different sharpening degrees. The steps of the error image. Among them, the feature of the two types of ergonomic images in the ke field is that the step of adding the error image may include a best matching process. In some examples of the present invention, the image to be processed may further include a step of sharpening the image to be processed before performing the enlargement process. The sharpening process may include the step of performing the image sharpening process. In some examples of the invention, the magnifying process of the image to be processed may include an interpolation process. Wherein, the interpolation method may include - double cubic interpolation processing. Wherein, in the double-cube interpolation process, at least one of the parameters such as the finest and most healthy matching threshold value, the smoothing control threshold value, the local 峨 brightness value, the correction value, and the adjacent pixel standard deviation value may be further adjusted, and the pair is adjusted. The step of cubic interpolation calculation. The method of this month may also include the step of sharpening the resulting image after the correction and amplification are completed. The above objects and advantages of the present invention will become more apparent from the detailed description and appended claims. [Embodiment] The image enlargement method of the present invention mainly comprises three steps of step 4, and the image to be processed is subjected to the first step p, and the image is reduced from the image to be processed (d〇wn kindergarten teacher, obtaining one) The reduced-sampled image with a smaller number of pixels is extracted from the reduced-sampled image towel—the error image, which represents the error portion that is generated when the image is enlarged. The image to be processed is enlarged by the _ method in the thinnest section and is zoomed in. In the wrong part of the image, the result of the magnifying effect is obtained. Fig. 1 is a flow chart showing the method of enlarging the shirt image of the present invention. As shown in Figure 9, the computer device 201232475 first obtains the image to be processed! The image to be processed is sampled to obtain a sampled pure (10) small sample in the reduced number of pixels, 丨 (4). In this step, the sample can be sampled in the form of I, as long as the number of pixels is small. The image can be used in the process - the ratio 'takes pixels in the image to be processed, the ratio can usually be 丨 / 1 〇 = 衫 像 image size (number of pixels), system resource costs and Required Analysis of other conditions. Bureau on 1 / 1G or below 1Λ㈣ ratio is also applicable in the present invention.
本例中’該取樣比例訂為^，也就是將該待處理影像的像素，每2 個相鄰像素作為—組，取其中—像素，將所取得的像素組合成為—縮: 樣的影像〇，D=i/2⑴。 於步驟將該縮小取樣的影像放大，成為原來的尺寸得到映射影像 助。在放大時，可以將該縮小取樣的影像，直接按原比例放大，成為該映 射影像。但在本實施财，則是以通常稱為「雙立方内插法批灿ic __ati〇n)」的方法’將該影像放大。該雙立方_放大方法是以—似 的矩陣内的像素計算出位在該4職陣最中心的未知像素之值。該方2 有岐可以得顺清晰的放大影像，但其缺闕是計算較為繁複費時。在 本實例中，魏小取樣的影像經過放錢，成Μ射影像处，^扣。 接者進行誤差影像的取得。於步驟]〇4將該待處理影像『與該 Rb相減’得到誤差影像E。E=I-Rb。 〜 古在T步驟中，尚可包括一修復該映射影像Rb之步驟。該步驟乃是將 刪樣像Rb以最佳匹配法(best她hing)處理。其方式是將該映射影像 处與該待處理影像I，—最低銳化節L，_中等銳化影像Μ，及—?最高 201232475 尖化影像_t最佳㈣，贿魏包含細議最健配。取得該最 低纖麻，中等銳化影像IM，及最高銳化影像m的方法包括以「非 屏障銳化」處理(unmask sharpening)，得到不同程度的邊緣銳化影像。其中， 依照所得邊緣銳化程度，分別訂為最低銳化影像il，中等銳化影_，及 最尚銳化影像IH。 在本發明的實例中，可將銳化時的參數，作如下設定：對最低銳化影像 IL，設定其半徑控制參數㈣㈣為3 〇(像素），銳化程度值咖_咖6 ; 對中等銳化影細，則設定其半徑控制參數為3 〇(像素），銳化程度值為 〇·9 ;對最高銳化影像IH，則設定其半徑控制參數為3 〇(像素），銳化程度值 為1_2。當然’其他參數設定，也可適用在本發明，只要能夠得到3種不同 銳化程度的影像 <=而該銳化程度，自然也不限於3種。少於3種或多於3 種，均無不可。全視所需影像品質與電腦處理成本決定。另外，在進行該 最佳匹配處辦’可將其臨倾設定為4，代表熟麵界值，即平均方差 (mean square error)。實驗發現，該臨界值乃是在一最佳匹配結果與訊雜比之 間的權衡。如果設定太高，則影像中的雜訊將會增加，反之則不容易找到 最佳匹配4實施例將該值設定為4的結果，可使得到的修復影像平均結 構性近似度(structural similarity-SSIM)達到0.95左右，而其峰值訊雜比 (peakSignalton〇iserati〇_PSNR)均優於習知技術。當然，如前所述，該值 的設定乃是-種獅的結果，高於或低於4，皆可適用在本發明，並非任何 技術上的限制。 在這種實财，將齡郷像Rb⑽最佩化雜IL，巾魏化影像 201232475 IM’及最高銳化影像IH進行最佳匹配，即得到一修復的 驟104將該待處理旦，你τ ή 〜像R。