JP6588745B2 - Scene change detection device and program thereof - Google Patents

Description

  The present invention relates to a scene change detection device, a scene change detection method, and a program for detecting a scene change of a moving image.

  Scene change detection is a technique for detecting a turning point of a scene, and is used to control encoding parameters in a moving image encoding apparatus. Further, in a consumer video recorder, it is used for detecting a turning point of a commercial message and a program program, encoding parameter control at the time of re-encoding, and the like.

  As a method for detecting a scene change, there is a method using a change in pixel value between frames. For example, in the case of using a pixel value change between frames, a pixel change is compared between two adjacent frames, and a scene change is made when the difference is large (see, for example, Non-Patent Document 1).

  By decoding once the encoded video signal, the technique of Non-Patent Document 1 can be used, but there is a problem that the calculation amount is large because the technique is decoded once. Further, the technique of Non-Patent Document 1 has a problem that it is easy to be erroneously detected in a video with a large movement. Therefore, there are coding parameters that use DCT (discrete cosine transform) coefficient correlation and motion vector directions among the coding parameters for scene change detection of the coded video signal. For example, for example, a technique (for example, refer to Non-Patent Document 2) using a correlation of DCT coefficients of an intra prediction frame (Intra: hereinafter referred to as “I frame”, which is a frame encoded using only intra prediction). , “P frame” (Predicted: a frame that can be predictively encoded from one direction) or “B frame” (Bi-Directional Predicted: a frame that can be predictively encoded from the past and the future) There is a technique using a vector direction (see, for example, Non-Patent Document 3).

  Furthermore, in the B frame, a motion vector reference frame is counted for each macroblock in the frame, and a scene change is detected based on the change in the statistics (see, for example, Patent Document 1), or the type of macroblock Another technique is to detect the scene change by counting the number of macroblocks and using the bias in the motion compensation prediction direction (see, for example, Patent Document 2).

JP-A-6-54315 JP 2011-66682 A

Nagasaka Goro and Tanaka Joe, “Automatic Indexing and Object Searching in Color Video Images”, Information Processing Society of Japan, Journal of Information Processing Society of Japan, Vol.33, No.4, pp.543-550, 1992 Issued on April 15 Boon-Lock Yeo, et al., “Rapid scene analysis on compressed video”, IEEE, Circuits and Systems for Video Technology, IEEE Transactions on Volume 5; Issue: 6, pp.533-544, December 1995. Toshimitsu Kaneko, Osamu Hori, “High-speed cut detection from MPEG video using motion vector coding”, IEICE, IEICE technical report, PRMU, Pattern recognition / media understanding, Vol. 96, No.385, pp.55-62, Published November 22, 1996

  As described above, the technique of Non-Patent Document 1 has a problem that a calculation amount is large because decoding is performed once, and a problem that false detection is likely to occur in a video with a large motion.

  In addition, in the technique using the correlation of DCT coefficients and the technique using the motion vector direction among the encoding parameters such as Non-Patent Documents 2 and 3 and Patent Documents 1 and 2, in order to obtain DCT coefficients and motion vectors, There is a problem that the amount of calculation must be increased due to partial decoding, and a detectable GOP (Group Of Pictures) structure has a GOP length M = 3 (I / P, B, B, P / I) and the like, and there is a problem that it cannot be applied when the GOP structure has a hierarchical structure.

  Therefore, there is a demand for a technique for specifying a scene change frame with a small amount of calculation from encoding parameters in an encoded stream regardless of whether the GOP structure has a hierarchical structure.

  Accordingly, the present invention has been made in view of the above-described problems, and an object of the present invention is to provide a scene change detection apparatus and scene change detection that efficiently detect a scene change for a moving image having an arbitrary GOP structure. It is to provide a method and a program thereof.

  The present invention generally includes a change in a statistic related to the number of coding unit blocks between consecutive similar frames (for example, between I frames, between P frames, between B frames, etc.) in a video having an arbitrary GOP structure. Based on this, a scene change is detected.

