CN103647963A - Video quality evaluation method based on Gop scene complexity - Google Patents

Abstract

Video objective quality evaluation is one of important research points in QoE services in the future. The invention proposes a video quality evaluation method based on GoP scene complexity. Firstly, through analysis of a video sequence, separated frames and GoPs of a video are obtained. Through use of the obtained separated frames and GoPs, scene complexity and time-domain complexity of a Gop are obtained. According to the obtained scene complexity and time-domain complexity, an original quality and visual masking effect of the GoP are calculated respectively. According to the error masking effect, the objective quality of a single GoP is obtained. At last, a proposed weighting formula is used to perform weighting on a plurality of GoPs in the video so that the objective quality of the video sequence is obtained.

Description

Method for evaluating video quality based on GoP scene complexity

Technical field

The present invention relates to video quality to carry out the method for objective evaluation, particularly a kind of calculating based on GoP scene complexity and GoP scene complexity are to the method for the mapping of video quality.

Technical background

Along with digital video service is dissolved into people's life more and more in every way, video source is more and more important to the video quality assessment of user side.The most reliable way of video quality evaluation is by subjective assessment, and people removes to observe truly video, by watching, video is provided to a subjective evaluation of estimate.In VQEG, just stipulated a kind of single continuous mass (SSCQE) evaluation method that stimulates, observer provides a subjective scores after watching one section of continuous videos, and the finally evaluation by comprehensive multidigit observer draws subjective assessment value.Yet in a lot of occasions, this evaluation method has serious limitation, its needs a large amount of personnel to participate in, and wastes time and energy, and is also difficult to integratedly be applied to all multipair QoE and have in the system of demand simultaneously.

For this reason, people's Recent study the multiple method for objectively evaluating for video quality.According to different to the demand of reference video in algorithm evaluation procedure, method for objectively evaluating can be divided into following three classes:

Entirely with reference to evaluation method.The method needs the full detail of reference video in evaluation procedure, and reference video is all without compressing conventionally.Algorithm, by the difference between contrast reference video and test video, quantizes the damaged or quality of test video, thereby draws the subjective assessment value about test video.Must, full reference video should be the most accurately, because it can or must be about the whole information of reference video.Meanwhile, desirable full reference method should damaged what cause be really to have good robustness to various.

Half with reference to evaluation method.The method is applied to obtain about reference video part information, or only need to obtain the occasion of a part of information, and video quality is made to evaluation.It extracts some feature between reference video and test video conventionally, by the contact between contrast characteristic, realizes the evaluation to video quality.

Nothing is with reference to evaluation method.When evaluating, the method without any need for the information about reference video, therefore there is maximum flexibility ratio and applicable scene.Because cannot obtain the relevant information of reference video, accurate evaluation video has suitable challenge.A large amount of scholars is studied this problem from each different angles in recent years.

This problem proposes a kind of by calculating the no-reference video quality evaluating method of GoP scene complexity.The method has defined a main body characteristic based on video sequence, is called scene complexity.The method thinks, scene complexity is the leading factor that affects video quality.Any and structure of video sequence, the disappearance that content is relevant can reflect by scene complexity.By calculating the scene complexity of test video, change, can make effective evaluation to video quality.

Summary of the invention

To achieve these goals, solve corresponding technical problem, the present invention realizes by following scheme:

Whole methods and results is as shown in Figure 1:

Step 1: analytical test video sequence, isolate single frames and frame type.

Step 2: according to the frame type drawing in step 1, video sequence is separated into single GoP sequence.

Step 3: according to the GoP sequence obtaining in step 2, extract respectively the texture maps of IBP frame, calculate the scene complexity of single GoP by formula 1.

Step 4: according to the GoP sequence that must fall in step 2, calculate the time domain complexity of single GoP according to formula 2.

Step 5: according to the single GoP time domain complexity obtaining in step 4, calculate the visual error shielding effect in single GoP.

Step 6: according to the GoP visual error shielding effect in the scene complexity obtaining in step 3 and step 5, calculate the quality of single GoP.

Step 7: according to the single GoP in step 2, the duration of calculating single GoP.

Step 8: according to the duration of this GoP in the single GoP quality obtaining in step 6 and step 7, draw the objective quality of whole video sequence according to formula 3, see accompanying drawing 2.

Accompanying drawing explanation

Accompanying drawing can provide a further understanding of the present invention, and is comprised in the specification part as content, and it shows embodiments of the invention, and comes together to explain principle of the present invention with specification.Wherein,

Fig. 1 shows the framework of whole method.

Fig. 2 shows the hierarchical mode of system.

Embodiment

First with reference to accompanying drawing, the preferred embodiments of the present invention are described, as much as possible, identical identical or similar label or textual representation for part in whole accompanying drawing.

Fig. 1 shows method frame of the present invention, specifically comprises:

Step 1: analytical test video sequence, isolate single frames and frame type.For a video sequence, at decoder, read video packets, when decoding, just can learn what frame that belongs to when pre-treatment.Conventionally, a video sequence has some I frames, and P frame and B frame form.I frame is called reference frame, and it includes the complete information of original image, therefore often also maximum; P frame is called single directional prediction frame, and the difference by coding present image and previous I frame or P frame obtains; B frame is bi-directional predicted frames, carries out Two-way record down by former and later two P frames.

