JP7291478B2 - Video encoding device and its program - Google Patents

Description

The present invention relates to a video encoding device and its program.

In order to transmit moving images (video) over broadcast waves, communication lines, etc., and to store them on recording media, research has been conducted on moving image coding methods that compress and encode moving images. there is
Conventional moving picture coding methods include MPEG (Moving Picture Experts Group)-1, MPEG-2, MPEG-4, AVC (Advanced Video Coding)/ITU-T Recommendation H.264. AVC/H.264 standard. 264, etc.
Also, as the latest international standardization of moving image coding, HEVC (High Efficiency Video Coding)/ITU-T Recommendation H.264. HEVC/H.265 standard. H.265 (see Non-Patent Document 1).
As a next-generation international standard, there is VVC (Versatile Video Coding), which is being developed by JVET (Joint Video Exploration Team) (see Non-Patent Document 2).

HEVC/H. H.265 and other conventional video coding systems use inter prediction (inter-frame prediction) that uses temporal correlation between video frames and intra prediction (intra-frame prediction) that uses spatial correlation within a frame. are appropriately switched to generate a predicted image. In the conventional moving image encoding method, the encoded data of the moving image is compressed by encoding the difference between the original image and the predicted image.
When inter-prediction is used, prediction errors between frames propagate in the time direction.
In order to prevent this error from propagating, conventional video coding methods employ an intra-refresh technique in which intra-prediction-encoded frames are periodically inserted into inter-prediction-encoded coded data. is used.

Note that intra refresh is not necessarily performed periodically. For example, a technique is disclosed in which the transmitting device performs intra-refresh at the timing of receiving a report such as packet loss of encoded data notified to the transmitting device (encoding device) from the receiving device (decoding device) ( See Patent Document 1).

Japanese translation of PCT publication No. 2014-528200

Sakae Okubo [Supervisor], "H.265/HEVC Textbook", Impress Japan Co., Ltd., October 21, 2013 “Versatile Video Coding (VVC)” [online], Fraunhofer Heinrich Hertz Institute for Communication Technology, [searched November 21, 2018], Internet <URL: https://jvet.hhi.fraunhofer.de>

In the conventional moving image coding method, intra frames are inserted periodically to prevent error propagation while suppressing the code amount.
However, in the conventional video encoding method, if the image distortion that accompanies the passage of time within the cycle of inserting the intra frame, for example, the distortion due to the fluctuation of the water surface in the video image of the bottom of the water taken from above the water surface, , the quality of the encoded image is degraded. Moreover, if many intra frames are inserted in order to avoid this problem, there is a problem that the code amount increases.

The present invention has been made in view of such problems. It is an object of the present invention to provide a moving image encoding device and a program therefor capable of encoding a moving image while suppressing degradation of image quality.

In order to solve the above-mentioned problems, a video encoding device according to the present invention is a video encoding device for encoding a video, comprising encoding means and intra refresh control means, wherein intra refresh control is performed. The means includes optical flow estimation means, distortion amount calculation means, distortion amount accumulation means, and refresh determination means.

In such a configuration, the moving image encoding device uses the optical flow estimation means to estimate motion vectors of pixels having similar features between frames as optical flows for each frame of the moving image by optical flow estimation using the gradient method. do. The optical flow estimating means performs optical flow estimation using the gradient method. Even if changes such as movement, rotation, enlargement/reduction occur between frames due to feature points containing gradients, motion between frames can be estimated. It becomes possible to detect vectors with high accuracy.

Then, the moving picture coding apparatus calculates the amount of distortion between frames based on the magnitude and direction of the motion vector estimated by the optical flow estimation means by the distortion amount calculation means. Furthermore, the moving image encoding apparatus calculates the accumulated distortion amount by accumulating the distortion amount for each frame by the distortion amount accumulating means.