於步 衫像I /、該映射影像R相減，得到誤差影像e。叫士。 ,、後本發明之方法進入影像放大的階段。於此階段中 =將該她1影像1放大1倍數。在此，所―無—定的 次放大到所需倍數，或以數步驟逐心^ 心數。在本發明的實施例巾 所需倍數。在這嶋之下 ⑽，嶋步驟放大到 用^ 之下，母姐大倍數赠大㈣倍至Μ倍較為適 二以放大㈣倍在處理效果與運算軸^ 仁並非任何技術上的限制。 在此步射，，也可额經過初步處理的待處簡，作為該待處理影 顺化的最低舰f彡像1L，__ im，及最高銳化 :田之一者。該等影像已經出部的銳化處理，可以提供較優異的放 "、中’在本發明的較佳實例中，採用該最低銳化影像IL，得到較 為優異的效果。 ❿ 在執行該放大處理時’可以使用任何已知的放大技術。例如業界常用的 内插法，即侧。域編物術，娜侧嫌制。只要 能將該影職大，朗舦絲柯。耐本發明_佳實咐，是建議 ^用上述雙立方缝法’作域⑽處理方式。這種方式對於經過銳化的 衫像，能夠提供優異的放大品f。妓其他的放大方法，仍可制在本發 明。 在上述雙立方内插法的處理步驟中， 可以使用若干參數值，來改良放大 201232475 的效果。例如前述Sreedevi等人的論文中所提出的調整參數，即最佳匹配 臨界值、平滑化控制臨界值、本地動態調整值、鄰近像素標準偏差值等， 即可用來調整該雙立方内插法的計算，獲得更優異的放大影像品質。其中， 该最佳匹配臨界值(threshold limit)是用來控制相關像素之值，使其不落入誤 差範圍。s亥平滑化控制臨界值(sm〇〇thcontr〇丨threshol〇d)是用來控制使平滑 化處理的不致過向。該本地動態調整值(l〇cal dynamics)是用來調整像素之亮 度。该鄰近像素標準偏差值(_η _Γ 〇f neighb〇ring啦也)為從該映射影像 經過最佳匹配處理後，所得的值。 其次’在步驟106將該誤差影像e也放大相同倍率。該放大方法可使用 任何已知的影像放大方法，可與該待處理影像〗之放大方法相同或不同。於 步驟！〇7在該待處理影像中減除該誤差影像，得到修正後影像z，z斗E。 於步驟⑽判斷該修正後影像z的放大倍率是否已經達到所需倍數。如 果判斷結果為否，則步驟回到1〇5，再將該修正後影像放大_定倍率。如是， 則判斷修正完成，進入下一步驟。 在本發明中，频及107可與步驟1〇8互換順序。申言之，該待處 理影像可料錄大到所需倍數之後，相雜差影正初經放大的处 待處理影像。但因本實娜取逐錢大的纽策略，_逐轉正的方式 可能較為適用。 經過上述朗之後’該f彡像已經放大顺需鲜，並_修正，成為夠 合需求的影像。在此可另外在步_將所得的影像，再度作銳化處理^ 以消除影像_繼，進-賴獅。如祕，賤化方法可以 201232475 使用任何已知的技術。於此不再贅述。 第2圖顯示林發·方法放A 3姆像的處理結果。在進行影像放大 時，所使㈣銳化軟體為業界常㈣·軟體。圖中顯示將圖a、b、c各 放大36倍’成為圖d、e、f的結果。顯示使用本發明的方法可以使放大的 影像仍然财清_品質。第3關顯示本發财法與心麟的效果比 較’其中’圖a為待處理影像，b為以丨earning㈣方法放大Μ倍的結果， c為以S〇ftedgeprior方法放大16倍的結果，d為以㈣娜。⑽祕叩方In this example, the sampling ratio is set to ^, that is, the pixels of the image to be processed, each of the two adjacent pixels is taken as a group, and the pixels are taken as the image, and the obtained pixels are combined into a reduced image. , D=i/2(1). In the step, the reduced-sampled image is enlarged to become the original size to obtain the mapped image. When zoomed in, the reduced-sampled image can be directly enlarged to the original scale to become the mapped image. However, in the present implementation, the image is enlarged by a method commonly referred to as "double-cube interpolation method 灿 __ati〇n". The double-cube-amplification method calculates the value of the unknown pixel located at the center of the 4-position array by pixels in a matrix. The party 2 has a clear and magnified image, but its lack of calculation is more complicated and time consuming. In this example, the image of Wei Xiao sampling is put into the image, and the image is taken at the image. The receiver obtains the error image. In step [4], the image to be processed is "subtracted from the Rb" to obtain an error image E. E=I-Rb. ~ In the T step, a step of repairing the mapped image Rb may be included. This step is to treat the deletion as Rb with best matching (best her hing). The method is that the mapped image is located with the image to be processed I, the lowest sharpening section L, the _ medium sharpening image Μ, and the highest highest 201232475 sharpened image _t (four), bribe Wei contains the most detailed Match. The method of obtaining the minimum fiber, the medium sharpening image IM, and the highest sharpening image m includes "unmask sharpening" to obtain different degrees of edge sharpening images. Among them, according to the degree of edge sharpening, the minimum sharpened image il, the medium sharpened image _, and the most sharpened image IH are respectively set. In the example of the present invention, the parameter at the time of sharpening can be set as follows: for the lowest sharpened image IL, the radius control parameter (4) (4) is set to 3 〇 (pixel), the sharpness value is _ coffee 6; To sharpen the shadow, set the radius control parameter to 3 〇 (pixels) and the sharpness value to 〇·9. For the highest sharpened image IH, set the radius control parameter to 3 〇 (pixels), sharpening degree. The value is 1_2. Of course, other parameter settings can also be applied to the present invention, and as long as three different sharpening images can be obtained, the degree of sharpening is naturally not limited to three. Less than 3 or more than 3 are not available. The image quality required for all-view and the processing cost of the computer are determined. In addition, in the case of performing the best matching, the inclination can be set to 4, which represents the cooked boundary value, that is, the mean square error. The experiment found that the critical value is a trade-off between the best match result and the signal-to-noise ratio. If the setting is too high, the noise in the image will increase, otherwise it is not easy to find the best match. 