Step 2: according to the frame type drawing in step 1, video sequence is separated into single GoP sequence.A common video sequence adds some predictive frame by a leading I frame conventionally, consists of, as IPPP time sequencing ...Conventionally, the part between two I frames, is called a GoP.At encoder during to Video coding, the switching of scene is meaned in the insertion of I frame conventionally, or predictive frame is more than enough, and present image and first I frame difference are enough large.Therefore, a kind ofly reasonably think, a GoP can be regarded as a scene.We are for the calculating of scene complexity, also based on GoP.

Step 3: according to the GoP sequence obtaining in step 2, extract respectively the texture maps of IBP frame, pass through formula

$\mathrm{SC}=\frac{\underset{i=1}{\overset{N}{\∪}}{\mathrm{map}}_i-\underset{i=1}{\overset{N}{\∩}}{\mathrm{map}}_j}{m\×n}$

Map _ithe texture maps that represents i frame, m, n represents frame resolution

Calculate the scene complexity of single GoP.Texture maps is by extracting a width bianry image of image edge information.Conventionally, extract canny operator, sobel operator, log operator etc.Adopt canny operator herein:

1. use Gaussian filter smoothed image

$h(x,y,\mathrm{\σ})=\frac{1}{{2\mathrm{\π\σ}}^2}{e}^{\frac{x^2+y^2}{-{2\mathrm{\σ}}^2}}$

g(x,y)＝h(x,y,σ)*f(x,y)

Wherein to represent that Gaussian filter impacts corresponding for the first formula, and the second formula represents image after filtering.

2. by single order local derviation finite difference compute gradient amplitude direction

$M[x,y]=\sqrt{G_x{(x,y)}^2+G_y{(x,y)}^2}$

θ[x,y]＝arctan(G _x(x,y)/G _y(x,y))

M[x, y] reacted image border intensity, θ [x, y] has reacted the direction at edge.G _x(x, y) and G _y(x, y) is both direction derivative.

3. pair gradient magnitude carries out non-maximum transplanting

The Grad obtaining in 2 is not enough to determine edge, therefore, for determining edge, must retain the point of partial gradient maximum, and suppress non-maximum.

4. with dual threshold algorithm, detect and be connected edge

Utilize formula

Q _s=m·e ^n·sc

Calculate GoP quality.M and n value can obtain by regression test.

Step 4: according to the GoP sequence that must fall in step 2, according to formula

$P=\underset{m,n}{\overset{\mathrm{all}}{\mathrm{\Σ}}}{[I(t,m,n)-I(t-1,m,n)]}^2$

I (t, m, n) represents the m of t frame, the motion vector of n piece

Calculate the time domain complexity of single GoP.Time domain complexity can be thought the variation severe degree of video content along with the time.By the result to the poor summation of frame, can regard the degree that a video content changes in a GoP as.

Step 5: according to the single GoP time domain complexity obtaining in step 4, calculate the visual error shielding effect in single GoP.Error concealment effect should be directly proportional with time domain complexity.Time domain complexity is higher, and during appearance wrong, human eye is more not easy to discover.

Q _e＝a*P

Wherein, parameter a can obtain by regression test.

Step 6: according to the GoP visual error shielding effect in the scene complexity obtaining in step 3 and step 5, calculate the quality of single GoP.Single GoP quality should add the lifting of the quality of error concealment for original GoP quality,

Q _gop＝Q _s+Q _e

Step 7: according to the single GoP in step 2, the duration of calculating single GoP.

Step 8: according to the duration of this GoP in the single GoP quality obtaining in step 6 and step 7, utilize formula below

$\mathrm{VQE}=\frac{\underset{n\∈S_G}{\mathrm{\Σ}}(Q_{\mathrm{gop}}\·T_n)}{\underset{n\∈S_G}{\mathrm{\Σ}}{T}_n}$

Comprehensively draw the objective video quality of whole video sequence.

Claims (5)

1. the method for evaluating video quality based on GoP scene complexity, is characterized in that, at least comprises the steps:

Step 1: calculate the scene complexity of single GoP,

Step 2: calculate the time domain complexity of single GoP,

Step 3: calculate the quality of single GoP,

Step 4: the objective video quality of calculating whole video sequence.

2. the method for evaluating video quality based on GoP scene complexity according to claim 1, is characterized in that:

In step 1, calculate the scene complexity of single GoP:

The 1st step: analyze video sequence, draw isolated frame, single GoP,

The 2nd step: according to formula

, calculate the scene complexity of single GoP.

3. the method for evaluating video quality based on GoP scene complexity according to claim 1, is characterized in that:

In step 2, calculate the time domain complexity of single GoP:

The 1st step: the isolated frame according to obtaining in step 11, obtain encoding block,

The 2nd step: according to the encoding block obtaining in step 21, the interframe movement vector of computing block,

The 3rd step: utilize the interframe movement vector obtaining in step 22, according to formula

Calculate time domain complexity.

4. the method for evaluating video quality based on GoP scene complexity according to claim 1, is characterized in that:

In step 3, calculate the quality of single GoP:

The 1st step: utilize GoP scene complexity, utilize formula

Q _s=m·e ^n·sc

Calculate the quality of GoP,

The 2nd step: utilize GoP time domain complexity, utilize formula

Q _e＝a*P

Calculate the visual masking effect of GoP,

The 3rd step: according to visual masking effect, the GoP quality that step 42 is drawn adds visual masking effect, utilizes formula

Q _gop＝Q _s+Q _e

Draw the quality of GoP.

5. the method for evaluating video quality based on GoP scene complexity according to claim 1, is characterized in that:

In step 4, calculate the objective video quality of video sequence:

According to formula

, calculate the quality of whole video sequence.

Images