Image quality estimation device and program

Abstract

To provide an image quality estimation device and program capable of obtaining a relation between video quality and a bit rate at a fixed calculation cost or less.SOLUTION: An image quality estimation device includes a parameter derivation unit configured to derive a first parameter according to a constant and resolution of a video, and a second parameter according to encoding results obtained by encoding the video once. The image quality estimation device includes a curve estimation unit configured to estimate a relation between a bit rate and video quality of the video based on the constant, the first parameter, and the second parameter.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to an image quality estimation device and a program. [Background technology]

[0002] The radio wave bandwidth available for video distribution services is limited. For this reason, video may be compressed and encoded. In this case, if the encoding bit rate is set below a certain level, the image quality will deteriorate and the service quality will also fall below a certain level. On the other hand, if the encoding bit rate is set above a certain level, the amount of unnecessary information will increase and the radio wave bandwidth may be strained. For this reason, the video quality is monitored and encoded at the minimum necessary bit rate to maximize the service quality.

[0003] One of the methods for Video Quality Assessment (VQA) is the Full-Reference (FR) method. The FR method compares the original image with a degraded image (e.g., coded image) to objectively evaluate the quality of the degraded image. There are standard specifications for FR-type video quality assessment, such as ITU-T J341.

[0004] Examples of FR-type video quality assessment methods include PSNR (Peak Signal to Noise Ratio), SSIM (Structural Similarity Index Measure), and VMAF (Video Multimethod Assessment Fusion). Among these, VMAF has been confirmed to have a certain level of high correlation with MOS (Mean Opinion Score), a subjective image quality index (see Non-Patent Document 1 below). VMAF is a video quality assessment method that combines multiple known video quality assessment methods (e.g., VIF (Visual Information Fidelity)), weights each of them, and uses a learning algorithm to obtain assessment results. VMAF assesses each video on a scale from "0" (low image quality) to "100" (high image quality).

[0005] On the other hand, there is a technology in which a relationship between bit rate and resolution is stored in advance in a video distribution system, and video is distributed at a predetermined bit rate for each video content title (see Non-Patent Document 2 below). Non-Patent Document 2 shows, for example, that video with the same resolution as that of a fixed bit rate can be distributed at a lower bit rate than when video is distributed at a fixed bit rate.

[0006] Furthermore, predicted MOS (P-MOS) has been proposed, which predicts MOS, a subjective image quality index, from PSNR, an objective image quality index (see Non-Patent Document 3 below). Non-Patent Document 3 shows that by fitting the relationship between PSNR and MOS to a curve expressed as predicted MOS, the correlation with MOS can be increased to a certain level. [Prior art documents] [Non-patent literature]

[0007] [Non-Patent Document 1] Netflix Technology Blog: “Toward A Practical Perceptual Video Quality Metric”, 2016 / 6 / 6. [Non-patent document 2] Netflix Technology Blog: “Per-Title Encode Optimization”, 2015 / 12 / 14. [Non-patent document 3] M. Alizadeh and M. Sharifkhani, “Subjective video quality prediction based on objective video quality metrics”, 2018 4th Iranian Conference on Signal Processing and Intelligent Systems (ICSPIS), Tehran, Iran, 2018 / 12, pp. 7-9, doi: 10.1109 / ICSPIS.2018.8700561. Summary of the Invention [Problem to be solved by the invention]

[0008] The above-mentioned FR-type video quality assessment method aims to assess the quality of encoded video, for example, and does not take into account the bit rate. In other words, the conventional FR-type video quality assessment method does not take into account the relationship between video quality and bit rate.

[0009] Although the above-mentioned Non-Patent Document 2 shows the relationship between bit rate and resolution, encoding must be performed under multiple bit rate conditions to obtain each bit rate, which increases the computational cost to a certain extent. In particular, the computational complexity of VVC (Versatile Video Coding), the latest encoding method, is greater than that of HEVC (High Efficiency Video Coding), the encoding method that preceded VVC. Therefore, when encoding using VVC, the computational cost is expected to increase more than ever before.

[0010] Furthermore, the above-mentioned Non-Patent Document 3 shows the relationship between PSNR and predicted MOS, but the predicted MOS is expressed by three parameters α, γ, and β. Therefore, to obtain the predicted MOS, at least three encoding operations are performed. Therefore, even in Non-Patent Document 3, the calculation cost is above a certain level.

[0011] An object of the present disclosure is to provide an image quality estimation device and program that are capable of obtaining the relationship between video quality and bit rate with calculation costs below a certain level. [Means for solving the problem]

[0012] An image quality estimation device according to a first aspect includes a parameter derivation unit that derives a constant, a first parameter corresponding to a resolution of an image, and a second parameter corresponding to an encoding result obtained by encoding the image once. The image quality estimation device also includes a curve estimation unit that estimates a relationship between a bit rate and image quality of the image based on the constant, the first parameter, and the second parameter.

[0013] A program according to the second aspect is a program that causes a computer to function as the image quality estimation device according to the first aspect. [Effects of the Invention]

[0014] According to the present disclosure, it is possible to provide a video quality estimation device and program that are capable of obtaining the relationship between video quality and bit rate at a calculation cost below a certain level. [Brief explanation of the drawings]

[0015] [Figure 1] FIG. 1 is a diagram showing an example of the configuration of an image quality estimation device according to the first embodiment. [Figure 2] 2A and 2B are diagrams showing the relationship between the bit rate and the image quality according to the first embodiment. [Figure 3] FIG. 3 is a diagram illustrating an example of operation according to the first embodiment. [Figure 4] FIG. 4 is a diagram illustrating another example of operation according to the first embodiment. [Figure 5] FIG. 5 is a diagram illustrating another example of operation according to the first embodiment. [Figure 6] FIG. 6 is a diagram illustrating an example of the configuration of a coded stream generating device according to the second embodiment. [Figure 7] FIG. 7 is a diagram illustrating an example of generating a bit rate ladder according to the second embodiment. [Figure 8] FIG. 8 is a diagram illustrating an example of operation according to the second embodiment. [Figure 9] FIG. 9 is a diagram illustrating an example of a bandwidth distribution according to the second embodiment. [Figure 10] FIG. 10 is a diagram illustrating another operation example 3 according to the second embodiment. [Figure 11] FIG. 11 is a diagram illustrating an example of multi-layer coding according to another operation example 4 of the second embodiment. [Figure 12] 12(A) and 12(B) are diagrams illustrating an example of a bit rate ladder according to the second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0016] [First embodiment] The image quality estimation device according to the embodiment will be described with reference to the drawings. In the description of the drawings, the same or similar parts are denoted by the same or similar reference numerals.

