CN106447656B - Rendering flaw image detecting method based on image recognition - Google Patents
Rendering flaw image detecting method based on image recognition Download PDFInfo
- Publication number
- CN106447656B CN106447656B CN201610842103.1A CN201610842103A CN106447656B CN 106447656 B CN106447656 B CN 106447656B CN 201610842103 A CN201610842103 A CN 201610842103A CN 106447656 B CN106447656 B CN 106447656B
- Authority
- CN
- China
- Prior art keywords
- pixel
- matrix
- image
- rendering
- neighborhood
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/005—General purpose rendering architectures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Graphics (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides the rendering flaw image detecting methods based on image recognition, method includes the following steps: inputting render video image to be detected；Calculate the difference matrix of rendering image；Detection template size is determined according to rendering image change rate；It is poor that each pixel neighborhood of a point minimum pixel in each frame image is calculated according to detection template；Bad frame image and flaw present position are judged according to neighborhood minimum pixel difference.In view of the problems of the existing technology, its purpose is to provide a kind of video flaw detection methods by the present invention, are used for quickly detecting to the flaw part in rendering result video image.
Description
Technical field
The present invention designs a kind of rendering image Defect Detection technology, especially designs a kind of for rendering image Defect Detection
Method.
Background technique
Since house ornamentation and video display rendering computing resource and storage resource demands are larger, completed using large-scale cloud computing technique
It is such to be rendered to as mainstream technology.Thus the system of rendering is provided in face of numerous different types of rendering tasks, is loaded in resource
There is certain probability in the process makes part calculate node that can not ensure all resource whole normal loads, thus will cause part
Rendering node rendering is not normal.When rendering result image file summarizes, it is adjacent thereto that those render scenes are screened by artificial detection
The result images that image has differences, and defining such image is abnormal frame image, it is rendered again to correct mistake.Although
This abnormal image belongs to contingencies in rendering result, and general isolated presence will appear continuous in the case where only a few
Two frame Rendering errors images occur, but this kind of mistake causes entire render process heavy dependence manually to post-process misarrangement, greatly
The automatic flow and efficiency of rendering are limited, for this purpose, this patent, which designs a kind of algorithm automatic identification, renders abnormal frame image, with
Rendering process automation level is promoted, rendering efficiency is promoted.
Above rendering its file size of abnormal image, format and normal rendering result have no significant difference, using conventional
It is difficult to screen when information based on file format, the present invention designs a kind of method based on machine vision thus, utilizes sequential chart
The local characteristics of picture variation detect the bad frame image in rendering result automatically, and specific summary of the invention is as follows.
Summary of the invention
In view of the problems of the existing technology, its purpose is to provide a kind of video flaw detection methods, to wash with watercolours by the present invention
Flaw part in dye result video image is used for quickly detecting.
To achieve the above object, the present invention provides a kind of rendering flaw image detecting method based on image recognition, should
Method is the following steps are included: input render video image to be detected；Calculate the difference matrix of rendering image；According to rendering image
Change rate determines detection template size；Each pixel neighborhood of a point minimum pixel in each frame image is calculated according to detection template
Difference；Bad frame image and flaw present position are judged according to neighborhood minimum pixel difference.
As a further improvement of the present invention, the method also includes following steps: to the rendering result image of input into
Row serializing processing, is read in the form of RGB and is arranged to result images, and in chronological order.Obtain four-matrix data M
[X, Y, Z, T], wherein X, Y are video image location of pixels, and Z is tri- color of RGB, and T is time shaft.
As a further improvement of the present invention, the method also includes following steps: according to the variation of rendering result image
Rate determines single pixel neighborhood detection template ST, because of the special nature of rendering image, when rendering contexts are identical, and rendering result
The numerical value of each pixel is relative constant in image.To the adjacent two field pictures to sort in temporal sequence, if prospect or background
Ohject displacement relatively less under the premise of, then can centainly be found in closing on frame image completely corresponding another with current pixel point
One pixel, and this two pixel of correspondence has neighborhood similarity in position.Therefore according to display foreground or background position
The rate of shifting can determine a neighborhood template, so that each pixel can be using this neighborhood template as model in current frame image
It encloses, same pixel is found in a neighborhood for closing on this pixel corresponding position of frame, neighborhood size is the template of ST
Size.In actual application, this template size is generally obtained by artificial experience.