Scene change detection device of the present invention, there is provided a scene change detection device for detecting a scene change of a moving image, statistic compiled for each frame the number of put that coded unit block in allogeneic consecutive frames in the video A scene for identifying a scene change frame by means of the same frame inter-frame determination processing means for determining a scene change section by comparing the change in a threshold according to the frame type, and an overlapping section of the scene change section determined by all frame types Change detection means, and the same-type inter-frame determination processing means uses, as the statistic, the number of coding unit blocks having the smallest size among all block sizes constituting a frame, and a threshold value according to the frame type. Comparing and determining the presence or absence of the scene change section, the moving image is When the OP structure has a hierarchical structure including one or more lower frames of the P frame, the B frame, and the non-reference B frame with respect to the highest I frame, the I located at the highest level of the hierarchical structure Means for determining a scene change section based on a change in the statistic in order from a frame, and determining whether or not the scene change section exists in a lower frame only when it is determined that the scene change section exists in an upper frame; all the moving image as the GOP structure when consisting of I frames, and wherein Rukoto that Yusuke means for determining scene change interval based on the change of the statistical amount for each I consecutive frames, the.

Furthermore, the program of the present invention is a program for causing a computer to function as the scene change detection device of the present invention .

  According to the present invention, it is possible to detect a scene change of an arbitrary GOP structure only by processing with a light load by comparing the number of encoding unit blocks without requiring a high load calculation such as DCT.

It is a block diagram which shows schematic structure of the scene change detection apparatus of one Embodiment by this invention. It is a flowchart of the main routine which shows the operation example of the scene change detection apparatus of one Embodiment by this invention. It is a flowchart of the subroutine which shows the operation example of the scene change detection apparatus of one Embodiment by this invention. It is a flowchart of the subroutine which shows the operation example of the scene change detection apparatus of one Embodiment by this invention. It is a figure which shows the operation | movement of the same kind frame determination process of one Example in the scene change detection apparatus of one Embodiment by this invention. It is a figure which shows the scene change detection result of one Example in the scene change detection apparatus of one Embodiment by this invention.

  Hereinafter, a scene change detection apparatus 1 according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a schematic configuration of a scene change detection apparatus 1 according to an embodiment of the present invention.

(Device configuration)
The scene change detection apparatus 1 performs scene change based on a change in a statistic regarding the number of coding unit blocks (hereinafter also referred to as “CB: Coding Block”) between consecutive similar frames in a moving image having an arbitrary GOP structure. The number of CBs that can be set as one of the encoding parameters between consecutive similar frames is greatly different depending on the design.

  Then, the scene change detection apparatus 1 compares the number of CBs between consecutive similar frames, and determines a section where the CB number greatly changes as a “scene change section”. In this example, as frame types, an I frame, a P frame, a B frame, and a non-reference B frame (hereinafter referred to as “NRB frame”: a B frame that is not referred to by other frames). Can be handled. That is, the scene change detection apparatus 1 determines whether or not there is a scene change section between the I frame, the P frame, the B frame, and the non-reference B frame, and determines that it is a scene change section in all frame types. The detected frame is detected as a scene change frame. Note that the CB number greatly changes means that the difference in the CB number between two frames to be compared in a certain frame type (X frame) exceeds the threshold Th (X) set for each frame type.

  More specifically, as shown in FIG. 1, the scene change detection device 1 according to the present embodiment includes an inter-frame determination processing unit 11 and a scene change detection unit 12. The homogenous interframe determination processing unit 11 includes a homogenous frame extraction unit 111, a block number calculation unit 112, and a scene change section detection unit 113.

  The inter-frame determination processing unit 11 inputs a video signal of a moving image having an arbitrary GOP structure, stores a predetermined number of frames (for example, 32 frames) for determining a scene change as a “detection target section”, and In this detection target section, the presence / absence of a scene change section is sequentially determined for each frame type. This is a functional unit that outputs to the change detection unit 12. Here, the detection target section for determining the scene change can be any number of frames as long as the number of CBs between the same type frames other than the I frame that are continuous across the I frames can be compared. .

  The homogeneous frame extraction unit 111 is a functional unit that sequentially extracts consecutive X frames existing in a detection target section in order to determine whether or not there is a scene change section in a certain frame type (X frame).

  The block number calculation unit 112 is a functional unit that counts the number of CBs of consecutive X frames that are sequentially extracted by the homogeneous frame extraction unit 111.

  The scene change section detection unit 113 compares the number of CBs of consecutive X frames counted by the block number calculation unit 112, and when the difference in the number of CBs exceeds a threshold Th (X) set for each frame type. This is a functional unit that determines the scene change section.

  By using the functions of the same frame extraction unit 111, the block number calculation unit 112, and the scene change section detection unit 113, the presence / absence of a scene change section is sequentially determined for each frame type in the detection target section. The frame number indicating the scene change section can be determined.