[0017] (Image quality estimation device according to the first embodiment) First, the image quality estimation device according to the first embodiment will be described.

[0018] FIG. 1 is a diagram showing an example of the configuration of an image quality estimation device 100. The image quality estimation device 100 shown in FIG. 1 estimates the relationship between the bit rate when video is compression-encoded (hereinafter sometimes referred to as "encoding") and the video quality of the video. In particular, the image quality estimation device 100 according to the first embodiment estimates the relationship between video quality and bit rate using the encoding result obtained from a single encoding run. While Non-Patent Document 2 mentioned above calculates the predicted MOS through at least three encoding runs, the image quality estimation device 100 shown in FIG. 1 can estimate the relationship between video quality and bit rate through a single encoding run, making it possible to obtain the relationship between video quality and bit rate with calculation costs below a certain level.

[0019] In the first embodiment, the relationship between the bit rate and the video quality is expressed by the following formula:

[0020]

number

[0021] Furthermore, equation (1) makes it possible to obtain video quality ("Quality") that is compatible with any bit rate ("Rate"). Therefore, it can be said that equation (1) properly represents the correspondence between video quality and bit rate.

[0022] Figure 2 shows a graph when the horizontal axis is the bit rate and the vertical axis is the video quality for Equation (1). As shown in Figure 2, Equation (1) is represented as a sigmoid function. A sigmoid function is an S-shaped curve and is a point-symmetric function centered around the inflection point. For a sigmoid function, if the inflection point is (xs, ys), it is convex downward in the range of x < xs, convex upward in the range of x > xs, and has the maximum slope at x = xs. Also, for a sigmoid function, if the asymptotes are y = y1 and y2, it is a curve that approaches y = y1 as x increases and approaches y = y2 as x decreases. However, in Figure 2, only a part of the sigmoid function is shown. Specifically, Figure 2 is a graph of the part above the inflection point (the point on the curve when "Quality" is "0").

[0023] Thus, the relationship between the video quality and the bit rate is shown by a curve (Figure 2(A)), and the curve is shown as a (part of) sigmoid function.

[0024] As shown in Figure 2(A), the curve includes an ascending region and a relaxation region. The ascending region indicates the region on the curve where the video quality ("Quality") increases with a certain slope or more as the bit rate ("Rate") increases. On the other hand, the relaxation region indicates the region on the curve where the curve approaches the asymptote. In the ascending region, the video quality also increases with a certain slope or more as the bit rate increases, but in the relaxation region, compared with the ascending region, the video quality only approaches the asymptote and does not increase so much even as the bit rate increases.

[0025] As shown in equation (1), it includes three parameters a, b, and c. Parameter a represents the asymptote for the curve, as shown in Figure 2(A). That is, parameter a represents the video quality that the curve asymptote to as the bit rate increases. Parameter b represents the bit rate when the video quality is "0." Parameter c represents a parameter related to the slope of the ascending region of the curve. That is, the three parameters a, b, and c shown in equation (1) are parameters that represent the characteristics of the curve (Figure 2(A)) that shows the relationship between video quality and bit rate. It can be said that the curve shown in Figure 2(A) is characterized by the three parameters a, b, and c.

[0026] In the first embodiment, by specifying the parameters a and b as constant values, it is possible to calculate equation (1) from one data point, which means that it is possible to estimate video quality from a single bit rate condition.

[0027] As shown in FIG. 1, the image quality estimation device 100 includes a parameter derivation unit 110 and a curve estimation unit 120.

[0028] The parameter derivation unit 110 derives the three parameters a, b, and c in equation (1). The parameter derivation unit 110 includes a parameter a derivation unit 111, a parameter b derivation unit 112, a target bit rate determination unit 113, a video encoding unit 114, and a parameter c derivation unit 115.

[0029] (Derivation of parameter a) The parameter a derivation unit 111 derives the parameter a. As described above, the parameter a represents an asymptote for the curve. In the first embodiment, the video quality (Quality) is assumed to be VMAF, and therefore the parameter a can range from "0" to "100". Since the "Quality" becomes "100" when the "Rate" is made very large, the parameter a derivation unit 111 derives "100" as the parameter a. In this way, the parameter a is a constant. The parameter a derivation unit 111 may derive the parameter a by reading the parameter a stored in the memory from the memory. The parameter a derivation unit 111 outputs the derived parameter a to the curve estimation unit 120.

[0030] (Derivation of parameter b) The parameter b derivation unit 112 derives parameter b (first parameter). As described above, parameter b represents the "Rate" (bit rate) when "Quality" (video quality) becomes "0". The bit rate when video quality becomes "0" is expected to differ depending on the content. In the first embodiment, it is assumed that the variation in the bit rate for each content is smaller than a constant. However, in the first embodiment, the bit rate (i.e., parameter b) depends on the video resolution. Therefore, the parameter b derivation unit 112 derives parameter b, for example, as follows.

[0031]

number

[0032] In this way, the parameter b derivation unit 112 according to the first embodiment derives the parameter b according to the resolution of the video. The parameter b derivation unit 112 may acquire the resolution from the input video and derive the parameter b according to the resolution using equation (2). The parameter b derivation unit 112 outputs the parameter b to the curve estimation unit 120. However, the parameter b may be a fixed value rather than a value according to the resolution. In this case, the parameter b derivation unit 112 may derive the parameter b by reading the parameter b (fixed value) stored in the memory from the memory.

[0033] (Derivation of parameter c) As described above, the parameter c (second parameter) represents a parameter related to the slope of the ascending region of the curve. In the first embodiment, the parameter c is derived from the encoding result when the input video is encoded. Here, the (actual) bit rate when the input video is encoded at the target bit rate Rt is R, and the video quality (VMAF) of the encoded video is V. When the formula (1) is transformed, the following formula (3) is obtained.

[0034]

number

[0035] It is desirable to select the target bit rate Rt as a bit rate in the rising region shown in Figure 2(A). The curve shown in Figure 2(A) is expected to be different depending on the content, but it is the rising region rather than the relaxation region that characterizes the curve, and by using the numerical values ​​in the rising region as much as possible, it becomes possible to identify the curve for each content shown in Figure 2(A). The bit rate in the rising region is expected to differ for each input video, and the target bit rate Rt is expected to differ for each input video. In the first embodiment, it is assumed that the target bit rate Rt depends on the resolution. Therefore, in the first embodiment, the target bit rate Rt is determined using the following equation (4), as an example.

[0036]

number

[0037] The video encoding unit 114 encodes the input video so that the video reaches the target bit rate Rt. The video encoding unit 114 performs encoding only once. As an example, the encoding method may be VVC, but other methods capable of compression encoding may also be used.