Then, when the cumulative amount of distortion exceeds a predetermined threshold value, the moving picture encoding apparatus outputs a control signal instructing intra-refresh to the encoding section and initializes the cumulative amount of distortion. do. This causes the encoding means to perform intra-refresh and insert intra-frames.
The video encoding device can operate with a video encoding program for causing a computer to function as the means described above.

ADVANTAGE OF THE INVENTION This invention has the outstanding effect shown below.
According to the present invention, by performing intra-refresh before the accumulated amount of image distortion of a moving image becomes large, it is possible to encode the moving image while suppressing deterioration in image quality.
As a result, the present invention minimizes the insertion of intra frames and suppresses the deterioration of the image quality to reproduce a moving image even in a moving image that rotates, expands, and shrinks, such as fluctuations in the surface of the water. can be encoded.

1 is a block configuration diagram showing the configuration of a video encoding device according to an embodiment of the present invention; FIG. 2 is a block configuration diagram showing the configuration of intra refresh control means in FIG. 1; FIG. FIG. 10 is an explanatory diagram for explaining an optical flow estimation result; 2 is a block configuration diagram showing the configuration of encoding means in FIG. 1; FIG. 4 is a flow chart showing the operation of the video encoding device according to the embodiment of the present invention; FIG. 10 is a block configuration diagram showing the configuration of a moving image encoding device according to an embodiment of a modified example of the present invention; FIG. 2 is an explanatory diagram for explaining the difference in image quality of encoded data between the video encoding device according to the embodiment of the present invention and a conventional video encoding device, and FIG. FIG. 2B is an example of an image obtained by encoding and decoding a moving image by the apparatus, and FIG.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the drawings.
<<Configuration of Video Encoding Apparatus>>
First, referring to FIG. 1, the configuration of a video encoding device 1 according to an embodiment of the present invention will be described.
A moving image encoding device 1 compresses and encodes a moving image, which is a video signal. As shown in FIG. 1, the video encoding device 1 includes intra-refresh control means 2 and encoding means 3 .

The intra-refresh control means 2 controls the timing of intra-refresh for inserting the intra-frame in the coding means 3 into the coded data.
The intra-refresh control means 2 accumulates the amount of distortion of the moving image over time for each frame, and outputs a control signal instructing to perform intra-refresh when the accumulated amount of distortion exceeds a predetermined threshold value. Output to the encoding means 3 .

The encoding means 3 encodes moving images by an intra-refresh method.
The encoding means 3 uses HEVC/H. Intra refresh, which periodically inserts intra frames, which are frames encoded by intra-prediction (intra-frame prediction), into encoded data encoded by inter-prediction (inter-frame prediction) using a video encoding method such as H.265. Using the method of compression encoding the moving image. In addition to periodically inserting an intra frame into the encoded data, the encoding unit 3 inserts an intra frame into the encoded data at the timing when intra refresh is instructed by a control signal from the intra refresh control unit 2. do.
As a result, the video encoding device 1 can prevent the image distortion propagated by the inter-prediction code from increasing within the period of the intra-frame.
A detailed configuration of the video encoding device 1 will be described below.

<Intra Refresh Control Means>
The configuration of the intra-refresh control means 2 of the video encoding device 1 will be described with reference to FIG.
As shown in FIG. 2, the intra refresh control means 2 includes an optical flow estimation means 20, a distortion amount calculation means 21, a distortion amount accumulation means 22, an accumulated distortion amount storage means 23, and a refresh determination means 24. Prepare.

The optical flow estimating means 20 estimates an optical flow between frames input in time series of moving images. An optical flow is a motion vector between frames of pixels that are feature points such as edges.
The optical flow estimator 20 estimates motion vectors of pixels having similar features between frames as optical flows by a general optical flow estimation method using a gradient method. For example, the optical flow estimator 20 calculates motion vectors by the Lucas-Kanade method.

Note that the optical flow estimating means 20 searches for pixels with similar local feature amounts including gradients between frames by SIFT (Scale-Invariant Feature Transform), SURF (Speeded-Up Robust Features), or the like. A vector may also be obtained.

