TWI255140B - Caption detection and compensation for interlaced image - Google Patents

Abstract

This disclosure presents a video de-interlacing, which relates to caption detection and compensation for an interlaced image. Since the pattern of caption is fine and changes more complicatedly than other objects within the same image, it is difficult to obtain a qualified caption de-interlaced result via common video de-interlacing. For a caption that moves horizontally, the disclosure detects a caption region of continuous scan lines containing caption pixels on a san-line basis. Next, a brightness histogram for pixels within the caption region is analyzed to obtain a brightness bound, and caption pixels are identified by reference to the brightness bound. Next, un-filled pixels within the image are motion-compensated by reference to corresponding pixels in preceding/succeeding fields, and the compensated caption region is further examined to eliminate undesired distortion.

Description

1255140 IX. Description of the Invention: [Technical Field of the Invention] The present invention relates to a deinterlacing method for interlaced images, which is a method for compensating for subtitles. ^There are _fine interlaced images in the word [previous technique] The sequential image device is non-interlaced (NGn_interla (10) progressive scan (ProgressiveScan), such as the computer's Gugan 5 § 5 a ...., α for the cycle * 敫 fine (10), presented in Image on the computer screen, =J_(Frame) with - line - line from top to bottom «; 3^:;? J"75ΗΖ(## 75 ^ 柊 柊 乂 乂: 衫 像 like device ' as traditional TV, using the interlaced image display error _ 2 or more than the tear (10) to represent, such as the map field, tyrant (object (4) field, 3 〇 two 曰, work two *, sequentially these two The field is repeatedly scanned and inserted dozens of times per second (the NTSC system is "_" This Weizhong will feel that it is related to the ability of the human eye to receive images. The human eye is fine for every fit, and then the sweep is improved. Second, the Z day will become inconspicuous. In the 1930s when the TV system specification was formulated, it was impossible to afford such a large amount of money for the scanning of such a high scanning frequency. Using interlaced scanning, you don’t need to update 60 frames in two or two seconds. In order to meet the demand for image quality acceptable to the human eye. However, the technology of the passer-by is changing with each passing day. Now the sequential scanning display technology relies on updating (10) times per second, and the problem of insufficient transmission rate and bandwidth is already light and secret. The reason for the initial interlaced scanning has been lost, L is due to the long-term use of the TV system, and the video recording device is also interlaced with image-to-record images. In order to convert the image of the interlaced image into a sequential Image gauge 1255140 Image playback in the device, there is an important problem that we need to solve, that is, the conversion. When we want to convert the interlaced image into a sequential image, the field and the image field are recorded at different time points. If the shadow to be displayed will be * mi, the static image is like the 'mQtiQn', so as long as the straight field is combined into a frame, you can get the complete image. 'However, if 32 =;! If you don't move, there may be an offset between the fields. If you directly combine the fields into a picture I: the resulting frame may produce such as a tooth edge (_(1) plus), a line (10) e Crawllng) Comfortable distortion (artifact s). Multi-solutions, Π Π 鳞 骑 骑 骑 骑 骑 像 像 像 像 , , , , , , , , , , , , , 许 许 许 许 许 绿 绿 绿 绿 绿 绿 绿

Lr ft^r^(intra'field interpolati〇n)' = Qing (four), moving image adaptive deinterlacing adaptive

Clng) and so on. The interpolation method in the field is for the same field to be spatially Ιί ϋ discarded. The two scanning lines adjacent to the selected field are averaged to obtain the complete picture. Inter-field interpolation method is used to convert the odd-field and the side-field into a complete range of '===== moving images in time for different map fields, and decide to do nothing to convert to a sequential image. If the ratio of moving images between the graphs is low, the direct shed will be combined; otherwise:::::7Γθ1Γ 0 lion mines and corrections 4 silk scale vectors (4) i〇n vector), sub-transparent image orientation operation to achieve solution Interlaced purpose. The above-mentioned transfer of wealth has its own secrets, such as the image and image distortion when the image is scanned and edited, the code = in the vertical «Ξ2:, ί: ί;:;: ; :^:::::rblurdng) ° hard Miscellaneous touch on the material research.姆 调 较 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 ^ The image processed by the interleaving method is a general animated image. However, if there is a subtitle in the image = for example, there is often a marquee subtitle below the news screen, because the edge of the text (·) is the subject ^rt 'and - The image is secret. If the subtitles are blurred, or the text is 以, the phenomenon is == LCD monitor (10). Compared with the general monitor, although it is deinterlaced = like the second: but more for general moving images, it is not specifically targeted.