[0038] The video encoding unit 114 measures the bit rate of the encoded video ("R" in equation (3)). The bit rate of the encoded video may be measured using a known method. The video encoding unit 114 also measures the video quality using VMAF based on the input video and the encoded video. The measurement of the video quality using VMAF may also be performed using a known method. The video encoding unit 114 outputs the bit rate R and VMAF(V) to the parameter c derivation unit 115.

[0039] The parameter c derivation unit 115 derives the parameter c using equation (3). As shown in equation (3), the parameter c is a parameter corresponding to the encoding result (for example, the bit rate R and VMAF(V)). Specifically, the parameter c derivation unit 115 derives the parameter c by substituting the bit rate R and VMAF(V) received from the video encoding unit 114 into equation (3). The parameter c derivation unit 115 outputs the parameter c to the curve estimation unit 120.

[0040] The curve estimation unit 120 estimates the relationship between bit rate and video quality based on three parameters a, b, and c. Specifically, the curve estimation unit 120 estimates a curve (FIG. 2(A)) showing the relationship between bit rate and video quality using equation (1). FIG. 2(B) shows a curve (measured value) showing the relationship between bit rate and video quality experimentally obtained by encoding 4K video, and the estimated curve (predicted value). The "data points used for prediction" are R=1.0 Mbps and VMAF=90.5. Since the input video is 4K video, parameter a=100 and parameter b=0.1, and parameter c is calculated from equation (3). The estimated curve (predicted value) can be calculated from the three parameters a, b, and c and equation (1).

[0041] The image quality estimation device 100 may output a curve showing the relationship between bit rate and video quality as a Rate-Quality curve. The Rate-Quality curve may be transmitted to, for example, a video content distribution system. The video content distribution system can use the Rate-Quality curve to distribute video content at a bit rate corresponding to VMAF. For example, the video content distribution system can determine a lower limit value for VMAF and distribute video content at a bit rate corresponding to the lower limit value. Alternatively, the image quality estimation device 100 may include a video content distribution unit. In this case, the video content distribution unit, like the video content distribution system, can input the Rate-Quality curve output from the curve estimation unit 120 and distribute video content using the Rate-Quality curve.

[0042] (Operation example according to the first embodiment) Next, an example of operation according to the first embodiment will be described.

[0043] FIG. 3 is a diagram illustrating an example of operation according to the first embodiment.

[0044] As shown in FIG. 3, in step S10, the image quality estimation device 100 starts processing.

[0045] In step S11, the parameter a derivation unit 111 derives the parameter a.

[0046] In step S12, the parameter b derivation unit 112 derives the parameter b. The parameter b derivation unit 112 may derive the parameter b according to the resolution of the input video.

[0047] In step S13, the target bit rate determination unit 113 determines the target bit rate Rt. The target bit rate determination unit 113 determines the target bit rate Rt according to the resolution of the input video.

[0048] In step S14, the video encoding unit 114 encodes the input video at the target bit rate Rt.

[0049] In step S15, the parameter c derivation unit 115 derives the parameter c. The parameter c derivation unit 115 inputs the encoding result (bit rate R and VMAF(V)) received from the video encoding unit 114 into equation (3) to derive the parameter c.

[0050] In step S16, the curve estimation unit 120 estimates a curve representing the relationship between the bit rate and the video quality based on the three parameters a, b, and c. The curve estimation unit 120 estimates the curve by substituting the three parameters a, b, and c into equation (1).

[0051] Then, in step S17, the image quality estimation device 100 ends the series of processes.

[0052] (Another operation example 1 according to the first embodiment) In the first embodiment, VMAF has been described as an example of an index representing video quality, but other objective image quality indices may be used as the index representing video quality. Examples of other objective image quality indices include the above-mentioned PSNR and SSIM. PSNR is a method for evaluating image quality by comparing an input video and an encoded video on a pixel-by-pixel basis and determining the average change. SSIM is a method for evaluating image quality based on changes in luminance values, contrast, and the structure of objects in the input video and the encoded video.

[0053] Furthermore, a subjective image quality index may be used as an index representing image quality in the first embodiment. The above-mentioned MOS may be used as the subjective image quality index. Taking MOS as an example, the following is given. That is, the minimum value of MOS is "1." Therefore, when image quality ("Quality") is expressed in MOS, Equation (1) can be transformed into Equation (5) below.

[0054]

number

[0055] Similarly, for the other indices mentioned above, by modifying equation (1) or setting parameters according to the range of possible values, it is possible to obtain a curve that can take on a range of values ​​corresponding to the other indices.

[0056] (Another operation example 2 according to the first embodiment) The three parameters a, b, and c described in the first embodiment may be changed depending on the performance of the encoding encoder included in the video encoding unit 114. In particular, the value of parameter b represents the bit rate when video quality is "0," and is therefore likely to be affected by the performance of the encoding encoder. For example, if the performance of the encoding encoder is higher than a certain level, it is expected that the video quality will be higher than a certain level even if the bit rate is lower than a predetermined value. Therefore, the parameter b derivation unit 112 sets the value of parameter b to a small value below a certain level for an encoding encoder with performance higher than a certain level. On the other hand, the parameter b derivation unit 112 sets the value of parameter b to a large value above a certain level for an encoding encoder with performance lower than a certain level. This makes it possible to maintain the accuracy of the relationship between video quality and bit rate at a certain level or higher, even for an encoding encoder with video quality below a certain level, just as in the case of an encoding encoder with performance higher than a certain level.

[0057] (Another operation example 3 according to the first embodiment) In the first embodiment, the video encoding unit 114 performs encoding at a target bit rate Rt to obtain encoded video. Regarding the bit rate R of the encoded video, when the bit rate R is in the relaxed region, the accuracy of the relationship between video quality and bit rate may decrease below a certain level. As shown in FIG. 2(A), in the relaxed region, compared to the increasing region, even if the bit rate increases, the video quality does not change much, and the accuracy of the relationship between video quality and bit rate is better when the relationship is obtained in the increasing region than when the relationship is obtained in the relaxed region.

[0058] Therefore, in the image quality estimation device 100 according to another operation example 3, when the bit rate R is in the relaxed region, encoding is performed again. By re-encoding, for example, it is possible to predict that the bit rate R will be in the increasing region, and therefore it is possible to prevent a decrease in the accuracy of the relationship between the video quality and the bit rate.

[0059] Fig. 4 is a diagram showing an operation example according to Alternative Operation Example 3. In Fig. 4, the same processes as those in the first embodiment are denoted by the same reference numerals.