In this way, the optical flow estimating means 20 uses the motion vector calculated using the feature amount including the gradient as the optical flow. The optical flow can be estimated even when changes in .
As a result, the optical flow estimator 20 can obtain a motion vector even for moving images that involve changes such as movement, rotation, enlargement, and reduction.

FIG. 3 shows an example of images of frame F(t) at time t and frame F(t+1) at time (t+1).
The optical flow estimator 20 estimates the optical flow between frame F(t) and frame F(t+1). For example, in the example of FIG. 3, the optical flow estimation means 20 determines that pixels P1, P2, and P3 of frame F(t) are Three motion vectors are calculated to indicate that the pixel has been moved to pixels P1', P2' and P3'. Note that the three feature points shown here are merely examples, and there are actually more feature points.

The optical flow estimating means 20 calculates the magnitude L (the number of pixels) of the motion vector, which is the estimated optical flow, and the direction θ (angle: 0°≦θ<360°) of the motion vector, the number of the estimated feature points Only n are output to the distortion amount calculation means 21 as optical flow estimation results.

The distortion amount calculator 21 calculates the distortion amount of the image for each frame from the optical flow estimated by the optical flow estimator 20 .
The distortion amount calculator 21 calculates the amount of distortion from the previous frame based on the degree of dispersion of the magnitude and direction of the motion vector, which is the optical flow estimated by the optical flow estimator 20 . Here, the distortion amount calculation means 21 calculates the amount of distortion for each frame by calculating the variance of the magnitude of the motion vector and the variance of the direction of the motion vector and multiplying them.
Specifically, the distortion amount calculation means 21 calculates the variance _sL of the motion vector magnitudes L (L ₁ , L ₂ , . . . , L _n ) using the following equation (1).

The distortion amount calculation means 21 also calculates the variance _s θ of the motion vector directions θ (θ ₁ , θ ₂ , . . . , θ _n ) by the following equation (2). Note that the strain amount calculating means 21 calculates the variance s _θ in the range of 0°≦θ≦180° in order to prevent cancellation due to strain in the opposite direction. Specifically, for the direction θ of 180°<θ<360°, θ is converted to (360°−θ) and the following equation (2) is calculated.

In equations (1) and (2), n is the number of optical flows (motion vectors) for one frame estimated by the optical flow estimation means 20 . α is a weighting factor of the magnitude of the motion vector in the amount of distortion. β is a weighting factor in the direction of the motion vector in the amount of distortion. These weighting coefficients α and β are constants that respectively set the weight of the magnitude and the weight of the direction of the motion vector in the amount of distortion. For example, α=β=1, or α=10 and β=1. The weighting factors α and β may be set externally.
Then, the distortion amount calculating means 21 calculates the variance s _L of the motion vector magnitude L calculated by the formula (1) and the variance of the motion vector direction θ calculated by the formula (2), using the following formula (3): s _θ is multiplied to calculate the strain amount d.

Thereby, the distortion amount calculation means 21 can quantitatively calculate the distortion with respect to the previous frame for each frame.
The distortion amount calculation means 21 outputs the distortion amount d for each frame calculated by the equation (3) to the distortion amount accumulating means 22 .

The distortion amount accumulating means 22 accumulates the distortion amount for each frame calculated by the distortion amount calculating means 21 .
The strain amount accumulating means 22 reads the strain amount accumulated up to the previous frame from the accumulated strain amount storage means 23 and adds the strain amount input from the strain amount calculating means 21 to obtain the accumulated strain amount up to the current frame ( Cumulative strain amount) is calculated.
The strain amount accumulating means 22 stores the accumulated strain amount in the accumulated strain amount storage means 23 and outputs it to the refresh determination means 24 .

