TW589876B - Method for 3:2 pull-down film source detection - Google Patents

Abstract

A method for a 3:2 pull-down film source detection. First, a source is received. Then, field differences of two fields of the same type in the source and an average field difference according to the field difference corresponding to at least one prior field in the source are calculated. The source is established as a 3:2 pull-down film source by checking whether a 3:2 pull-down signature is in the source according to the field difference and the average field difference, and a bad editing point is detected according to an interlaced frame information of the source.

Description

589876 V. Description of the invention (1) The technical field to which the invention belongs, the present invention relates to a film source (Film Source) measurement method, and in particular to a method that can consider the difference between the two fields (Field) and interleaving ( Interlaced) Frame 2: 3: 2 pull,

Pul 1-Down) Detection method of movie source. Prior art In practice, moving picture sequences can be roughly classified into movie sources and video sources. Frame rate of film source (Frame is 24 frames per second, and frame rate of NTSC video source is 30 frames per second 60%. Therefore, in order to display movies in the TV system, the The frame rate must be converted from 24 to 30. This frame rate conversion is usually called a 2 (or 2: 3) pull-down procedure. Figure 1 shows an application to a frame with four frames (A, B, C, and D) The movie fragment 3: 1: 2 pull down the program example. In this example, the original movie frame is three or two interlaced fields. Frame 4 and frame c are divided by copying their fields. Three fields, while frame B and frame D are only divided into two fields. Therefore, the frame rate is converted from 24 frames per second of the original movie sequence to 60 fields per second of the interlaced field sequence, and It is then converted to an interlaced frame sequence of 30 frames per second. Since a 3: 2 pull-down movie sequence is suitable for playing on a NTSC interlaced TV, a pull-down movie sequence is on a progressive (scanning) TV or computer screen During play, interleaved frames combined by different movie frames, such as the frame “A + B” and the frame in Figure 1 ", B + C" will have some annoyances

Page 5 589876 Λ, description of the invention (2) ^ t irregularities, such as comb line artifacts (comb artifacts), saw = ,,, water, etc. In order to remove these wave processing, it is important to detect the 3: 2 pull down the movie ^ and execute an InVerse Telecine program to the 3: 2 pull: shirt sequence. It is important to restore the original movie frame of. Therefore, when used in the same device, it is very important to identify and detect the 3: 2 pull-down movie sequence or interleaved video sequence. — In addition, due to the duplication of the interlaced field 3.. .2, the pull-down movie sequence has an identification mark of (1). This identification mark is shown in Figure 2 and explained below. The fragment of this interlaced field sequence contains a 3: 2 pull-down movie to compare the interlaced% of the same type (eg, Top or Bottom) with a comparison result of "丨 〇 〇〇〇 丨 〇〇〇〇〇 丨, ..., where i represents the match (Match), 0 represents the match does not match. Therefore, in 2 pull down movie sequence or interleaved video sequence detection can be This is done by alternately aligning fields of the same type and looking for identification numbers in the sequence 10000100001 ... ". Figure 3 shows the method of detecting the source of a movie 3: 2. Refer to Figure 2 and Figure 3 at the same time, which will be described below. First, the field index (Field

Index) n, Match Counter, and Mode Counter are set to 0 (step S301). The pairing counter is used to ^ record how many times the 3: 2 pull-down identification mark π 1 〇00〇 ”has been detected, and the mode counter is used to indicate whether the 3 ·· 2 pull-down identification mark is correct. These two The counters are 3: 2 to pull down the key indicator of film source detection, and will be explained in more detail below. 3 Π Then 'as in step S 3 0 2, receive two fields of the same type and Enter