[0060] As shown in FIG. 4, in step S20, video encoding unit 114 determines whether or not the bit rate R of the encoded video is in the increasing range. If the bit rate R is in the increasing range (Yes in step S20), the process proceeds to step S15. On the other hand, if the bit rate R is not in the increasing range (No in step S20) (i.e., if the bit rate R is in the relaxed range), video encoding unit 114 performs video encoding again in step S21. In this case, video encoding unit 114 performs encoding with a target bit rate of Rt / 2. Then, video encoding unit 114 outputs the bit rate R and VMAF(V) for the encoded video to parameter c derivation unit 115.

[0061] However, the determination of whether the bit rate R is in the increasing region may be made based on whether the VMAF of the encoded video is greater than a quality threshold (e.g., "90"). That is, if the VMAF of the encoded video is less than the quality threshold, the video encoding unit 114 determines that the bit rate R is in the increasing region (Yes in step S20). On the other hand, if the VMAF of the encoded video is greater than the quality threshold, the video encoding unit 114 determines that the bit rate R is in the relaxed region (No in step S20).

[0062] (Another Operation Example 4 According to the First Embodiment) Parameter b represents the bit rate at which the video quality is at its lowest. Therefore, for video that is difficult to encode, such as video that includes scene changes, it is expected that the video quality of the encoded video will be below a certain level, and the value of parameter b will be a small value below a certain level. On the other hand, for video that is easy to encode, it is expected that the video quality of the encoded video will be above a certain level, and the value of parameter b will be a large value above a certain level. Therefore, in another operation example 4, the parameter b derivation unit 112 derives parameter b according to not only the resolution but also the encoding difficulty of the encoded video. Specifically, the parameter b derivation unit 112 determines the encoding difficulty using the VMAF acquired from the video encoding unit 114. Figure 5 is a diagram showing another operation example 4 according to the first embodiment. The following will explain this using Figure 5.

[0063] That is, if the VMAF is less than the encoding threshold, the parameter b derivation unit 112 determines that the video is difficult to encode and sets the parameter b to a large value equal to or greater than a certain value (step S30). When the VMAF is low video quality and less than the encoding threshold, it is expected that the bit rate at the lowest video quality will be a low value equal to or less than a certain value. Therefore, by setting the parameter b to a large value equal to or greater than a certain value, it is possible to obtain the relationship between video quality and bit rate with the same accuracy as when the VMAF is equal to or greater than the encoding threshold.

[0064] On the other hand, if the VMAF is equal to or greater than the encoding threshold, the parameter b derivation unit 112 determines that the video is easy to encode and sets the parameter b to a small value less than a certain value (step S30). For video that is easy to encode, even if the parameter b is set to a small value less than a certain value, the VMAF is high video quality equal to or greater than the encoding threshold, so it is possible to accurately obtain the relationship between video quality and bit rate.

[0065] Specifically, the parameter b derivation unit 112 may derive the parameter b using the following equation (6).

[0066]

number

[0067] [Second embodiment] Next, a second embodiment will be described. In the second embodiment, an example will be described in which information indicating the relationship between video quality and bit rate (i.e., a Rate-Quality curve) is input from the image quality estimation device 100 described in the first embodiment, and a bit rate ladder is created based on the information.

[0068] (Bitrate Ladder) As described above, the bitrate ladder is a list showing the relationship between bitrate and resolution. Fig. 12(B) is a diagram showing an example of a bitrate ladder. The bitrates included in the bitrate ladder are the bitrates of the encoded video when the video is encoded. The bitrates are sometimes called target bitrates. As shown in Fig. 12(B), multiple correspondence relationships between bitrates and resolutions are shown.

[0069] A video (content) distributor can use the bit rate ladder to perform, for example, the following process. That is, the distributor creates a bit rate ladder, encodes the video (content) at the bit rate indicated in the bit rate ladder, and stores the encoded stream in storage in advance. Then, the distributor uses the bit rate ladder to distribute an encoded stream at a bit rate corresponding to a resolution available to the user. The user can view the video at the resolution indicated in the bit rate ladder by decoding the encoded stream.

[0070] As described above, when creating a bitrate ladder on the distribution side, video is encoded in advance at multiple bitrates (or bitrate conditions). The distribution side decodes the encoded video, measures the video quality of the decoded video, and checks whether the video quality is above a certain level.

[0071] However, it is expected that video with a higher resolution than a certain level, such as 4K or 8K, will be targeted for encoding. It is also expected that encoding methods with higher performance than previous encoding methods, such as VVC, will be used. Therefore, it is expected that the time required to create a bitrate ladder will become enormous.

[0072] Therefore, the second embodiment aims to enable generation of an appropriate bit rate ladder. Also, the second embodiment aims to enable generation of a bit rate ladder with a certain amount of calculation or less.

[0073] (Configuration example of an encoded stream generating device) Next, an example of the configuration of the coded stream generating device according to the second embodiment will be described.

[0074] 6 is a diagram illustrating an example of the configuration of the coded stream generating device 200. In the second embodiment, the coded stream generating device 200 creates a bit rate ladder. Then, the coded stream generating device 200 generates a coded stream using the bit rate ladder.

[0075] As described above, the bit rate ladder includes a resolution for each bit rate. In the second embodiment, the resolution may be determined arbitrarily. Therefore, in the following description, the resolutions included in the bit rate ladder may be omitted.

[0076] As shown in FIG. 6, the coded stream generating device 200 includes a bit rate ladder generating unit 210, a video coding unit 220, and a storage unit 230.

[0077] The bit rate ladder generator 210 inputs the curve (hereinafter sometimes referred to as the "Rate-Quality curve") output from the image quality estimation device 100, and generates a bit rate ladder using the Rate-Quality curve. As in the first embodiment, FIG. 2(A) shows an example of a Rate-Quality curve. The Rate-Quality curve in the second embodiment may be the same as that in the first embodiment. In the second embodiment, the Rate-Quality curve will be described as being represented by the following equation (8).

[0078]

number

[0079] (Bitrate ladder generation) Here, a method for generating a bit rate ladder in the bit rate ladder generating unit 210 will be described.

[0080] Focusing on Figure 2(A), let's assume that a bitrate included in the bitrate ladder is set in the relaxed range. In this case, even if encoding is performed using two adjacent bitrates included in the bitrate ladder, the user may not be able to distinguish the difference in video quality (especially subjective image quality such as MOS). Therefore, even if a bitrate is set in the relaxed range, it may not contribute to the user's quality of experience (QoE).

[0081] Therefore, it is desirable that the bit rates included in the bit rate ladder are included in the ascending range. However, even if a bit rate included in the ascending range is used, if the interval between bit rates becomes too wide or too narrow, it may be difficult for the user to perceive the advantage of changing the video quality according to the available bandwidth. Therefore, it is desirable that the bit rates be spaced appropriately in the ascending range.