The cumulative strain amount storage means 23 stores the cumulative strain amount accumulated by the strain amount accumulating means 22 . The cumulative strain amount storage means 23 can be configured with a general storage medium such as a semiconductor memory.
In the cumulative strain amount storage means 23, the cumulative strain amount is updated by the strain amount accumulating means 22, and the cumulative strain amount is initialized ("0") by the refresh determination means 24. FIG. It is assumed that the cumulative distortion amount stored in the cumulative distortion amount storage means 23 is initialized in advance when the video encoding device 1 is activated.

The refresh determination means 24 determines whether or not to instruct the encoding means 3 to perform intra refresh.
The refresh determination means 24 outputs a control signal instructing intra refresh to the encoding means 3 when the accumulated distortion amount accumulated by the distortion amount accumulation means 22 exceeds a predetermined threshold value. A threshold that serves as a reference for instructing this intra-refresh may be set externally.

The refresh determination means 24 initializes the accumulated distortion amount stored in the accumulated distortion amount storage means 23 when the control signal instructing intra refresh is output to the encoding means 3 .
In this way, the intra refresh control means 2 can accumulate the image distortion amount in frame units of the moving image, and instruct the coding means 3 to perform intra refresh when the accumulated distortion amount exceeds a predetermined threshold value. .

<Encoding Means>
The configuration of the encoding means 3 of the video encoding device 1 will be described with reference to FIG. Note that the encoding means 3 can be MPEG-2, AVC/H. 264, HEVC/H. Any encoding method such as H.265 may be used. Here, as an example, HEVC/H. An example using the H.265 coding system will be described. Note that HEVC/H. H.265 is well known from Non-Patent Document 1 and the like, so detailed description thereof will be omitted.

As shown in FIG. 4, the encoding means 3 includes a subtracting means 30, a transforming means 31, a quantizing means 32, an inverse quantizing means 33, an inverse transforming means 34, an adding means 35, and a filtering means 36. , frame storage means 37 , prediction means 38 , and entropy coding means 39 .

The subtraction means 30 calculates a difference image between the input frame image of the moving image and the predicted image predicted by the prediction means 38 . The subtraction means 30 outputs the calculated difference image to the conversion means 31 .

The transforming means 31 orthogonally transforms the difference image generated by the subtracting means 30 for each block of a predetermined size, and calculates an orthogonal transform coefficient. This orthogonal transform is, for example, a discrete cosine transform (DCT). The transform means 31 outputs the calculated orthogonal transform coefficients to the quantization means 32 .

The quantization means 32 quantizes the orthogonal transform coefficients calculated by the transform means 31 based on a preset quantization table. The quantization means 32 outputs the quantized orthogonal transform coefficients (quantized orthogonal transform coefficients) to the inverse quantization means 33 and the entropy encoding means 39 .

The inverse quantization means 33 performs inverse quantization, which is the reverse of the process performed by the quantization means 32 , on the quantized orthogonal transform coefficients quantized by the quantization means 32 . The inverse quantization means 33 outputs the inversely quantized orthogonal transform coefficients to the inverse transform means 34 .

The inverse transforming means 34 performs inverse orthogonal transform (for example, inverse discrete cosine transform), which is the reverse of the process performed by the transforming means 31, on the orthogonal transform coefficients inversely quantized by the inverse quantizing means 33. is. The image for each block transformed by the inverse transformation means 34 becomes a prediction error image for a decoded image decoded by a decoding device (not shown) that decodes the encoded data output from the moving image encoding device 1 . The inverse transforming means 34 outputs the prediction error image generated by the inverse transforming to the adding means 35 .

The addition means 35 adds the prediction error image generated by the inverse transformation means 34 and the prediction image predicted by the prediction means 38 . The addition unit 35 adds the prediction error image and the prediction image to generate a decoded image to be decoded by a decoding device (not shown). The adding means 35 outputs the generated decoded image to the filtering means 36 and the predicting means 38 .