HlHir - the general law of the wrong and the characteristics of the tree and the amount of money, the mouth of this book is unveiled - the method of de-interlacing the subtitles of the image towel. [Summary of the Invention] The subtitles of the subtitles are for the subtitles of the left and right chickens, for example, the image of the marquee image below the screen =, 'The method of the invention is proposed - the method of deinterlacing, especially in 1? And 77 ^ knife does not have different motion compensation parts for the sub-screen and the non-subtitle image part < : The two parts of the job combination and the output (four) system are deinterlaced. Obtaining the method of mm' in the mention of (4) on the fiber rotation wheel, the process of the fruit curtain sweeping method is as follows, round the round, and analyze the pure line detection of the sub-screen of the word 暮 计 ,, and obtain the scan line definition including subtitles. The drawing axis of the ___ is a method of motion compensation for the feature of the screen finder; and the range of subtitles in the second image of the subtitle is checked and used. The method t_4b produced in the following discloses the processing of the shirt as the surrounding, and the condition of obtaining the deinterlacing is only an example, and the invention is not limited to the invention. 1255140 can only be implemented according to the following embodiments. The method will be described in detail below. However, except for the detailed description, which is implemented by the embodiment of the present invention, and the scope of the present invention is not limited, the main steps of the subtitle compensation deinterlacing method disclosed in the present invention (10) the scanning line containing the subtitle in the gamma field are first described. , "Miao = line markings come. Then analyze the words corresponding to the field before the interlacing of the field]

=2 The rough range of the subtitle block in the front field (Step 3=J is the pixel reduction in which the subtitle to be compensated is compensated by using the previous or the previous position and the prime position (step 140). When the operation is used as the subtitle compensation, we do not want to make the subtitles. This is not yet finished. 'When the subtitles are based on the image, the interlaced result is used to generate the distortion. The result of the deinterlacing is correct. 1W, the decision is a dynamic operation, and the combination of the two outputs is a figure for the subtitle block to be directly calculated, whether if the |^=cross_middle-picture field is deinterlaced compared to the current picture field and the front-frame phase 3 The range of the moving image in the field. We are the range of the moving image in the field, and the difference between the two is corresponding to the part of the area that may contain the subtitle part; (10)7 because the sub-screen moves left and right, The moving image is based on the scanning line. The more the horizontal movement of the subtitles can be obtained by shifting the left and right sides of the subtitle. The first observation is the scanning line, if it is scanned (four) by the scanning line t belongs to the moving line containing mobile subtitles The morpheme can be inferred that I should account for a certain proportion of 5 hai, and a large part of the sputum is not in the area of the area where the difference in the value of the 曰昼1255140 / month is very small. Suppose the moving image in the scan line _ Two = 3 can be used as the number of subtitles for the subtitles N, "and M > W / 3 and N > W / 2 in the smooth area, the width of the towel w is wide \ \ then the initial judgment of the fine level of the pot. Only when a =7 line satisfies the upper secret part, 'will perform the step-by-step operation on the scan line (the subtitle scan line is said to be "1. Here must be _ tune, find the moving picture picture _ method and boots In the above method, and the ratio of the above-mentioned specific conditions is _, _, only _ feasible _ green, limited. If the scan line meets the above conditions, the number of scalar pixels of the scan line is consistent with the image of the premise ί It Step by step f county line is pure subtitle part, you need to get the characteristics of the movement and the speed of the movement. In this embodiment, the subtitle system is left to use these features to help distinguish the subtitles. Referring to the second map, the frame is composed of two places (such as the NTSC system), and the current interlaced image field is the second two = The reference picture field should be the n-2th field. Because of the subtitles in the image, right == one of the pixel blocks in the field n can be in the block of the search window =. ",", "Χ" indicates the position of the field in the previous picture. ^The field η is empty (Null). In this example, the size of the search window 26 is set to 2, which can be used to move the subtitles. For the block 2() of the field η towel, in the position of the field η - the block 20, the search window 26 is used as the range, and the block 22 with the block by = is searched horizontally. The η rhyme block 22 is compared with the block 2 图 of the field η, which is the moving image vector of the corresponding position of the block 22 (such as the vertical axis), and ^ν