[0082] Therefore, the bit rate ladder generator 210 according to the second embodiment generates a bit rate ladder as follows.

[0083] Fig. 7 is a diagram showing an example of generating a bit rate ladder according to the second embodiment. In Fig. 7, the horizontal axis represents bit rate ("Rate") and the vertical axis represents video quality ("Quality"). The curve shown in Fig. 7 is a Rate-Quality curve.

[0084] Generally, the bandwidth (bit rate) available for video distribution is limited, whether it is via wireless or the Internet. The lower limit of the bandwidth (bit rate) available for video distribution is R. L , the upper limit is R U Let's say.

[0085] In addition, the upper limit of video quality ("Quality") is set to Q t and the bit rate corresponding to the upper limit value Qt on the curve is Rmax.

[0086] At this time, the bit rate ladder generator 210 determines the lower limit R of the bandwidth available for video distribution. L The bit rate ladder generator 210 sets the upper limit R of the bandwidth available for video distribution. U (representing the bit rate) and the upper limit of video quality Q t The bit rate Rmax corresponding to U ,Rmax)) to the maximum bit rate R N-1Set to.

[0087] Then, the bit rate ladder generating unit 210 sets the video quality corresponding to the lowest bit rate R0 as the minimum value Qmin of the video quality, and sets the video quality corresponding to the highest bit rate R N-1 The video quality corresponding to this is set as the maximum video quality Qmax, and the video quality is divided equally between the minimum value Qmin and the maximum value Qmax so as to be at equal intervals.

[0088] However, the bit rate ladder generator 210 uses the following formula for the maximum video quality Qmax: Qmax=max(Q t ,f(R N-1 )) as the highest bitrate R N-1 The video quality f(R N-1 ) is the upper limit of video quality Q t Limit the number of users to the following:

[0089] The bit rate ladder generator 210 obtains, from the Rate-Quality curve, bit rates that provide each video quality obtained by dividing the video quality at equal intervals between the minimum video quality value Qmin and the maximum video quality value Qmax. Then, the bit rate ladder generator 210 generates a bit rate ladder including the plurality of bit rates. The bit rate ladder includes a minimum bit rate R0 and a maximum bit rate R N-1 This includes:

[0090] In the example shown in Fig. 7, the number of divisions N is 6, and the lower limit of the bandwidth (bit rate) R L =0.5 (Mbps), upper limit of bandwidth R U = 20 (Mbps), upper limit of video quality Q t This shows an example where the upper limit of video quality Q is set to 93. t = 93, the bit rate R max = 14Mbps, so the maximum bit rate R N-1 =min(R U ,R max )=R max = 14 (Mbps). The minimum video quality Qmin is the lower limit of the bandwidth R L= 0.5 (Mbps). The maximum video quality Qmax is max(Q t ,f(R N-1 ))=Q t =93. The bit rate ladder generation unit 210 determines the remaining four bit rates so that the intervals between the minimum video quality Qmin=20 and the maximum video quality Qmax=93 are equal. The bit rate ladder generation unit 210 determines the four bit rates R i Determine (i=1,2,3,4).

[0091]

number

[0092]

number

[0093] The method for generating a bit rate ladder in the bit rate ladder generator 210 has been described above.

[0094] Returning to FIG. 6, the video encoding unit 220 selects a plurality of bit rates R included in the bit rate ladder. i (i=0,...,N-1). The video encoding unit 220 generates a plurality of encoded videos. The video encoding unit 220 outputs the encoded videos as an encoded stream. The video encoding unit 220 may perform encoding using, for example, VVC, or may perform encoding using another encoding method. The video encoding unit 220 outputs the encoded stream to the storage unit 230.

[0095] The storage unit 230 includes a memory. The memory may be a non-volatile memory. The storage unit 230 may be a storage device. The storage unit 230 may be fixed to the encoded stream generating device 200, or may be portable and moveable to another device. The storage unit 230 stores the encoded stream. Furthermore, the storage unit 230 is capable of outputting the encoded stream to an external device. The encoded stream may be output to a video distribution system. The video distribution system may be a distribution system using the Internet, or may be installed in a broadcasting station using wireless radio waves.

[0096] (Operation example according to the second embodiment) Next, an example of operation according to the second embodiment will be described.

[0097] 8 is a diagram illustrating an example of operation according to the second embodiment. Before the operation shown in FIG. 8 is started, the bit rate ladder generator 210 determines the lower limit R of the bandwidth (bit rate) available for video distribution. L and upper limit R U and the upper limit of video quality Q t may be set (or determined) in advance.

[0098] As shown in FIG. 8, in step S50, the coded stream generating device 200 starts the process.

[0099] In step S51, a rate-quality curve is estimated by the image quality estimation device 100. The rate-quality curve is input to the coded stream generation device 200. Thereafter, the bit rate ladder generation unit 210 performs processing using the rate-quality curve.

[0100] In step S52, the bit rate ladder generation unit 210 calculates the minimum bit rate R0 by dividing the minimum bit rate R0 by the lower limit R of the bandwidth (bit rate) available for video distribution. L (R0=R L), the minimum value Qmin of the video quality corresponding to the minimum bit rate R0 is calculated from the Rate-Quality curve.

[0101] In step S53, the bit rate ladder generator 210 uses the Rate-Quality curve to determine the upper limit Q of the video quality. t The bit rate Rmax corresponding to

[0102] In step S54, the bit rate ladder generator 210 calculates the upper limit R U and bit rate Rmax, the smaller of which is the maximum bit rate R N-1 Set to (R N-1 =min(R U ,Rmax)).

[0103] In step S55, the bit rate ladder generator 210 uses the Rate-Quality curve to determine the maximum bit rate R N-1 Specifically, the bit rate ladder generator 210 calculates the maximum video quality Qmax corresponding to the upper limit of video quality Q t and the maximum bit rate R N-1 and the corresponding video quality, Qmax=max(Q t ,f(R N-1 )) to calculate the maximum value Qmax of the video quality.

[0104] Note that steps S53 to S55 may be performed before step S52.

[0105] In step S56, the bit rate ladder generator 210 calculates the video quality Q i (i=1,...,N-2) is calculated. Equation (10) divides the video quality into equal intervals between the minimum value Qmin and the maximum value Qmax of the video quality, and represents each of the divided video qualities.

[0106] In step S57, the bit rate ladder generator 210 calculates the bit rate Ri (i=0,...,N-1) is calculated. Equation (9) is used to calculate each video quality Q divided by equation (10) in the Rate-Quality curve. i The bit rate ladder generator 210 generates a bit rate ladder including the calculated bit rates. The bit rate ladder generator 210 outputs the generated bit rate ladder to the video encoder 220.