  When the GOP structure of the input video signal is known in advance, the homogenous interframe determination processing unit 11 may determine the scene change section only for the frame type related to the GOP structure. When the GOP structure is not known in advance, the frame type is first identified, and the presence / absence of a scene change section is determined for each frame type. For example, the frame type may be identified by referring to the leading GOP header of each GOP or the leading frame header of each frame.

  For this reason, when the scene change determination of a certain detection target section is completed, the same-type inter-frame determination processing unit 11 proceeds to the scene change determination of the next detection target section. In order to make it possible to compare the number of CBs between the same type of frames that are continuous across these detection target sections, for example, at least one GOP may be included in the continuous detection target sections. .

  The scene change detection section 12 is a scene change section when the same frame inter-frame determination processing section 11 determines that there is a scene change section for each frame type within a predetermined number of frames (for example, 32 frames). This is a functional unit that obtains overlapping sections, detects a frame determined to be a scene change section in all frame types as a scene change frame, and outputs the frame number as a detection result. However, the scene change detection unit 12 is configured to detect whether or not a scene change frame has been detected for each detection target section in which a scene change is determined, in addition to configuring the scene change frame as a detection result. Can also be configured to output.

(Device operation)
Hereinafter, an operation example of the scene change detection apparatus 1 according to the present embodiment will be described with reference to FIGS. 2 to 3. FIG. 2 is a flowchart of a main routine showing an operation example of the scene change detection apparatus 1 of the present embodiment, and FIGS. 3 and 4 are flowcharts of the subroutine.

Here, the frame type X frame included in the scene change determination section is described as fX (n) = {x ₀ , x ₁ , x ₂ ,..., X _n }. That is, the I frame is fI (n) = {i ₀ , i ₁ , i ₂ ,..., I _n }, and the P frame is fP (n) = {p ₀ , p ₁ , p _{2 ,.} }, B frame is fB (n) = {b ₀ , b ₁ , b ₂ ,..., B _n }, NRB frame is fNRB (n) = {nrb ₀ , nrb ₁ , nrb ₂ ,..., Nrb _n }. Moreover, the CB number in a certain frame _{x m} and CB. Note that n is an integer greater than or equal to 0 and indicates the frame order within the detection target section for each frame type, and the frame number of the moving image can be associated. m represents a value to be determined among 0 to n.

Further, when the scene change determination section is [x _m−1 , x _m ], the difference in the number of CBs is expressed as diffCB [x _m−1 , x _m ] and can be calculated by the equation (1).

diffCB [x _m−1 , x _m ] = | CB (x _m ) −CB (x _m−1 ) | (1)

The scene change determination flag indicating that it is determined that there is a scene change section and the frame number is represented by CCf [x _m−1 , x _m ].

  The threshold value Th (X) used for determining the scene change section can be a fixed value corresponding to the resolution and the picture of the moving image, or a predetermined statistical value, for example, a value calculated by equation (2). can do.

Th (X) = (average value of diffCB [x _m−1 , x _m ])
+ 2 × (standard deviation of diffCB [x _m−1 , x _m ]) (2)

  3 and 4 show the subflows in the scene change detection with respect to the main flow in the scene change detection shown in FIG. 2, but the frame type W in FIG. Indicates the frame type corresponding to the hierarchy. For example, when X indicates a P frame, W indicates an I frame, when X indicates a B frame, W indicates a P frame, and when X indicates an NRB frame, W indicates a B frame.

  First, referring to FIG. 2, the scene change detection apparatus 1 uses a similar frame extraction unit 111 of the same frame inter-frame determination processing unit 11 to start a continuous I when a scene change determination starts in a certain detection target section. Frames are extracted, and scene change determination between successive I frames, that is, interframe determination (I) is performed (steps S1 and S2). Inter-frame determination (I) is performed by the subroutine shown in FIG.

Referring to FIG. 3, the inter-frame determination processing unit 11 uses the block number calculation unit 112 to calculate the number of CBs of two consecutive I frames [i _m−1 , i _m ] in the detection target section. Counting is performed, and the scene change section detecting unit 113 calculates the difference absolute value diffCB [i _m−1 , i _m ] (equation (1)) of the CB number of consecutive I frames counted by the block number calculating unit 112. (Step S11). Subsequently, the inter-frame determination processing unit 11 uses the scene change section detection unit 113 to compare diffCB [i _m−1 , i _m ] with Th (I) to determine the presence / absence of a scene change (step S12). ). If determined that there is no scene change (step S12: No), the scene change decision flag _{CCf [i m-1, i} m] = a false (False) (step S13), and if it is determined that a scene-change has occurred (step S12: Yes), the scene change determination flag CCf [i _m−1 , i _m ] = true (True) (step S14).