As a further improvement of the present invention, the method also includes following steps: to the rendering result figure of time series
As temporally axis sequence progress difference processing, difference processing are divided into two steps, i.e. matrix displacement and Difference Calculation:
M '=Mt-Mt+1T=[1...T-1]
Matrix displacement is to adjacent two field pictures matrix MtWith Mt+1Middle MtMatrix carries out shifting processing.First according to ST mould
The shape of plate determines member point position in template, except each pixel in member point exterior sheathing determines a MtMovement (Δ x, Δ y)
Operation.MtIt is translated as unit of pixel according to the size of Δ x, Δ y, original matrix MtIn each pixel it is mobile after obtain new picture
Element (x+ Δ x, y+ Δ y).If removing matrix M after pixel is mobilet+1Directly give up the part of range；And if occurring after pixel movement
In matrix Mt+1In range and original pixel is not in matrix MtIn range, then need to carry out initialization assignment to the new pixel, assignment is big
Small is twice single pixel maximum value pmax(pmax=max (m) m ∈ Mt)。
Matrix M after shifting is calculated after shifting every time according to single pixel neighborhood detection templatei tWith Mt+1Matrix difference matrix Δ Mi t
Once (i ∈ { ST-0 }), matrix by RGB dimension seeks 2 norm of vector after all differences, obtains differential mode matrix | Δ Mi t|.Finally to institute
There are in differential mode matrix each corresponding pixel points minimize matrix Δ Mmin t(ΔMmin t=mini{|ΔMi t|,i∈{ST-0}})。
As a further improvement of the present invention, the method also includes following steps: to matrix Δ Mmin tIt is post-processed,
Because there are identical pixel, matrix Δ M in neighborhood in consecutive frame for most of pixelmin tTheoretically big portion
Subregion is 0.But due to the influence in image there are noise, Δ Mmin tIn will appear individual non-zero isolated points, use Mathematical Morphology
Method can be to Δ Mmin tIn isolate abnormal point denoising.Concrete operations are to determine the template of Mathematical Morphology Method first
SE (disc template is used, template size is | XY |/20000), then to Δ Mmin tCarry out closing operation of mathematical morphology, and by result into
Row morphology opening operation obtains correction matrix sequence Δ M 'min t。
As a further improvement of the present invention, the method also includes following steps: to differential mode correction matrix sequence Δ
M’min tDiscriminant analysis.Since based on similar principle in neighborhood of pixels in rendering result image, Δ M ' in matrix sequencemin tGreatly
Part is 0 matrix, therefore to matrix sequence Δ M 'min tIn each matrix carry out non-zero pixels statistics, when wherein non-zero pixel number
It then can be determined that there are bad frames in the corresponding rendering result of current time t when surge, and check that this rendering result image is divided
It distinguishes.
The utility model has the advantages that
Context of methods is significant in the industrial production, can realize that the bad frame of rendering image detects function automatically using this method
Can, detect the abnormal results in rendering automatically under conditions of being not necessarily to manual intervention, and be marked.It is rendered in extensive outdoor scene
In task, automatic bad frame detection method can assist screening bad frame rendering image, improve the quality of rendering result, reduce identification bad frame
Cost of labor.
Detailed description of the invention
Rendering image Defect Detection flow chart Fig. 1 of the invention.
Rendering bad frame instance graph Fig. 2 of the invention.
Difference matrix non-zero region figure Fig. 3 of the invention.
Difference matrix Fig. 4 of the invention calculates schematic diagram.
Bad frame detection curve figure Fig. 5 of the invention.
Bad frame region labeling figure Fig. 6 of the invention.
Specific embodiment
Hereinafter, the present invention will be described in detail with reference to various embodiments shown in the accompanying drawings.
It show preferable system flow chart of the present invention refering to fig. 1, be divided into input picture and carries out serializing processing, determine list
Neighborhood of pixels template carries out displacement and Difference Calculation to sequence image matrix, is post-processed to difference modular matrix, statistic mixed-state
Five steps such as bad frame image.
It is normal rendering result picture refering to left figure shown in Fig. 2, right figure is rendering abnormal results picture.Because material loads
Problem, it is evident that the rendering of lower right corner meadow part is abnormal in right figure, there are notable differences with left figure corresponding region.This two frame
Image subtraction, there are large area non-zero regions on corresponding position.In other words this adjacent two field pictures is on this position
Each pixel can not find corresponding same pixel in template ST neighborhood.The difference result image of this two field pictures is as schemed
3, the wherein visible continuous non-zero region of flaw location, and other scatterplots be two field pictures prospect and background it is mobile caused by difference
Dissimilarity.Since prospect and background movement speed are unhappy, so such discrepancy is isolated zonule, monomer area is little.