[0107] In step S58, the video encoding unit 220 selects each bit rate R included in the bit rate ladder. i (i=0,...,N-1) The video encoding unit 220 encodes the video as an encoded stream.

[0108] In step S59, the storage unit 230 stores the coded stream.

[0109] Then, in step S60, the coded stream generating device 200 ends the series of processes.

[0110] As described above, the coded stream generating device (coded stream generating device 200) according to the second embodiment has a bit rate ladder generating unit (e.g., bit rate ladder generating unit 210), a video coding unit (e.g., video coding unit 220), and a storage unit (e.g., storage unit 230).

[0111] The bitrate ladder generator generates information showing the relationship between bitrate and video quality (e.g., Rate-Quality curve) and the upper limit of the bitrate available for video distribution (e.g., R U ) and the lower limit (e.g., R L ) and the upper limit of the video quality (e.g., Q t ) and the video quality is divided into multiple bit rates (e.g., R i Generate a bitrate ladder containing (i=0,...,N-1).

[0112] Furthermore, the video encoding unit 220 encodes the video at a plurality of bit rates included in the bit rate ladder to generate a plurality of encoded streams.

[0113] The storage unit 230 stores a plurality of coded streams.

[0114] In this way, for example, in the coded stream generating device 200, the upper limit value Q of the video quality t By appropriately setting, the bit rate ladder generating unit 210 can generate a bit rate ladder including bit rates in the ascending region of the Rate-Quality curve. Therefore, the coded stream generating device 200 can generate an appropriate bit rate ladder.

[0115] Furthermore, the multiple bit rates included in the bit rate ladder are bit rates corresponding to video qualities divided at equal intervals. Therefore, the video qualities corresponding to adjacent bit rates are not too far apart or too close, allowing the user to recognize differences in video quality, which contributes to the user's quality of experience (QoE). Therefore, the coded stream generating device 200 can generate an appropriate bit rate ladder.

[0116] Furthermore, since the video encoding unit 220 can encode video using an appropriately generated bit rate ladder, it is possible to confirm the relationship between the bit rate and video quality using the minimum amount of encoded video required, and ultimately determine the bit rates included in the bit rate ladder. Thus, the encoded stream generating device 200 can generate a bit rate ladder with a certain amount of calculation or less.

[0117] (Another operation example 1 according to the second embodiment) Next, another operation example 1 according to the second embodiment will be described.

[0118] In the second embodiment, the upper limit R of the bandwidth (bit rate) available for video distribution isU and the lower limit R L In another operation example 1 according to the second embodiment, an example will be described in which the bit rate is determined by using "the distribution of the bandwidth (bit rate) available to the user" instead of "the bandwidth (bit rate) available for video distribution."

[0119] FIG. 9 is a diagram showing an example of bandwidth distribution according to the second embodiment. FIG. 9 shows an example of a normal distribution with an average of 5 Mbps and a standard deviation σ=2 Mbps. In this case, if the highest bitrate value of the bitrate ladder described in the second embodiment is 14 Mbps, the proportion of users who can receive it will be 1% or less of the total. Therefore, the contribution to user quality of experience (QoE) will be below a certain level. In order to improve user quality of experience, the bitrate ladder should be created within a range with a high proportion of users.

[0120] Therefore, in another operation example 1 according to the second embodiment, the bit rates included in the bit rate ladder are determined within the bandwidth available to the user. In the example shown in Fig. 9, the bit rate ladder is determined so as to be within a 2σ range centered on the average value. That is, the bit rate ladder generation unit 210 sets a bit rate that is -2σ from the average value as the minimum bit rate R0, and a bit rate that is +2σ from the average value as the maximum bit rate R1. N-1 Thereafter, as in the second embodiment, the bit rate ladder generator 210 determines the minimum value Qmin of the video quality corresponding to the minimum bit rate R0 and the maximum bit rate R N-1 The bitrates R included in the bitrate ladder are calculated by dividing the video quality into equal intervals between the maximum video quality Qmax for i (i=0,...,N-1).

[0121] The percentage of users included in ±2σ is 95% of the total. By setting the bit rate within this range, it is possible to create an appropriate bit rate ladder, as in the second embodiment, and it is possible to create a bit rate ladder with a certain amount of calculation or less.

[0122] Note that if the minimum value of the bit rate ladder is set higher than a predetermined value, the number of users who cannot view the content will increase. Therefore, the bit rate ladder generating unit 210 may include bit rates outside the range of bandwidth available to users in the bit rate ladder.

[0123] 9 is an example, and the bandwidth (bit rate) available to the user does not necessarily follow a normal distribution. Or, even if it follows a normal distribution, it does not have to be limited to the range from -2σ to +2σ. In another operation example 1 according to the second embodiment, a bit rate ladder including bit rates within the bandwidth available to the user can be generated by utilizing the distribution of the bandwidth available to the user.

[0124] (Another operation example 2 according to the second embodiment) In the second embodiment, an example has been described in which VMAF is used as video quality ("Quality"). Compared to other quality indices, VMAF has a high correlation with subjective image quality. Therefore, VMAF has an almost linear relationship with subjective image quality. Therefore, in the second embodiment, by dividing video quality ("Quality") at equal intervals, it is possible to divide subjective image quality at equal intervals as well.

[0125] On the other hand, there are video quality indices that are nonlinear with subjective image quality. For example, the video quality index SSIM has a nonlinear relationship with the subjective image quality MOS. Therefore, the MOS may be estimated from the SSIM using the following equation (11). In equation (11), a, b, c, and d are constants.

[0126]

number

[0127] As a result, for example, the encoded stream generating device 200 can include bit rates corresponding to MOSs divided at equal intervals in the bit rate ladder, and thus, as in the second embodiment, can create an appropriate bit rate ladder that contributes to the user's quality of experience (QoE).

[0128] Note that equation (11) is an example of an equation for converting a video quality index into MOS, and the video quality interval may be adjusted using an equation or table that can convert a video quality index into MOS (or another subjective image quality index).

[0129] (Another operation example 3 according to the second embodiment) In the first embodiment, an example has been described in which the image quality estimation device 100 estimates a Rate-Quality curve from the encoding result of a single encoding run. For example, if the image quality estimation device 100 can estimate a Rate-Quality curve based on the encoding results of multiple encoding runs, it is expected that the estimation accuracy will be improved. Then, if the coded stream generation device 200 generates a bit rate ladder using a Rate-Quality curve with improved estimation accuracy, it will be possible to generate a high-precision bit rate ladder with accuracy above a certain level.