That is, referring to FIG. 2, first, diffCB [i ₀ , i ₁ ] is obtained in the scene change determination section [i ₀ , i ₁ ], and when it is determined that there is no scene change (step S2: False). Then, the inter-frame determination (I) is performed in the next scene change determination section [i ₁ , i ₂ ] (S1, S2 through steps S3, S4). That is, whether it is determined that a scene change (step S2: True), or for all of I-frames existing in the detection target interval until the scene change determination is completed (in other words, until i m _{+ 1>} i _n) Then, the inter-frame determination (I) is performed (step S3: S1 through No and S4). A scene change determination section _{[i m-1, i m} ] If it is determined that the scene change (step S2: True), the scene change decision flag _{CCf [i m-1, i} m] = True (True (That is, step S14 shown in FIG. 3), the process shifts to scene change determination between P frames, that is, determination between frames (P) (steps S5 and S6). In the inter-frame determination (I), when it is determined that there is no scene change until the end within the detection target section (step S3: Yes), it is determined that there is no scene change within the determination section, and the process is terminated ( It moves to the next detection target section of the moving image).

  Inter-frame determination (P) is performed by the subroutine shown in FIG. Here, x shown in FIG. 4 can be replaced with p, and w can be replaced with i.

Referring to FIG. 4, the determination start frame at this time is the frame P _k having the smallest frame number among the P frames included in the scene change determination section [i _m−1 , i _m ] of the I frame. (Step S21). In the inter-frame determination (P), the same-type inter-frame determination processing unit 11 uses the block number calculation unit 112 to calculate the number of CBs of two consecutive P frames [p _k−1 , p _k ] in the detection target section. Each difference is counted, and the difference absolute value diffCB [p _k−1 , p _k ] (formula (1)) of the number of CBs of consecutive P frames counted by the block number calculation unit 112 is counted by the scene change section detection unit 113. Calculate (step S22). Subsequently, the inter-frame determination processing unit 11 compares the diffCB [p _k−1 , p _k ] with Th (P) by the scene change section detection unit 113 to determine the presence / absence of a scene change (step S23). ). If it is determined that there is no scene change (step S23: No), the scene change determination flag CCf [p _k-1 , p _k ] = false (false) (step S24), and the next scene change determination section [p _k , p _{k + 1} ], the inter-frame determination (P) is performed (steps S25 and S26, and then S22 and S23).

That is, it is determined that a scene change has occurred (step S23: Yes) or until the scene change determination for all P frames existing in the detection target section is completed (in other words, p _k-1 > im _). Until the inter-frame determination (P) is performed (step S23: No, S24, S23: No and S26 through S22). When it is determined that a scene change has occurred in a certain scene change determination section [p _k−1 , p _k ] (step S23: Yes), the scene change determination flag CCf [p _k−1 , p _k ] = true (True) ) (Step S27, ie, step S6: True in FIG. 2), the process shifts to a scene change determination between B frames, ie, an inter-frame determination (B) (steps S7 and S8 in FIG. 2). In the inter-frame determination (P), when it is determined that no scene change has occurred until the end of the detection target section (step S25: Yes), it is determined that the scene change detection in the inter-frame determination (I) is a false detection. (Ie, step S6: False in FIG. 2).

Inter-frame determination (B), which is a scene change determination between consecutive B frames, is processed in the same manner as inter-frame determination (P) by replacing x with b and w with p shown in FIG. (Steps S7 and S8). Therefore, the inter-frame determination (B) determines that a scene change has occurred in a certain scene change determination section [b _k−1 , b _k ], and the scene change determination flag CCf [b _k−1 , b _k ]. = True (S8: True), the process shifts to a scene change determination between NRB frames, that is, an inter-frame determination (NRB). In the inter-frame determination (B), when it is determined that no scene change has occurred in the detection target section, the scene change determination flag CCf [b _k−1 , b _k ] = false (False) and the inter-frame determination ( If the scene change detection in I) is an erroneous detection, the subflow is terminated (S8: False).