0702-8851TWF (N1); 91P59; yianhou.ptd Page 6 589876-V. Description of the invention (3) '~ Line comparison to see if they are the same because of copying. This is done by calculating the Field Difference FieldDiff ', 逯 (S303), and its formula is as follows: = 23 Σ Yf n -l · 2) | where M and N are the height and width of the field, respectively. If the field difference is less than a threshold Fi_th (YES in step S304), as in step S305, the two fields are considered to be in accordance with each other, the pairing counter is increased, and the mode counter is cleared to 0. 1 0 0 0 0 ”. On the other hand, if the field difference is greater than the threshold "Blaze 1_1: 11 (No in step 3304), as in step 3306, the two field-view comparisons match, and the mode counter is increased by 1." No later 'source sequence The mode can be judged based on the value of the pairing counter and the mode counter shown in Figure 2. When the pairing counter A is 1 and the mode counter is equal to 0 (YES in step $ 3 〇07), such as step magic &, Flag-The movie mode (pi imMode) is set to 1 to indicate that a 3 ·· 2 pull-down movie sequence has been detected. Otherwise, the value of the mode counter can be used to determine whether the sequence meets the identification mark. "丨 〇〇0 〇 ”. If the mode counter is less than or dedicated to 4 (No in step S309), it means that the movie mode has not changed (step S310). However, if the mode counter is greater than 4 (YES in step S309), this indicates this The sequence no longer meets the 3: 2 pull-down identification mark, f 1 0 0 0 0 ″ ”, as in step S 3 11 ′, the movie plate is set to 0 to indicate that the sequence is not a 3: 2 pull-down movie sequence 歹 1J. To prevent overflow (Overf low), when the mode counter exceeds a predetermined maximum value (YES in step S312), as in step S313, the mode counter is set

0702-8851TWF (Nl); 91P59; yianhou.ptd page 7 589876 5. Description of the invention (4) This is the maximum value. This processing flow is repeated as the sequence is input to dynamically Jon »Control 3: 2 to pull down the identification marks (step S 3 15 and return to step s 3 0 2) until the end of the sequence (YES in step S314). β However, the conventional method has two disadvantages. First, information is lost due to digital video compression and processing. The field difference between the replicated fields may exceed the threshold F i -th. Therefore, it is not accurate enough to judge whether the copied field is consistent using only the field difference and the threshold. Second, the detection of bad edit points (Bad Ed 111 ng P0 1 n t) is not efficient. For example, in Figure 4] a 3: 2 pull-down sequence contains good and bad edit points. Since the good edit points follow the 3: 2 pull-down order, the identification mark "丨 〇〇〇〇" is still maintained. Conversely, if a bad edit point violates the 3: 2 pull-down order, the recognition mark will continue to be 1 〇〇〇0 ". However, such a bad edit point will not be detected until 4 days later. Therefore, the output movie frames between the bad edit point and the point at which the bad edit was detected will be erroneously reconstructed using inverse video. That is, there are two frames from the merged field π Ηπ, and I " reconstruction ' causes serious reconstruction errors. In view of this, the main purpose of the invention t 1 β Α & 枘 UU h " is to provide

The difference between the source == and the interlaced frame information 3: 2 pull down the movie The editing point is to use the interlaced frame information to check "in order to achieve the above purpose of the invention" can be referred to by the present invention #

589876. V. Description of the invention (5) 3: 2 Pull down the detection method of the source of the film to achieve. According to the method for detecting the source of a movie in accordance with step 3 of the present invention, first, a source is received, and the correlation coefficient is initially set to zero. The coefficients include a field index η, a pairing counter, a mode counter, an accumulated field difference accDiff, and an average field difference aveDiff. Next, calculate the field difference FieldDiff of two fields of the same type, and compare the field difference FieldDiff with a first threshold Fi — ^^ and an adaptive threshold Adapt — th. If the field difference is less than Fi-th

Adap_th, the pairing counter is incremented, and the pattern count is 7 /, and the talk is 0. Conversely, if the field difference is not less than ^^: 1, the branch count state is increased by 1. FieldDiff is added to accDiff, and accDiff is divided by the mode counter. Among them, Adapt_th is multiplied by a coefficient κ. When coming,

Finally, when the pairing counter is greater than 丨 and the mode counter is equal to 0 == the source of the movie is pulled down for 3: 2 and when the source of the pairing counter is greater than 4, it is detected that the source is not 3: 2. The mode counter is equal to 3 and the mode counter is equal to: depending on the order and intersection of the movie source in accordance with 3: 2. “Order ¥”-Bad edit point is measured in the source. Change the old ^ error% is the bottom field, determine whether this view is a