The difference matrix for seeking t frame is shown refering to Fig. 4.As it was noted above,It is in t frame image MtOn the basis of
The set of several obtained subgraphs is translated, the number of subgraph is related with selected template size, and the bigger needs of template translate
Number is more, and subgraph is more.As shown in Fig. 4 (1),In MtOn the basis of horizontal direction be shifted x pixel of Δ, Vertical Square
To y pixel of Δ is shifted, (Δ x, Δ y) need to be traversed for whole pixels on template ST except 0 point.One is had in translation motion
Partial region removes Mt, while also can the immigration of some regionFor moving intoThis partial region need at the beginning of it
Beginningization assignment, that is, image grey area, assignment size are twice single pixel maximum value pmax(pmax=max (m) m ∈
Mt)。
Several image collections obtained after translationBy with latter image Mt+1Phase
Subtract, obtains difference matrix
The number containing abnormal frame in this section of consecutive image sought using the present invention is shown refering to Fig. 5.Such as Fig. 3 institute
Show, scatterplot region caused by continuous non-zero region and prospect as caused by bad frame, background movement is marked in image.Use form
Method first carries out closed operation to image, then carries out morphology opening operation to result, it can removal scatterplot region is protected simultaneously
Stay continuum caused by bad frame.It can be drawn by the elemental area of the remaining regions in image after computation of morphology operations
Such as the histogram of Fig. 5.
It show using the present invention refering to Fig. 6 in abnormal frame shown in Fig. 2, the accurate location for the unusual part sought,
That is lower right corner black region in Fig. 6.It can be told also by carrying out interpretation or threshold value screening to remaining regions in previous step
Bad frame as shown in FIG. 6 can be obtained by the way that non-zero region continuous in this number image is marked in the picture number of bad frame
Area image.
Certainly the above embodiments merely illustrate the technical concept and features of the present invention, and its object is to allow be familiar with technique
People can understand the content of the present invention and implement it accordingly, it is not intended to limit the scope of the present invention.It is all according to this hair
The modification that the Spirit Essence of bright main technical schemes is done, should be covered by the protection scope of the present invention.
Claims (5)
1. the rendering flaw image detecting method based on image recognition, which is characterized in that method includes the following steps:
Step 1: inputting render video image to be detected；
Step 2: calculating the difference matrix of rendering image；
Step 3: determining detection template size according to rendering result image change rate；
Step 4: it is poor to calculate each pixel neighborhood of a point minimum pixel in each frame image according to detection template；
Step 5: judging bad frame image and flaw present position according to neighborhood minimum pixel difference；
Single pixel neighborhood detection template ST is determined according to rendering result image change rate in step 3, because of the spy of rendering image
Different property, when rendering contexts are identical, the numerical value of each pixel is relative constant in rendering result image；To arranging in temporal sequence
The adjacent two field pictures of sequence, if prospect or background object displacement relatively less under the premise of, found in closing on frame image
One other pixel point completely corresponding with current pixel point, and this two pixel of correspondence has neighborhood similarity in position；
Therefore a neighborhood template determined according to the rate of display foreground or background displacement so that in current frame image each pixel with
This neighborhood template is range, finds same pixel, neighborhood in a neighborhood for closing on this pixel corresponding position of frame
Size is the template size of ST；
To the rendering result image of time series, temporally axis sequence carries out difference processing, and difference processing is divided into two steps, i.e.,
Matrix displacement and Difference Calculation:
M '=Mt-Mt+1T=[1...T-1]
Matrix displacement is to adjacent two field pictures matrix MtWith Mt+1Middle MtMatrix carries out shifting processing: first according to ST template
Shape determines member point position in template, except each pixel in member point exterior sheathing determines a MtMovement (Δ x, Δ y) behaviour
Make；MtIt is translated as unit of pixel according to the size of Δ x, Δ y, original matrix MtIn each pixel it is mobile after obtain new pixel
(x+ Δ x, y+ Δ y)；If removing matrix M after pixel is mobilet+1Directly give up the part of range；And if being appeared in after pixel movement
Matrix Mt+1In range and original pixel is not in matrix MtIn range, then need to carry out the new pixel initialization assignment, assignment size
For twice single pixel maximum value pmax(pmax=max (m) m ∈ Mt)；
Matrix M after shifting is calculated after shifting every time according to single pixel neighborhood detection templatei tWith Mt+1Matrix difference matrix Δ Mi tOnce (i
∈ { ST-0 }), matrix by RGB dimension seeks 2 norm of vector after all differences, obtain differential mode matrix | Δ Mi t|；Finally to all differential modes
Each corresponding pixel points are minimized matrix Δ M in matrixmin t(ΔMmin t=mini{|ΔMi t|,i∈{ST-0}})。
2. rendering flaw image detecting method according to claim 1, it is characterised in that: to matrix Δ Mmin tProgress after
Reason, because there are identical pixel, matrix Δ M in neighborhood in consecutive frame for most of pixelmin tTheoretically
Most of region is 0；But due to the influence in image there are noise, Δ Mmin tIn will appear individual non-zero isolated points, use mathematics
Morphological method can be to Δ Mmin tIn isolate abnormal point denoising.