Also, the interframe determination (NRB), which is a scene change determination between consecutive NRB frames, is processed in the same manner as the interframe determination (P) by replacing x shown in FIG. 4 with nrb and replacing w with b. (Steps S9 and S10). Accordingly, the inter-frame determination (NRB) determines that a scene change has occurred in a certain scene change determination section [nrb _k−1 , nrb _k ], the scene change determination flag CCf [nrb _k−1 , nrb _k ] = True (S10: True), the process proceeds to step S11. In the inter-frame determination (NRB), when it is determined that there has not been a scene change until the end in the detection target section, the scene change determination flag CCf [nrb _k−1 , nrb _k ] = false (False) and the inter-frame determination ( The subflow is terminated assuming that the scene change detection in I) is an erroneous detection (S10: False).

  Here, in the scene change detection device 1, when it is determined by the inter-frame determination processing unit 11 that there is a scene change in the inter-frame determination (NRB), the scene change detection unit 12 determines the previous determination, that is, the frame Among the sections determined to have a scene change in the inter-frame determination (I), the inter-frame determination (P), the inter-frame determination (B), and the inter-frame determination (NRB), all the frame types are obtained. In step S11, the frame determined as the scene change section is detected as a scene change frame. This overlapping section is two adjacent frames in the moving image, and it can be seen that the characteristics of the moving image change greatly during this period.

  From the above, the scene change frame can be specified as the subsequent frame of the two adjacent frames in the overlapping section. In this way, for example, as shown in FIG. 5, for a moving image having a GOP structure with a GOP length M = 8 and a random access format, an interframe determination (I), an interframe determination (P), an interframe determination ( B) and inter-frame determination (NRB) can be used to determine the scene change section, and the scene change frame can be specified in the overlapping section. Therefore, no heavy load calculation such as DCT is required, and the number of CBs A scene change of an arbitrary GOP structure can be detected only by processing with a light load by comparison. In particular, in order to determine the scene change section in order from the top of the hierarchical structure, for example, when it is determined that there is no scene change section in the interframe determination (I), the interframe determination (P) between the frames, The determination (B) and the determination between frames (NRB) can be omitted, the processing load is reduced, and the processing speed is improved.

(Example)
FIG. 6 shows a scene change detection result of an example in the scene change detection apparatus 1 of the present embodiment. FIG. 6 shows a specific example of the number of CBs by CB size in the case of encoding with HEVC / H.265. The horizontal axis represents the frame number, and the vertical axis represents the count number for each CB size (64 × 64 pixels, 32 × 32 pixels, 16 × 16 pixels, and 8 × 8 pixels).

The conditions used for encoding are as follows.
Quantization parameter (QP): 22 (fixed),
GOP structure: Random access method (GOP length M = 8)
・ Scene change frame: 140th frame

The frames included in fI (n), fP (n), fB (n), and fNRB (n) are as follows.
FI (n) = {0,32, ..., 128,160, ...}
FP (n) = {8,16, ..., 120,136,144,152, ...}
・ FB (n) = {2,4,6,10,12, ..., 134,138,140,142,146,148, ...}
FNRB (n) = {1,3,5,7, ..., 137,139,141,143,145, ...}

  First, the scene change determination (I) is performed from the section [0, 32] that is the first detection target section. If no scene chain is detected in the section [0, 32], the scene change determination (in the order of the section [32, 64], the section [64, 96], the section [96, 128], and the section [128, 160]) I). As the scene change determination (I) proceeds, diffCB [128,160] is large and exceeds a certain threshold Th (I) in the section [128,160]. At this time, CCf [128, 160] = True, and inter-frame determination (P) is performed.

The first _{p k} to perform inter-frame determining (P) is the minimum P frame included in the interval [128,160], i.e. from becoming a 136 frame, the scene change determination (P) is the interval [120,136] If no scene change is detected in the section [120, 136], the process proceeds to the next section [136, 144]. In the interval [136, 144], diffCB [136, 144] is large and exceeds a certain threshold Th (P). At this time, CCf [136,144] = True, and the process proceeds to inter-frame determination (B).

The first b _k for performing the interframe determination (B) is the minimum B frame included in the section [136, 144], that is, 138 frames. Therefore, the scene change determination (B) is performed from the section [134, 138]. If no scene change is detected in the section [134, 138], the process proceeds to the next section [138, 140]. In the section [138, 140], diffCB [138, 140] is large and exceeds a certain threshold Th (B). At this time, CCf [138, 140] = True, and the process proceeds to inter-frame determination (NRB).