is expressed as 4). If the subtitle moves at the same speed in the horizontal direction, the displacement between the front field w and the image can be reasonably inferred to be half of the moving image vector mv, that is, mv/2 (the figure "^ = in other words, the field n-1" Medium - the scan line containing the subtitles, the position of the line _ 昼 对应 对应 图 图 图 η η η η η η η η η η η η η η η η η η η η η η η η η 255 255 255 η η 255 255 Should be left to the left of the axis 2 block 21. The information heart of the word =_^= vector can be used to confirm whether the scan line contains the characteristics of the text. For example, we set several conditions to filter the pixels that may contain subtitles as the subtitles of the subtitles, adjacent pixels. Miscellaneous may be the pixel of the subtitle ^ knife' because the movement of the subtitle element should be ^1. Because there is a phase, _ sifting tools::=:s4—嶋-phase'== In summary, for a scan line, (N>W/2) and (M>W/3) and (10) can be used. /4) and (s> (4)/4) ^ := Ground _ line is judged to contain subtitle _ scan line, and _ line mark _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The meal line is really up and down one by one, and the next line is the sweep line on the second line. (If it is conditional, mark these lines as a word, which is consistent with the above line. _, _, ^r== 10 1255140 degrees of image, can also be a different number, does not limit the number of scan lines. 1 After the range of the scan line where /ni screen is located, then step by step Block ==-Fig. 13G) 'and the subtitle block position of the subtitle block_ (the first picture steps must be obtained after the completion of the steps must first obtain the subtitle brightness information (step - Figure 12 (four) retinoicin method can be made - a simple way to implement. Observed - some of the subtitles containing mobile words found that most of the subtitles have brighter bright ships, so you can use the redundancy value of 竺 = for the base The brain of the subtitle block ^= obtains the brightness reference value as the judgment reference. Referring to the third figure ί: the second field of the third field, the G'Q block 3G can be selected; F can be the smooth area of the subtitles. Mv, the previous picture field μ is opposite to the second:: the moving image between the gods n_2 and the field n is chat/2, according to the moving image vector; the moving image between the blocks 30 should be pasted and deleted Input ======Pen brightness as the reference value of the subtitle brightness is the front, the brightness of the bright pixel of the example is as the moxibustion wire is selected as the ratio of 11/16, that is, the first 5/16 of all the pixels. Base: .: sample (four) - value and = can be: = prime = pixel surrounded by reference material AB HgH can be up and down the approximate range of the range of the scan line found above the range of ^ ^ ^ = = block up and down, can be Step 42 in the first figure. Because some of the characters have a number of scanning lines as the upper and lower boundaries of the text block, the upper half or the lower half of the text occupy the 'two'?, 1, small, large, medium, -...etc., more than one, may not match step 1〇(^ ;^ or not at all, when the scan line of such text in the subtitle block, the condition of the line will be judged and the scan is judged It does not include subtitles, and the extension can reduce the misjudgment caused by the special situation. After obtaining the upper and lower boundaries of the subtitle block, it is then to find the left and right boundaries of the subtitle block. First, find the matching pixels for each scan line of 1255140. The brightness A is the reference value of the brightness, the leftmost and rightmost pixels of the common image vector (for example, mv/2) == (° left border 4!) = ί in the middle of the starting point The joint starting value of the common starting position of the entire subtitle block is the leftmost pixel that meets the above conditions in the common line of the closing screen; ?==jt=:== set,: (in this case, Liu) As the subtitle block ^^点^^篡 all the end of the financial end as the entire subtitle part ^ common start ^ fine set the correction, the method of the object curtain circle to get the range of the subtitle block, then further define the subtitle block The explicit-method's method uses the pixel of the pixel's reference value within the subtitle block as the subtitle pixel, and the &=== /1 is the subtitle pixel position method_sub. Referring to the fourth map, 2 ! ί The subtitle block is empty inside the smear line, and is intended to compensate for the book. The book tiifi is judged as a subtitle pixel. Or centered on the request, using the search blocks 45 and 46 of i:=x= as the reference, if the search block is to be compensated = 46_--- (eg, three) pixels are larger than - the brightness reference value is often On the = image, the inside of the pixel, will be compensated for the pixel 47 mark is a subtitle pixel. The 'degree-of-degree I value limit is less than the reference value of the _brightness value, can be based on =) not the same ^ change" search t block - size position and the standard greater than the brightness reference value limit (example; _1. / Qin Mu block after the subtitle pixel position, and then the subtitle pixel position is disclosed