4 The result of setting the value to 4, the average structural approximation of the obtained repair image (structural similarity- SSIM) is around 0.95, and its peak signal ratio (peakSignalton〇iserati〇_PSNR) is superior to conventional techniques. Of course, as mentioned above, the setting of this value is the result of the lion, higher or lower than 4, which can be applied to the present invention, and is not a technical limitation. In this kind of real money, the age of the Rb (10) is the most suitable for the hybrid IL, the towel Weihua image 201232475 IM' and the highest sharpened image IH for the best match, that is, a repair step 104 will be processed, you τ ή ~ Like R. In the step shirt image I /, the map image R is subtracted, and an error image e is obtained. Called a gentleman. The method of the present invention then enters the stage of image enlargement. At this stage = magnify her 1 image 1 by a multiple. Here, the “none” is zoomed in to the desired multiple, or in a few steps. The multiples required in the embodiments of the present invention. Under this circumstance (10), the 嶋 step is enlarged to use ^, the mother-in-law multiplies the big multiple (four) times to Μ times more suitable to enlarge (four) times in the processing effect and the calculation axis ^ Ren is not any technical limitation. In this step, it is also possible to pass the preliminary processing of the standby, as the lowest ship to be processed, 1L, __ im, and the highest sharpening: one of the fields. These images have been subjected to sharpening treatment, and it is possible to provide a superior discharge. In the preferred embodiment of the present invention, the minimum sharpened image IL is used to obtain a superior effect.任何 Any known amplification technique can be used when performing the amplification process. For example, the interpolation method commonly used in the industry, that is, the side. Domain editing, Na's side is suspected. As long as the film can be a big job, recite the silk. It is recommended to use the above-mentioned double-cubitch method as the domain (10) treatment method. This method provides an excellent magnification f for a sharpened shirt image. Other amplification methods can still be made in the present invention. In the processing steps of the above-described double-cube interpolation method, several parameter values can be used to improve the effect of enlarging 201232475. For example, the adjustment parameters proposed in the aforementioned Sreedevi et al. paper, that is, the optimal matching threshold, the smoothing control threshold, the local dynamic adjustment value, the adjacent pixel standard deviation value, etc., can be used to adjust the bicubic interpolation method. Calculate for better image quality. The best matching threshold is used to control the value of the associated pixel so that it does not fall within the error range. The s-hai smoothing control threshold (sm〇〇thcontr〇丨threshol〇d) is used to control the smoothing of the smoothing process. The local dynamic adjustment value (l〇cal dynamics) is used to adjust the brightness of the pixel. The adjacent pixel standard deviation value (_η _ Γ nef neighb〇ring) is the value obtained after the best matching processing from the mapped image. Next, the error image e is also magnified by the same magnification in step 106. The amplification method may use any known image enlargement method, which may be the same as or different from the enlargement method of the image to be processed. In step! 