Wrong field = end field, determine whether this view is-mind view :: In addition, when the corresponding mode counter is equal to 3, the corresponding mode counter is equal to two of 3 and 4 ^ 4 under = when 'if the point is in the source Detect. Ma Xunxu is bad. "

Implementation Method The present invention roughly uses a field difference FieldDif and a conventional messenger to determine whether a duplicate field is found. However, the limit of = = determines whether the field difference FieldDiff is matched with the contrast ratio. The above assumption is based on the fact that no matter how big the error in the movie sequence is, the two replicated fields: m phase ί: d 2 two fields that do not match. However, because the field difference is strongly localized when the film segment is two precise, therefore, in the present invention, the threshold value of a forest will be dynamically updated by the comparison result of the previous field difference. d-ίΓ provides bad editing points by referring to the interlaced frame information ^ " The above is based on the actual situation of the movie sequence being pulled down at 3: 2. Another unique identification sign, —ρρ ,,, and P " indicates sequential and " I " indicates interlaced. In other words, in the = group 3: 2 pull-down interlaced frame, the first frame is sequential and the second It intersects with the third frame, and the last two frames are sequential. One / Social is clearly known in the example of the magic map. Because & 'if = ^ · the father's wrong frame does not follow this identification mark, Then there must be bad editing points for this source, and in order to avoid reconstruction errors, the inverse video program cannot be executed to restore the movie frame. Figure 5 shows the method for detecting the source of a movie according to 3 ·· 2 of the embodiment of the present invention. First, as in step S502, a source is received, and in step S502, the correlation coefficient used in the detection is initially set to zero.

589876 • V. Description of Invention (7). These coefficients include field index n, pairing counter, mode counter, cumulative field difference accDiff, and average field difference aveDiff. Similarly, the pairing counter is used to record how many times the 3: 2 identification mark has been pulled down, and “10000” has been detected in the source, and the mode counter is used to indicate whether the 3: 2 identification mark has been pulled down. Go for the right. Then 'as in step S503, two fields having the same type (F (η) and F (n + 2)) are received, and as in step S504, F (n) and F (n + 2) are calculated FieldDif f. Similarly, the formula for the field difference is as follows: η)-yf η + 2) |

7 = 0 G where M and N are the height and width of the field, respectively. After that, in step 3505, the field difference F i e 1 dD i f ′ is compared with two threshold values. The first threshold value is Fi-th 'which roughly defines the maximum allowable field difference between two coincident fields. The second threshold value is a variable threshold value Adapt-th, which is temporarily updated by determining the content in the reference field sequence to determine the maximum allowable field difference of the two coincident fields. In this embodiment, Adapt —th is a part of the average field difference aveDiff of the previous field comparison (Fraction) and is: where 'K is a real number between 0 and 1, and aveDiff is: aveDiff = aaoDlfflModeCoun- t ^: α The assumption here is that if there is highly active content in the field sequence, the average field difference will be higher, and similarly, the accumulated field difference acc]) iff will also be higher. Therefore, the allowable field difference between the replicated fields will be relatively high. However, ‘if there is low activity in the field sequence, the average field difference will be