The first nrb _k for performing the interframe determination (NRB) is the minimum NRB frame included in the section [138, 140], that is, 139 frames. Therefore, the scene change determination (NRB) is performed from the section [137, 139]. If no scene change is detected in the section [137, 139], the process proceeds to the next section [139, 141]. In the section [139, 141], diffCB [139, 141] is large and exceeds a certain threshold Th (NRB). At this time, CCf [139, 141] = True.

  With the above processing, the frame is determined in the interval [128, 160] in the frame determination (I), in the interval [136, 144] in the determination between frames (P), and in the interval [138, 140] in the determination between frames (B). In the intermediate determination (NRB), the scene change is detected in the section [139, 141]. The frames included in the above sections are two adjacent frames of 139 and 140 frames, and the video characteristics have changed greatly during this period, so that 140 frames can be specified as scene change frames.

  In the present embodiment, an example of scene change detection in the random access method with the GOP length M = 8 is shown, but inter-frame determination (I) is also possible in scene change detection in the all-intra (All Intra) method consisting of all I frames. This can be done by implementing In addition, with respect to a low delay system composed of an I frame and a P frame, a scene change can be detected by performing an inter-frame determination (I) and an inter-frame determination (P).

  The present invention has been described with reference to specific embodiments and examples. However, the present invention is not limited to the above-described examples, and various modifications can be made without departing from the technical concept thereof.

  For example, in the above-described embodiments and examples, an example in which scene change detection is performed based on the number of CBs for each frame has been described. However, when it is desired to further reduce the amount of calculation, scene change is performed based on the minimum size CB. Detection may be performed. For example, when a certain CB is set to the minimum size CB, in HEVC / H.265, the minimum size CB is 8 × 8 pixels. Therefore, the scene change may be detected only by the number of CBs of 8 × 8 pixels. .

  In the above-described embodiments and examples, an example in which scene change detection is performed based on the number of CBs for each frame has been described. However, a frame is divided into two or more regions, and some or all of them are divided. You may comprise so that the scene change detection mentioned above may be performed in each of several area | regions. In the case of dividing into a plurality of regions, the divided shape may be the same as a divided region by slice division or tile division in HEVC / H.265. If the result of scene change detection differs depending on the divided area, all the areas where scene change detection has been performed or frames that have been detected as scene changes in more areas by majority decision can be identified as scene change frames. Good. For example, the calculation amount can be further reduced by setting a partial region of the frame as a determination target.

  The scene change detection apparatus 1 of the present embodiment can function as a computer, and a program for causing the computer to realize each component according to the present invention is a memory (inside or outside of the computer). (Not shown). The scene change detection device according to the present embodiment is appropriately read from a memory in which a program describing processing contents for realizing the function of each component is controlled by a central processing unit (CPU) provided in the computer. The function of each component of 1 can be realized by a computer. Here, the function of each component may be realized by a part of hardware.

  According to the present invention, a scene change having an arbitrary GOP structure can be detected only by a light processing by comparing the number of coding unit blocks without requiring a heavy calculation such as DCT. This is useful for applications that require detection of a scene change in a GOP structure moving image.

DESCRIPTION OF SYMBOLS 1 Scene change detection apparatus 11 Homogeneous frame determination process part 12 Scene change detection part 111 Homogeneous frame extraction part 112 Block number calculation part 113 Scene change area detection part

Claims (2)

A scene change detection device for detecting a scene change of a moving image,
The change in the statistics were aggregated for each frame the number of put that coded unit blocks between successive homologous frame in the moving picture, inter determines allogeneic frames compared with a threshold value corresponding to the frame type of the scene change interval determination processing Means,
Scene change detection means for specifying a scene change frame by an overlapping section of the scene change section determined by all frame types, and
The same type inter-frame determination processing means includes:
The number of coding unit blocks of the minimum size among all block sizes constituting a frame is used as the statistic, and is compared with a threshold according to the frame type to determine the presence or absence of the scene change section,
When the moving picture has a hierarchical structure including one or more lower frames of the P frame, the B frame, and the non-reference B frame with respect to the highest I frame as the GOP structure, the highest level of the hierarchical structure The scene change section based on the change of the statistic is determined in order from the I frame located in the frame, and the presence or absence of the scene change section is determined in the lower frame only when it is determined that the scene change section exists in the upper frame. Means,
Means for determining a scene change section based on a change in the statistic for each successive I frame, when the moving image has a GOP structure and consists entirely of I frames;
Scene change detecting device according to claim Rukoto to have a. A program for causing a computer to function as the scene change detection device according to claim 1 .