[0130] Therefore, in another operation example 3 according to the second embodiment, the video encoding unit 220 performs encoding each time a bit rate is determined in the bit rate ladder generation unit 210. The image quality estimation device 100 also estimates a Rate-Quality curve each time an encoding result is obtained, and outputs the estimated curve to the coded stream generation device 200. Then, the bit rate ladder generation unit 210 generates a bit rate ladder based on the Rate-Quality curve.

[0131] 10 is a diagram showing another operation example 3 according to the second embodiment. In FIG. 10, as in the second embodiment, the bit rate ladder generator 210 determines the lower limit R of the bandwidth (bit rate) available for video distribution. L and upper limit R U and the upper limit of video quality Q t may be set (or determined) in advance.

[0132] As shown in FIG. 10, in step S70, the coded stream generating device 200 starts the process.

[0133] In step S71, the bit rate ladder generator 210 calculates the minimum bit rate R0 by dividing the minimum bit rate R by the lower limit R of the bandwidth (bit rate). L Since the bit rate R0 is determined as the minimum bit rate, the video encoding unit 220 performs encoding at the minimum bit rate R0. The encoded stream is transmitted to the image quality estimation apparatus 100 via the storage unit 230.

[0134] In step S72, a Rate-Quality curve is estimated in the image quality estimation device 100. For example, the following processing is performed in the image quality estimation device 100. That is, the coded stream is output to the video coding unit 114 of the image quality estimation device 100, and the video coding unit 114 measures the VMAF and bit rate based on the coded stream and outputs them as coding results to the parameter c derivation unit 115. Thereafter, as in the first embodiment, the parameter c derivation unit 115 derives the parameter c, and the curve estimation unit 120 estimates a Rate-Quality curve. The coded stream generation device 200 receives as input the Rate-Quality curve estimated by the image quality estimation device 100. The Rate-Quality curve is a curve estimated based on the coding result coded at the lowest bit rate R0.

[0135] Thereafter, the bit rate ladder generator 210 uses the Rate-Quality curve to determine the upper limit Q of the video quality. t The bit rate Rmax corresponding to the bandwidth is calculated (step S73), and the upper limit value R U and bit rate Rmax, the smaller of which is the maximum bit rate R N-1 (step S74), and the maximum bit rate R N-1 The bit rate ladder generator 210 calculates the maximum value Qmax of the video quality corresponding to the maximum bit rate R N-1 Therefore, the maximum bit rate R N-1 to the video encoding unit 220, which then outputs the maximum bit rate R N-1 The video is encoded at the maximum bit rate R N-1 The coded stream is output to the image quality estimation device 100 via the storage unit 230.

[0136] In step S77, the coded stream generation device 200 sets the variable i to "1".

[0137] In step S78, the image quality estimation device 100 compares the encoding result at the minimum bit rate R0 with the encoding result at the maximum bit rate R N-1 For example, the curve estimation unit 120 of the image quality estimation device 100 estimates a Rate-Quality curve based on the encoding result of the minimum bit rate R0 and the encoding result of the maximum bit rate R N-1 The image quality estimation device 100 converts the two curves, R0 and R1, into a single curve by using the least squares method. N-1 Two Rate-Quality curves are estimated from the encoding results by the image quality estimation device 100, and the two estimated Rate-Quality curves are converted into a single Rate-Quality curve. The conversion into a single Rate-Quality curve is not limited to the least squares method, and any method may be used. For example, the conversion into a single curve may be performed by averaging the two curves. The coded stream generation device 200 receives the converted Rate-Quality curve estimated by the image quality estimation device 100.

[0138] Then, in step S79, the bit rate ladder generator 210 divides the video quality between the minimum video quality Qmin and the maximum video quality Qmax acquired in step S75 at equal intervals once (=i) times, and calculates the divided video quality Q1. In this case, the intermediate value between Qmin and Qmax becomes Q1.

[0139] In step S80, the bit rate ladder generator 210 uses the Rate-Quality curve to calculate a bit rate R1 corresponding to the video quality Q1. The bit rate ladder generator 210 outputs the bit rate R1 to the video encoder 220.

[0140] In step S81, the video encoding unit 220 encodes the video at a bit rate R1.

[0141] In step S82, the coded stream generating device 200 determines whether the variable i is "N-2". N represents the number of bit rates included in the bit rate ladder. If the variable i is "N-2" (Yes in step S82), the process proceeds to step S84. On the other hand, if the variable i is not "N-2" (No in step S82), the process proceeds to step S83.

[0142] In step S83, the coded stream generation device 200 increments the variable i. Thereafter, the processes from step S78 to step S81 are repeated until the variable i becomes "N-2".

[0143] For example, step S78 is repeated to estimate a Rate-Quality curve from a plurality of encoding results. Specifically, the image quality estimation device 100 estimates a Rate-Quality curve for each added encoding result, and converts the estimated Rate-Quality curves into a single Rate-Quality curve. A known method may be used for the conversion into a single curve.

[0144] For example, when i=N-2, the image quality estimation device 100 estimates N Rate-Quality curves based on N encoding results, and converts the N Rate-Quality curves into one Rate-Quality curve. The bit rate ladder generation unit 210 uses the converted Rate-Quality curve to divide the video quality between the minimum value Qmin and the maximum value Qmax of the video quality into (N-2) equal intervals, and calculates the video quality Q i Then, the bit rate ladder generator 210 calculates the bit rate for each video quality Q using the converted Rate-Quality curve. i The corresponding bit rate R i is acquired (step S80).

[0145] In step S84, the coded stream generating device 200 generates N bit rates R iWhen the result of step S82 is "Yes", the encoded stream encoded in step S81 is stored in the storage unit 230.

[0146] Then, in step S85, the coded stream generating device 200 ends the series of processes.

[0147] (Another operation example 4 according to the second embodiment) The coded stream generating device 200 can also create a bit rate ladder for multi-layer coding.

[0148] 11 is a diagram illustrating an example of multi-layer coding according to another operation example 4. Multi-layer coding refers to, for example, coding multiple videos with different resolutions. For example, low-resolution video is coded in a base layer (BL), and a difference between a predicted image (inter-layer predicted image) from the BL-coded video and high-resolution video is coded in an enhancement layer (EL). In multi-layer coding, inter-layer prediction is performed to suppress the generation of redundant information, enabling efficient coding of videos with multiple resolutions.