589876 V. Invention description (8) is lower. Similarly, the accumulated field difference a c c D i f f will be lower. Therefore, the allowable field difference between the replicated fields will be relatively low. It is worth noting that the threshold value Adapt_th should be smaller than Fi__th, and Fi_th and K can be determined according to different applications or movie types, and are usually known by experience. If the current field difference FieldDiff is less than Fi_th and Adapt_th (yes at step S5 0 5), field F (n) is considered to be consistent with field F (n + 2) and field F (n + 2) is the field F (η) replication field. After that, in step S506, the pairing counter is incremented by one, and the mode counter, accDiff, and aveDiff are set to zero. Conversely, if the field difference is not less than Fi_th or Adapt_th (No in step S505), that is, field F (n) does not match field F (n + 2), then in step S50 7, the mode counter is increased by 1. FieldDiff Added to accDiff, aveDiff is equal to accDiff divided by the mode counter, and Adapt a th is equal to aveDiff multiplied by the coefficient K. After that, the mode of the source sequence can be judged based on the values of the pairing counter and mode counter as shown in Figure 2. When the pairing counter is greater than 1 and the mode counter is equal to 0 (YES in step S508), as in step S509, the flag-movie mode (Fi ImMode) is set to 1 to indicate that a 3 ·· 2 pull-down power sequence has been detected. Here. Conversely, the value of the mode counter is used to determine whether the sequence is in compliance with the standard identification mark 1 100 000 0π. If the pairing counter is greater than 4 (YES in step S 5 10), the sequence no longer obeys 3: 2 to pull down the identification mark, 1 0 0 0 0 ”, as in step S511, the movie mode is set to 0 to indicate that the sequence is not疋 3: 2 pull down the movie sequence. However, if the pairing counter is small or at 4 (No in step S510), the interlaced frame information is used to detect bad edit points. This interlaced frame information can indicate the frame It is a sequential or staggered frame. The detailed operation of bad edit point detection will be described below.

589876 1 V. Description of the invention (9) Fig. 6 shows an example of checking for bad edit points in Fig. 6 according to the embodiment of the present invention. Is a staggered field of 3. At the first checkpoint, the second error: should be a counter type, so it corresponds to the lower end field according to the 3: 2 pull-down sequence. "Staggered view" " B + C "should both be staggered view and the next is staggered view. This checkpoint is a bad checkpoint in the right eight. The intersection ... is an upper end. The grid "D" should be a sequential view. If = long, 2 pulls corresponding to it, this checkpoint is a bad edit point. T D does not follow sequential view, and the corresponding mode counter is 3 ^ w. Therefore, 'bad edit' The point will only be counted in the mode;! == This happens. The system is ^ early detection> then ', but it is not limited to if / bad editing points are detected S513, which means that the movie mode has not changed. Second step S512疋) 'as in step point (No in step s512), as in step S5U "," Bad editing detected. 0. The leather shirt sequence, and the pairing counter is set to note that the corresponding mode is checked by inspection. The counter is 3 ~ i, and the editing point detection can also be completed by (pr0gressiverJ). The sequence is correct under "sequentiality", otherwise, the ",", and "α" results are sequential. Then this 3 ·· 2 is pulled in The movie mode judges that the second point of the disk is a bad edit point. After the bad edit point of the grid F (n) is detected, it is page 13 07 02-8851TWF (Nl); 91P59; yianhou.ptd 589876, V. Description of the invention (10) Prevents overflow, when the mode counter exceeds a predetermined maximum value (YES in step S514), as in step S515, the mode counter It is set to the maximum value. In addition, in order to maintain the local threshold of the content of the movie content when the threshold value Adapt_th is reflected, accDiff and aveDiff are set to 0, and when the next frame (F (n + 1)) is checked It will be recalculated (S51 7 and back to S503). This process flow dynamically monitors the sequence as the input of the sequence is repeated. 3 · 2 pull down the source and check for bad edits until the sequence ends (the work of step% 16 · U, ^ = The method of 3: 2 pull down movie source provided by the present invention can consider the difference between the two venues, and whether it is: 3 · 2 pull down the source of the electric shirt. In addition, the present invention borrows Checking the Gongxun is used to detect the wrong method again and again in the place where it may occur. Surgeons are more accurate detection of bad edit points. Although the present invention has been disclosed in the preferred embodiment as above, the present invention is limited, anyone familiar with this, , Not for God and the scope, as this fight, , Without departing from the scope of the application finish when the view spot after attachment of the present invention; therefore the protection as defined by the present invention === J whichever girth.