〇7 subtracts the error image from the image to be processed to obtain the corrected image z, z bucket E. In step (10), it is determined whether the magnification of the corrected image z has reached a desired multiple. If the result of the determination is no, the step returns to 1〇5, and the corrected image is enlarged by a predetermined magnification. If yes, it is judged that the correction is completed and the next step is entered. In the present invention, the frequency sum 107 can be interchanged with the step 1 〇 8. According to the statement, after the image to be processed can be recorded up to the required multiple, the image is processed by the initial difference. However, due to Ben Senna's strategy of taking advantage of the big money, the _ turnaround method may be more applicable. After the above-mentioned lang, the image has been enlarged and replaced with _ corrections to become an image of sufficient demand. Here, you can additionally sharpen the resulting image in step _ to eliminate the image _ subsequent, into the lion. As secret, the method of deuteration can use any known technology in 201232475. This will not be repeated here. Figure 2 shows the results of the processing of the A3 image by the Linfa method. When performing image enlargement, the (4) sharpening software is the industry (4) software. The figure shows the result of magnifying 36 times each of Figs. a, b, and c into graphs d, e, and f. It is shown that the magnified image can still be saved _ quality using the method of the present invention. The third level shows the comparison between the effect of this method and Xinlin's 'where' picture a is the image to be processed, b is the result of magnification by the 丨earning method, c is the result of 16 times magnification by the S〇ftedgeprior method, d For (four) Na. (10) Secret Party
法放大16倍的結果，e為以sparse represemati〇n方法放大μ倍的結果，『 為以本發明方法放大16倍的結果。顯示本發明的方法可以得到較之習知技 術更清晰的放大影像。 事實上’使用本發明的方法，不僅可以得到清晰的放大效果，同時因為 在影像的放大與修正中，只朗由待處理影像所產生的訓練餘如㈣对) 影像，不紐科來_丨練娜像，因此可以達到自主處理，不需選擇適 用的s;i丨練集影像，同時並可節省處理時間的優點。 以上疋對本發明影像放大方法實施例的說明，用來例示本發明。習於斯 藝之人士不難由以上制’作出各種變化與触。只要不超出本發明申請 專利範圍記載，都屬於本發明之專利範圍。 【圖式簡單說明】 第1圖表示本發明影像放大方法之流程圖。 第2圖顯示以本發明的方法放大3個影像的處理結果。 第3圖則顯示本發明方法與習知技術的效果比較示意圖。 201232475 【主要元件符號說明】 無The result was magnified 16 times, and e was the result of a magnification of μ times by the sparse represemati〇n method, "the result of amplification by 16 times by the method of the present invention. The method of the present invention is shown to provide a magnified image that is clearer than conventional techniques. In fact, using the method of the present invention, not only can a clear magnification effect be obtained, but also because in the enlargement and correction of the image, only the training residual generated by the image to be processed is as follows (four) pairs), not Newcomer _丨The Naina image can be processed autonomously, without the need to select the applicable s; i 丨 丨 影像 image, while saving the processing time advantages. The above description of the embodiment of the image enlargement method of the present invention is used to exemplify the present invention. It is not difficult for people who are in the arts to make various changes and touches. It is within the scope of the invention as long as it does not go beyond the scope of the patent application of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a flow chart showing an image enlargement method of the present invention. Fig. 2 shows the results of processing in which three images are enlarged by the method of the present invention. Fig. 3 is a schematic view showing the comparison of the effects of the method of the present invention and the prior art. 201232475 [Main component symbol description] None