[0149] The example shown in FIG. 11 illustrates a two-layer example. That is, the video encoding unit 220 has a BL encoding unit 221 and an EL encoding unit 222. The BL encoding unit 221 generates low-resolution encoded video from the input video. Meanwhile, the EL encoding unit 222 uses the decoded video output from the BL encoding unit 221 as predicted video to encode the difference from the input video and output it as high-resolution encoded video. The low-resolution encoded video output from the BL encoding unit 221 is transmitted as a base layer bitstream, and the high-resolution encoded video output from the EL encoding unit 222 is transmitted as an enhancement layer bitstream, and the two are combined and transmitted to the decoding side as a multi-layer bitstream. The BL+EL decoding unit 410 outputs a high-resolution decoded image using the EL encoded video and the BL encoded video, and the BL decoding unit 420 can output low-resolution decoded video using the BL encoded video.

[0150] In addition, in FIG. 11, an example of one EL encoding unit 222 is shown since there are two layers, but in the case of three or more layers, a plurality of EL encoding units 222 may be provided.

[0151] In such multi-layer coding, the bit rate of the coded video in BL is R L0 , the bit rate of the coded video in EL is (R L1 -R L0 ) Here, in the case where the video encoding unit 220 encodes two videos with different resolutions separately, rather than multi-layer encoding, the total bit rate is R L0 +R L1 On the other hand, when encoding using multi-layer coding, the total bits are R L0 +(R L1 -R L0 )=R L1 This becomes:

[0152] In other words, when encoding is performed at a bit rate included in the bit rate ladder, encoding using multi-layer encoding can reduce the total bits of the encoded video compared to encoding multiple videos with different resolutions separately, and it is possible to reduce the number of bits of the encoded stream stored in storage unit 230.

[0153] Fig. 12(A) shows an example of the bit rate ladder when multiple videos with different resolutions are coded separately, and when coded using multi-layer coding. Fig. 12(A) shows an example of up to six layers (five EL layers + one BL layer), but it can also be applied to multi-layer coding with seven or more layers.

[0154] [Other embodiments] A program may be provided that causes a computer to execute each process performed by the image quality estimation device 100 according to the first embodiment and the coded stream generation device 200 according to the second embodiment. The program may be recorded on a computer-readable medium. The computer-readable medium can be used to install the program on a computer. Here, the computer-readable medium on which the program is recorded may be a non-transitory recording medium. The non-transitory recording medium is not particularly limited, and may be, for example, a recording medium such as a CD-ROM or a DVD-ROM. Furthermore, circuits that execute each process performed by the image quality estimation device 100 according to the first embodiment and the coded stream generation device 200 according to the second embodiment may be integrated, and the device may be configured as a semiconductor integrated circuit (chip set, SoC).

[0155] Although the embodiments have been described in detail above with reference to the drawings, the specific configuration is not limited to the above, and various design changes can be made without departing from the spirit of the invention. Furthermore, the operation examples can be combined within a consistent range.

[0156] (Addendum) The above is summarized as follows, but the supplementary notes do not limit the above-described embodiment.

[0157] (Appendix 1) a parameter derivation unit that derives a constant, a first parameter according to a resolution of a video, and a second parameter according to an encoding result obtained by encoding the video once; a curve estimator that estimates a relationship between a bit rate and a video quality of the video based on the constant, the first parameter, and the second parameter. Image quality estimation device.

[0158] (Appendix 2) the curve estimator estimates a curve indicating a relationship between the bit rate and the video quality; the constant represents the video quality to which the curve asymptotically approaches as the bit rate increases; the first parameter represents the bit rate when the video quality is "0", the second parameter represents a parameter relating to the slope of an ascending region of the curve; The rising region represents a region on the curve where the video quality increases at a rate equal to or greater than a certain level as the bit rate increases. 2. The image quality estimation apparatus according to claim 1.

[0159] (Appendix 3) The curve is represented by a sigmoid function. 3. The image quality estimation device according to claim 1 or 2.

[0160] (Appendix 4) The curve is expressed as follows, where a is the constant, b is the first parameter, and c is the second parameter:

number

[0161] (Appendix 5) The video quality is expressed by an objective image quality index or a subjective image quality index. 5. An image quality estimation device according to any one of Supplementary Note 1 to Supplementary Note 4.

[0162] (Appendix 6) The parameter derivation unit derives the first parameter according to the performance of an encoder that performs the encoding. 6. An image quality estimation device according to any one of Supplementary Note 1 to Supplementary Note 5.

[0163] (Appendix 7) the parameter derivation unit performs the encoding again when the bit rate is a bit rate in a relaxed region on the curve; The relaxation region represents the region on the curve where the curve asymptotically approaches the constant. 7. An image quality estimation device according to any one of Supplementary Note 1 to Supplementary Note 6.

[0164] (Appendix 8) The parameter derivation unit derives the first parameter according to the degree of difficulty of the encoding. 8. An image quality estimation device according to any one of Supplementary Note 1 to Supplementary Note 7.

[0165] (Appendix 9) A computer is caused to function as the image quality estimation device described in Appendix 1. program. [Explanation of symbols]

[0166] 100: Image quality estimation device 110: Parameter derivation unit 111: Parameter a derivation unit 112: Parameter b derivation unit 113: Target bit rate determination unit 114: Video encoding unit 115: Parameter c derivation unit 120: Curve estimation unit

Claims

1. a parameter derivation unit that derives a constant, a first parameter according to a resolution of a video, and a second parameter according to an encoding result obtained by encoding the video once; a curve estimator that estimates a relationship between a bit rate and a video quality of the video based on the constant, the first parameter, and the second parameter. Image quality estimation device.

2. the curve estimator estimates a curve indicating a relationship between the bit rate and the video quality; the constant represents the video quality to which the curve asymptotically approaches as the bit rate increases; the first parameter represents the bit rate when the video quality is "0", the second parameter represents a parameter relating to the slope of an ascending region of the curve; The rising region represents a region on the curve where the video quality increases at a rate equal to or greater than a certain level as the bit rate increases. The image quality estimation device according to claim 1 .

3. The curve is represented by a sigmoid function. The image quality estimation device according to claim 2.

4. The curve is expressed as follows, where a is the constant, b is the first parameter, and c is the second parameter: [Equation 1] where Quality represents the video quality and Rate represents the bit rate. The image quality estimation device according to claim 2.

5. The video quality is expressed by an objective image quality index or a subjective image quality index. The image quality estimation device according to claim 1 .

6. The parameter derivation unit derives the first parameter according to the performance of an encoder that performs the encoding. The image quality estimation device according to claim 1 .

7. the parameter derivation unit performs the encoding again when the bit rate is a bit rate in a relaxed region on the curve; The relaxation region represents the region on the curve where the curve asymptotically approaches the constant. The image quality estimation device according to claim 2.

8. The parameter derivation unit derives the first parameter according to the degree of difficulty of the encoding. The image quality estimation device according to claim 1 .

9. A computer is caused to function as the image quality estimation device according to claim 1. program.

Images