589876 ^ Simplified illustrations of the drawings In order to make the above-mentioned objects, features, and advantages of the present invention more comprehensible, the following specific examples are given in conjunction with the accompanying drawings to explain in detail as follows: Figure 1 shows an application to Example 3 of movie frames (A, β, c, and D) 2 pull down the program example. Figure 2 shows a 3: 2 pull down the movie source and its identification mark. Figure 3 is a flow chart showing the method of detecting the source of a movie 3: 2. Figure 4 shows a 3: 2 pull down sequence with one good and one bad edit point, respectively. FIG. 5 is a flow chart showing an operation flow of a method for detecting a movie source according to 3: 2 of the embodiment of the present invention. Fig. 6 shows an example of bad edit point detection according to an embodiment of the present invention. Explanation of symbols

A, B, C, D, E, A + B, B + C η ~ field index;

FieldDiff ~ field difference; accDi f f ~ cumulative field difference; aveDiff ~ average field difference;

Fi_th ~ threshold;

Adap_th ~ threshold of variation; K ~ coefficient; S301, S3〇2 ..... S315 ~ operation steps S501, S5 02 ..... S517 ~ operation ^ step

0702-8851TWF (N1); 91P59; yianhou.ptd page 15

Claims (1)

589876-VI. Scope of patent application1. A 3: 2 detection method of pulling a movie to Mae / Origin, including the following steps: receiving a source; ;; the source having two types of the same type based on the corresponding at least field in the source Difference; and one of the field differences in the sequence; the field difference of the previous field is calculated based on the existence of the field difference and the average field mark in the source. Does Huanyi have a 3: 2 pull-down ^ As described in item 1 of the scope of patent application, the detection method of the 3: 2 pull-down movie source further includes detecting the edit point based on the source ^ μ, a # t. The original-the father misidentified the frame poor information to detect-the 3: 2 mentioned in the bad range item 2 pulls down the detection of the film source π ::: The parent misidentified frame information indication-the corresponding frame is-sequential or interlaced Sight 4 · A 3: 2 detection method of pulling down the source of a movie: method, including the following steps to receive a source; the initial setting of the complex coefficients is zero, and these coefficients include a field index η, a pairing counter, a mode counter, A cumulative field difference accDif f, and an average field difference aveDif f; the difference between one of the two fields p (n) and F (n + 2) of the same type in the source FieldDif f; the field difference FieldDif f Compare with a first threshold Fi_th and an adaptive threshold Adapt__th; 0702-8851TWF (N1); 91P59: yianhou.ptd page 16 589876, patent application scope If the field difference FieldDiff is smaller than the Fi — th and the Adapt th, increase the pairing counter by 1, and the pattern counter, The ac "iff and the aveDi ff are set to 0; 〆, if the field difference is not less than the Fi_th or the Adapt —th, increase the mode counter by 1, add the FieidDiff to the accDiff, and the aveDiff is equal to the core accDiff divided by The mode counter, and the Adapt_th is equal to the multiplication by a coefficient κ; and when the pairing counter is greater than 丨 and the mode counter is equal to 0, the source is measured as a 3: 2 pull down movie source. 5 Turu " Patent The 4th item in the range of 3: 2 pull down the source of the movie i 9 'robin includes when the pairing counter is greater than 4, the source is detected as a non 3: 2 pull down movie source. i: The 3: 2 pull down of the source of the movie described in the range of item 4 ^ by judging whether the image frame including the corresponding mode counter equal to 3 ^ 2 is a sequential frame or-interlaced格格 来 ',,] 3.2 pull down the bad sequence in the movie sequence Test 2: As described in Fan 3: 2, when the source of the movie is pulled down, the bad editing point in the source box is not an interlaced view. As shown in the example: = Fan 2 is used to pull down the source of the movie. The view frame is not the detection of the adjacent view after any one of the view frames of interlaced view G. Grid 4, the bad edit point in the source was pulled down by the movie 3: 2 as described in item 6 of the scope of patent application Debt of source 0702-8851TWF (N1); 91P59; yianhou.ptd page 17 589876-6. Patent application scope, testing method, when the field is an upper field and the view is not a sequential view, the bad edit point is at Detected in this source. N 1 0. As described in item 4 of the scope of patent application 3: 2, the source of the movie is pulled down, and the detection method further includes detecting by determining whether the two fields corresponding to the mode counters 3 and 4 are sequential. Bad edit point in this source. 11. The method for detecting the source of a movie according to 3: 2 described in item 10 of the scope of the patent application, wherein the field corresponding to the mode counter of 3 is the lower end field. . f 0702-8851TWF (N1); 91P59; yianhou.ptd page 18