in the present invention - subtitle compensation method, based on the pixel's moving image vector, the defined subtitle 12 1255140 prime = first via The picture of the previous or previous picture field corresponds to the operation of the picture, and the image of the aL is fflV/2. The _ front-picture field η-1 or the front and back field η-1 is The pixel of the word 暮t is _, that is, the pixel 5〇1 or the pixel 50 503, which is used as a reference for the image. The pixel 501 is directly copied or the pixel 50 503 is made: word== η: the value of the pixel to be compensated 5〇2. For all the pixels in the subtitle block, the sub-screen complement & the gift, that is, the result of deinterlacing the subtitles. The picture field or the front and back of the secret field of the field, but the compensation operation can also be expanded to include the information of the front and back _ field, the former i picture compensation method is also rare in the secondary _, for example, can also take ====: After 2°.6 ' .4) plus the weight of the total. The image of the compensation method is included in the subtitle of the present invention. The result of the verification of the final _ 'required verification check (step 150) is broken. In some special situations, for example, *^ corresponds to the corresponding %% and _ no subtitles Jian'. If the subtitle image of the crane is not implemented, the invention reveals that the anti-riding makes the subtitles lose. Knowing the subject _ field and the first two pictures before the subtitle compensation (that is, there is a staggered displacement phenomenon between the odd _ fields), so only the text of the subtitle block can be divided into several text areas. Block, it is too frequent to see if the situation is too large. First, the brightness of the two subtitles belonging to the subtitle is converted to 1, and the remaining pixels are converted to 0. The pixel transformation of the same-intercept (c〇lumn) in the text sub-region is too frequent, that is, the pixel (4) 1 appears interactively. The text of the job can be generated by improper subtitle compensation. In the text area=Fang 13 1255140, there is a ^^ phenomenon, indicating that the compensation result is not ideal, and L shift ==:= is established to indicate that the original current field and the front-picture field are established. The output image is the de-interlaced output of the subtitle image by combining the front field with the front-picture field (or the back-picture field). As described above, the image is deinterlaced in the image system, and the image is deinterlaced in the system, which can be = Η Η = = = = (10) information; and _fL will include: moving image area; Image, using the above method to solve the _ _ _ 靡 non-caption part:: sub-image _: = 1: species, de-interlacing method implementation (step _; finally word " sub-mouth knife like the father of the wrong The result is a combination of outputting the complete de-interlaced image coffee. It can also be applied, 'However, the present invention, its == two =^ == subtitles_image compensation principle does not have (four) difference, so the detailed description will be given to the invention The preferred embodiment is not intended to limit the application of the present invention. 1, the social wealth, for the well-known person who is familiar with the technology dr# can be clear and real [schematic description] 14 1255140 first The figure is a flow chart of the method proposed by the invention; the second picture A is a schematic diagram of obtaining a picture moving picture vector; the second picture B shows a subtitle range in the picture field; the third picture is a schematic diagram of obtaining a subtitle brightness reference value; The fourth A picture is not intended to obtain the rough range of the subtitle block, and the fourth B picture is to obtain the position of the subtitle pixel. In the meantime, the fifth figure is a schematic diagram of an embodiment of subtitle compensation; and the sixth figure is a flow chart of another embodiment of the present invention. [Description of main component symbols] 100, 110, 120, 130, 140, 150, 160: steps 20, 21, 22: Block 26: Search window 28: Subtitle range 30, 31: Block 41: Upper boundary 42: Lower boundary 43: Left boundary 44: Right boundary 45, 46: Search block 43 432, 433, 451, 455, 46 480: 昼素 5 (U, 503 · 昼 1 1255140 502: To compensate for 昼 600: interlaced image 650: deinterlaced images 610, 620, 630, 640: steps

Claims (1)

1255140 X. Patent Application Range 1. A method for subtitle detection compensation includes: inputting an interlaced image, which is a sequence of consecutive multiple frames, and the frames are composed of odd and even fields interlaced Execute-subtitle scan line side, the _ subtitle scan line range of the current map field; analyze the brightness of the plurality of pixels of the previous picture field corresponding to the subtitle scan line range to obtain a f-screen brightness reference value; The brightness reference value defines a subtitle range and a subtitle pixel position of the current picture field; a motion picture compensation is performed on the subtitle range; and a subtitle compensation output is obtained by performing a subtitle compensation check. Please patent _1 lion's word __ method, which looks for the subtitle scan line to compare the