Filtering means 36 removes distortion (block distortion, ringing, etc.) from the decoded image by filtering the decoded image generated by adding means 35 . Filters used in the filter means 36 are, for example, a deblocking filter, a sampling adaptive offset (SAO: Sample Adaptive Offset), and the like. The filter means 36 outputs the filtered decoded image to the frame storage means 37 .

The frame storage means 37 stores the decoded image from which the distortion has been removed by the filter means 36 as a frame image. The frame images stored in the frame storage means 37 are referred to by the prediction means 38 when generating predicted images by inter prediction.

The prediction means 38 generates a predicted image for the current frame, which is the object of encoding of the moving image. The prediction means 38 includes an intra prediction means 380 , an inter prediction means 381 and a switching means 382 .

The intra prediction unit 380 performs intra prediction (intra-frame prediction) to generate a prediction image for a frame to be encoded (encoding target frame) of an input moving image. The intra prediction means 380 performs intra prediction using the decoded image (reference frame) generated by the addition means 35 . The intra prediction means 380 generates a predicted image from the decoded image by using three algorithms of planar prediction, direct current (DC) prediction, and directional prediction. The intra prediction means 380 outputs the generated predicted image to the switching means 382 .

The inter prediction unit 381 performs inter prediction (inter-frame prediction) to generate a prediction image for a frame to be encoded (encoding target frame) of an input moving image. The inter prediction means 381 detects the motion of the image from the input frame to be encoded of the moving image and the decoded image (reference frame) of the previous frame stored in the frame storage means 37, and predicts the motion between the motion compensation frames. A predicted image is generated by making a prediction. The inter prediction means 381 outputs the generated predicted image to the switching means 382 .

The switching means 382 switches the predicted image to be output between the predicted image predicted by the intra prediction means 380 and the predicted image predicted by the inter prediction means 381 .
The switching means 382 selects the predicted image (intra frame) predicted by the intra prediction means 380 periodically, for example, once every 32 frames, and selects the predicted image predicted by the inter prediction means 381 at other timings. is selected and output to the subtracting means 30 .
In addition, independently of periodic switching, the switching means 382 changes the predicted image (intra frame) predicted by the intra prediction means 380 at the timing when intra refresh is instructed by the control signal from the intra refresh control means 2. Select and output to the subtracting means 30 .

The prediction means 38 outputs to the entropy encoding means 39 a parameter indicating whether intra prediction or inter prediction switched by the switching means 382 is performed as one of the encoding parameters.

The entropy coding means 39 performs entropy coding on the quantized orthogonal transform coefficients generated by the quantization means 32 and the coding parameters input from the prediction means 38 to generate coded data. The entropy coding means 39 generates coded data by, for example, context-based adaptive binary arithmetic coding (CABAC) as entropy coding.
Here, the entropy encoding means 39 receives encoding parameters only from the prediction means 38, but the quantization processing parameters in the quantization means 32, the filtering parameters in the filtering means 36, etc. Descriptions of parameters that are not directly related to are omitted.
In this way, the coding means 3 can insert an intra frame into the coded data at the timing when intra refresh is instructed by the control signal from the intra refresh control means 2 .

As described above, the moving image encoding device 1 performs intra-refresh when the accumulated distortion amount of the moving image exceeds a predetermined threshold value, suppresses the propagation of distortion between frames, and thereby prevents deterioration of image quality. Subdued encoded data can be generated.
The moving image encoding device 1 can be operated by a program (moving image encoding program) for causing a computer to function as each means described above.

<<Operation of Video Encoding Apparatus>>
Next, with reference to FIG. 5 (refer to FIGS. 1 and 2 as necessary for the configuration), the operation of the video encoding device 1 according to the embodiment of the present invention will be described. Here, the operation of the intra-refresh control means 2, which is the main component of the present invention, will be mainly described.