current picture field with the previous picture one corresponding field to get a moving image area; «The difference between the scan lines in the moving picture area The motion picture vector of each pixel is based on the motion _ each of the 昼 昼 向量 向量 向量 vector mosquito contains the word 暮 4 scan line; and the continuity of the scan lines including the subtitles is checked. The method for subtitle detection compensation according to item 2 of the moving range, wherein the plurality of books D are in the form of a shift between the target field and the corresponding field. ', Utr Patent Paragraph _2 item subtitle fine compensation method, its towel decision (four) 箄 spoon eight ϋ 线 line age in the moving image area of the plurality of pixels of the moving picture vector symbol r sigh conditions The scan line is judged to be the scan line containing the subtitles. Milk* 付合一17 1255140 5· The method for subtitle detection compensation according to Item 1 of the patent application scope, wherein the subtitle brightness reference value is used to analyze the brightness of the pixel corresponding to the subtitle scan line range of the previous picture field. The ratio of the brightness of one of the plurality of pixels of the range of the subtitle scan line is the brightness of the pixel as the reference value of the subtitle brightness, and the ratio is between 〇1. 6) The method for subtitle detection compensation according to the scope of claim 2, wherein defining a subtitle range and a subtitle pixel position according to the subtitle luminance reference value comprises: obtaining the current field and a subtitle area according to the subtitle luminance reference value Block range; when the subtitle block norm scan line--the luminance of the pixel is greater than the reference value of the vocalization brightness, the pixel is judged to be a subtitle pixel; and the thief of the subtitle area is checked. , bullying the plural to compensate for whether the picture is a subtitle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The rightmost plurality of pixels are processed-processed to obtain the upper, lower, left and right boundaries of the sub-screen block range. • The method of subtitle_compensation according to item 7 of the patent application, wherein the subtitle condition comprises: the brightness of the denier is greater than the reference value of the subtitle brightness; and the moving image of the aldarin is to the scan line of $ Common motion vector. 9. The method according to claim gamma method gamma method, wherein the processing is as follows: taking the middlemost value of the leftmost pixel and the rightmost side of the plurality of sweeping lines satisfying the subtitle condition The intermediate value of the book element is used as the left and right boundary of the subtitle block; the leftmost picture of the leftmost pixel of the subtitle condition that satisfies the subtitle condition is taken as the leftmost picture of the rightmost picture. Sigh 18 1255140 Take the plural scales, the caps, the words, the most pictures of the paintings, and the value of the rightmost picture. ==^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ More than one subtitle brightness limit tear _ wheel surface age n cap subtitle brightness limit is the sub-screen read reference value - smaller than paper, m case between (four) to i. A method for detecting the subtitle detection compensation, wherein the motion image compensation comprises: taking a pixel corresponding to the image of the front or the front and back of the image, and the front or the money field The image of the image to be compensated is obtained by the calculation of the image of the image. Face 12 _ turn ___, (4) operation for the following copy of the pixel to be compensated in the place of the previous map, the desired face painting 辩 触 触 U U U U U U U U U After the _w_ is registered after the occupation, the weight is added after the weighting; and the group that wants to delete the ___ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Checking the sub-regional soul range as a binarization process, the binarization process distinguishes the pixels of the subtitle block range 19 1255140 into pixels belonging to the subtitle and a pixel selection/character region not belonging to the subtitle a block, the text block is included in the subtitle range; and the number of times of the pixel of the text of the block (four) of the genus (four) is not changed by the number of pixels of the subtitle text; and according to the transformation The number of times determines the subtitle compensation output. 15. If the subtitle side compensation method is applied in the patent application, the interception record is greater than - (4), and the shutdown is performed. Error, otherwise it is judged to be correct.卞 补 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Like the affiliation of the affair, the job is abandoning the effect, and directly touches the current picture of the sub-scanning line range of the picture and the previous picture. 