In step S1, the optical flow estimator 20 estimates the optical flow between frames input in time series of moving images by a gradient method (for example, the Lucas-Kanade method). Thereby, the optical flow estimation means 20 estimates motion vectors (magnitude, direction) of a plurality of feature points as optical flows within the frame. Although illustration is omitted here, the video encoding device 1 retains only the first frame of the video in the optical flow estimating means 20, and does not perform optical flow estimation. The operation proceeds to S8.

In step S2, the distortion amount calculation means 21 calculates the amount of distortion from the previous frame based on the degree of dispersion of the magnitudes and directions of the motion vectors of the plurality of feature points that are the optical flows estimated in step S1.
Specifically, the distortion amount calculation means 21 calculates the variance of the magnitude of the motion vector using the above equation (1), and calculates the variance of the direction of the motion vector using the above equation (2).
Then, the distortion amount calculation means 21 calculates the distortion amount by multiplying the variance of the magnitude of the motion vector by the variance of the direction of the motion vector according to the equation (3).

In step S3, the strain amount accumulating means 22 reads out the strain amount accumulated up to the previous frame from the accumulated strain amount storage means 23, and adds the strain amount calculated in step S2 to obtain the accumulated strain amount up to the current frame. (Total strain amount) is calculated.
In step S4, the strain amount accumulating means 22 stores the accumulated strain amount (accumulated strain amount) in the accumulated strain amount storage means 23.

In step S5, the refresh determination means 24 determines whether or not the cumulative strain amount exceeds a predetermined threshold.
Here, if the accumulated distortion amount does not exceed the threshold value (No in step S5), the moving image encoding device 1 advances the operation to step S8.
On the other hand, if the accumulated distortion amount exceeds the threshold value (Yes in step S5), the refresh determination means 24 outputs a control signal instructing intra refresh to the encoding means 3 in step S6.
In step S7, the refresh determination means 24 initializes the cumulative strain amount value stored in the cumulative strain amount storage means 23 to "0".

In step S8, the encoding means 3 encodes the input moving image frame. At this time, when instructed to perform intra refresh by a control signal from the intra refresh control means 2, the encoding means 3 inserts an intra frame by intra prediction into the encoded data.

In step S9, when the next frame is input to the video encoding device 1 (Yes in step S9), the video encoding device 1 returns to step S1 and continues its operation.
On the other hand, if the next frame is not input (No in step S9), the video encoding device 1 terminates its operation.
Through the above operation, the video encoding device 1 performs intra-refresh when the accumulated distortion amount of the video exceeds a predetermined threshold value while the video is being input, thereby preventing the propagation of distortion between frames. can be suppressed, and encoded data can be generated with suppressed degradation of image quality.

<<Modification>>
Although the configuration and operation of the video encoding device 1 according to the embodiment of the present invention have been described above, the present invention is not limited to this embodiment.
Here, the moving picture encoding device 1 is configured to include the intra-refresh control means 2 and the encoding means 3 inside the device.
However, the intra-refresh control means 2 and the encoding means 3 may be configured separately. Specifically, as shown in FIG. 6, an intra-refresh control device 2B and a coding device 3B may constitute a video coding device 1B.
The intra refresh control device 2B has the same configuration as the intra refresh control means 2 described in FIG. Further, the encoding device 3B has the same configuration as the encoding means 3 explained in FIG.

When the video encoding device 1B is separated in this way, the intra-refresh control device 2B can operate the computer with a program (intra-refresh control) for functioning as each means described in FIG. can. Further, the encoding device 3B can be operated by a program (encoding program) for causing a computer to function as each means described in FIG.
By separating the intra-refresh control device 2B in this way, even if the encoding method is changed, the same intra-refresh control device 2B can be used by simply replacing the encoding device 3B. become.
Other actions and effects are the same as those of the video encoding device 1 .

Also, here, the encoding unit 3 is assumed to periodically insert an intra frame in advance. However, the coding means 3 may insert an intra frame into the coded data only by the control signal from the intra refresh control means 2 .