17· A method for image deinterlacing, comprising: · losing the father's fault~ like the error image is a sequence of consecutive multiple frames, and the frames are composed of odd and even fields; Performing a moving image area on the moving image side of the interlaced image; and detecting the sub-screen detection by distinguishing the interlaced image into a subtitle image and a non-subtitle image; a subtitle moving image compensation, outputting a subtitle deinterlacing output; performing a bribe for the non-subtitle part image, and outputting a tb-non-subtitle deinterlacing output; and, and the sub-screen moving image compensation output and the non-subtitle solution Interleaved output, outputting a deinterlaced image. The method for image deinterlacing according to Item 17, wherein the moving image detecting method compares the difference between the map fields and the moving image area. 20 1255140 19. Calculate the moving image area according to the method of image deinterlacing in item 17 of the application scope. The method of each of the screens includes: a plurality of elements according to the scanning lines in the moving image area The image vector 1 person - the scan line; and the image of the subtitle portion of the scan line containing the subtitles and the scan lines including the subtitles are obtained. 2〇·If the image deinterlacing method according to item 19 of the application scope is obtained, when the scanning of the plurality of pixels of the scanning lines of the eight-line system in the moving image area is determined, the scanning lines are determined as A scan line containing subtitles. ', °幻“-preset condition 21·, as in the application range 帛17 item image deinterlacing method, wherein the subtitle part silk red Wei weave material defines the subtitle element position according to the subtitle brightness reference value; - or corresponding pixels of the front and back plural fields, the 4 sides of the touch subtitles are operated - the values of the position of the subtitle pixels are obtained; and the subtitle pixel compensation result is checked to output the subtitle deinterlaced output. The method for deinterlacing the image according to the item 21, wherein the brightness of the pixel of the image of the subtitle portion is taken as a reference value of the subtitle brightness of the pixel ratio 23. The image deinterlacing method according to claim 21, wherein the subtitles are defined to include: 'According to the subtitle luminance reference value, a subtitle block range is obtained; when the subtitle block ranges within each scan line The brightness of a pixel is greater than the reference value of the subtitle brightness, and the pixel is determined to be a subtitle pixel; and 21 1255140 checks the brightness of the neighboring pixels of the plurality of pixels to be compensated for the subtitle block range, and determines The method for image deinterlacing according to claim 23, wherein obtaining the subtitle block range comprises extending a plurality of scan lines of the subtitle block range up and down, And all the plurality of scan lines satisfy a luminance greater than the subtitle luminance reference value, and the motion vector is the leftmost and rightmost plurality of pixels of the common motion vector of the scan line, respectively, The left and right borders of the subtitle block range are obtained, and the upper and lower boundaries of the subtitle block range are obtained. 25. The method for image deinterlacing according to item 23 of the application scope, wherein the neighboring pixels are in the form of Above and below, when the brightness of the pixel adjacent to the pixel is greater than the reference value of the subtitle brightness, or a certain number of pixels of the neighboring pixels are greater than a subtitle brightness limit, the pixel to be compensated is determined as The method for image deinterlacing according to claim 23, wherein the subtitle element compensation comprises: obtaining the pixels to be compensated before or before a pixel corresponding to the location; and an average operation of the pixel corresponding to the previous or previous map location to obtain the value of the pixel to be compensated. 27. The image deinterlacing method according to claim 21, wherein The result of checking the subtitles of the subtitle includes: ' The subtitle block range is subjected to a binarization process, and the binarization process divides all the pixels of the subtitle block range into subtitles belonging to the subtitle text and not subtitles. a character of the text; selecting a text block 'the text block is included in the subtitle block range; and counting the pixels belonging to the subtitle in each block of the text block and the pictures not belonging to the subtitle And determining, according to the number of times of the transformation, whether the subtitle pixel compensation output is correct. 22 1255140 28. The image deinterlacing method according to claim 27, wherein determining the subtitle compensation output comprises: if the text area When the number of blocks in the block that is too frequent is greater than a ratio, the caption complements the result and directly groups the caption part of the front- or front-end field, and vice versa. The result of the compensation as the pixel subtitle caption deinterleaving the wheel. Port 29. 'The method of image deinterlacing as described in the scope of the application, in which the interpolation method in the field, the interpolation between the fields, and the method of adapting the image to the six-fold thick 1 can be wrong. -. Reading, or moving image compensation, Lulu 23