≪Evaluation image≫
Finally, referring to FIG. 7, the difference in image quality between the encoded data generated by the video encoding device 1 and the encoded data generated by the conventional video encoding device was determined by decoding the encoded data. Description will be made using a decoded image.
FIG. 7(a) is an example of an image at a certain point in time when a moving image is encoded and decoded by a conventional moving image encoding device. FIG. 7(b) is an example of an image at the same point in time as in FIG. 7(a), which is obtained by encoding and decoding the same moving image as in FIG. 7(a) by the moving image encoding apparatus 1 of the present invention.
Here, the moving image is obtained by submerging a board with electronic components embedded in it in the bottom of water and photographing the board from above the surface of the water.
The image I1 shown in FIG. 7(a) is degraded in image quality compared to the image I2 shown in FIG. 7(b). For example, each leg cannot be identified in the IC chip foot portion E1 on the image I1, but each foot can be clearly identified in the IC chip foot portion E2 in the image I2.

As shown in FIG. 7(a), the conventional moving image coding apparatus has a moving image in which the movement of the subject changes not only by simple movement but also by rotation, enlargement/reduction, etc. due to the fluctuation of the surface of the water, such as a moving image of the bottom of the water. could not be encoded with high quality.
However, in the video encoding apparatus 1 according to the present invention, even if the subject changes due to rotation, enlargement/reduction, etc., the motion vectors of the feature points by the optical flow can be used to accurately detect the distortion of the image, and the cumulative distortion can be calculated. Intra refresh can be performed when the amount exceeds the threshold.
As a result, the moving picture coding apparatus 1 can code the moving picture while suppressing deterioration in image quality, as shown in FIG. 7(a).

1, 1B video encoding device 2, 2B intra refresh control means (intra refresh control device)
20 optical flow estimation means 21 distortion amount calculation means 22 distortion amount accumulation means 23 cumulative distortion amount storage means 24 refresh determination means 3, 3B encoding means (encoding device)
30 Subtracting Means 31 Transforming Means 32 Quantizing Means 33 Inverse Quantizing Means 34 Inverse Transforming Means 35 Adding Means 36 Filtering Means 37 Frame Storage Means 38 Predicting Means 380 Intra Predicting Means 381 Inter Predicting Means 382 Switching Means 39 Entropy Encoding Means

Claims (4)

A video encoding device for encoding a video,
encoding means for encoding the moving image by an intra-refresh method;
an intra-refresh control means for controlling the timing of performing intra-refresh of the encoding means;
The intra-refresh control means includes:
optical flow estimation means for estimating, as an optical flow, motion vectors of pixels having similar features between the frames by optical flow estimation using a gradient method for each frame of the moving image;
Distortion amount calculation means for calculating an amount of distortion between frames based on the magnitude and direction of the motion vector estimated by the optical flow estimation means;
a strain amount accumulating means for calculating a cumulative strain amount by accumulating the strain amounts calculated by the strain amount calculating means for each frame;
Refresh determination means for outputting a control signal for instructing intra refresh to the encoding means and initializing the accumulated distortion amount when the accumulated distortion amount accumulated by the distortion amount accumulation means exceeds a predetermined threshold value. and
The distortion amount calculation means calculates the distortion amount by calculating the variance of the magnitude of the motion vector estimated as the optical flow and the variance of the direction of the motion vector and multiplying them. Video encoding device. The distortion amount calculation means is characterized in that, for a direction θ in which the direction of the motion vector is 180°<θ<360°, θ is converted to (360°−θ) to calculate the variance of the direction. The moving picture encoding device according to claim 1 . 3. The distortion amount calculation means according to claim 1 , wherein the variance of the magnitude of the motion vector and the variance of the direction of the motion vector are multiplied by a predetermined weighting factor. video encoding device. A video encoding program for causing a computer to function as the video encoding device according to any one of claims 1 to 3 .

Images