3. rendering flaw image detecting method according to claim 2, which is characterized in that the abnormal point denoising tool
Gymnastics as: first determine Mathematical Morphology Method template SE, then to Δ Mmin tClosing operation of mathematical morphology is carried out, and by result
Morphology opening operation is carried out, differential mode correction matrix sequence Δ M ' is obtainedmin t。
4. rendering flaw image detecting method according to claim 3, it is characterised in that: the template SE uses disk mould
Plate, template size be | XY |/20000；Wherein X, Y are video image location of pixels.
5. rendering flaw image detecting method according to claim 4, which is characterized in that differential mode correction matrix sequence
ΔM’min tDiscriminant analysis: because based on similar principle in neighborhood of pixels in rendering result image, Δ M ' in matrix sequencemin t
Most of is 0 matrix, therefore to matrix sequence
ΔM’min tIn each matrix carry out non-zero pixels statistics, then determine t pairs of current time when wherein non-zero pixel number increases sharply
There are bad frames in the rendering result answered, and check that this rendering result image is differentiated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610842103.1A CN106447656B (en) | 2016-09-22 | 2016-09-22 | Rendering flaw image detecting method based on image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610842103.1A CN106447656B (en) | 2016-09-22 | 2016-09-22 | Rendering flaw image detecting method based on image recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106447656A CN106447656A (en) | 2017-02-22 |
| CN106447656B true CN106447656B (en) | 2019-02-15 |
Family
ID=58166855
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610842103.1A Active CN106447656B (en) | 2016-09-22 | 2016-09-22 | Rendering flaw image detecting method based on image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106447656B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107197325A (en) * | 2017-06-13 | 2017-09-22 | 深圳市瑞云科技有限公司 | A kind of bad frame automatic testing method |
| DE102018220236A1 (en) * | 2018-11-26 | 2020-05-28 | Heidelberger Druckmaschinen Ag | Fast image rectification for image inspection |
| CN111160160B (en) * | 2019-12-18 | 2022-08-05 | 河海大学 | Neighborhood box-based block mass acquisition computing method |
| CN115661145B (en) * | 2022-12-23 | 2023-03-21 | 海马云（天津）信息技术有限公司 | Cloud application bad frame detection method and device, electronic equipment and storage medium |
| CN115941914B (en) * | 2023-01-06 | 2023-05-23 | 湖南马栏山视频先进技术研究院有限公司 | Video rendering system based on video frame analysis |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101604458A (en) * | 2008-06-11 | 2009-12-16 | 美国西门子医疗解决公司 | The method that is used for the computer aided diagnosis results of display of pre-rendered |
| CN102903120A (en) * | 2012-07-19 | 2013-01-30 | 中国人民解放军国防科学技术大学 | Time-space condition information based moving object detection method |
| CN103945228A (en) * | 2014-03-28 | 2014-07-23 | 上海交通大学 | Video intra-frame copy-move tampering detection method based on space and time relevance |
| CN104539942A (en) * | 2014-12-26 | 2015-04-22 | 赞奇科技发展有限公司 | Video shot switching detection method and device based on frame difference cluster |
-
2016
- 2016-09-22 CN CN201610842103.1A patent/CN106447656B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101604458A (en) * | 2008-06-11 | 2009-12-16 | 美国西门子医疗解决公司 | The method that is used for the computer aided diagnosis results of display of pre-rendered |
| CN102903120A (en) * | 2012-07-19 | 2013-01-30 | 中国人民解放军国防科学技术大学 | Time-space condition information based moving object detection method |
| CN103945228A (en) * | 2014-03-28 | 2014-07-23 | 上海交通大学 | Video intra-frame copy-move tampering detection method based on space and time relevance |
| CN104539942A (en) * | 2014-12-26 | 2015-04-22 | 赞奇科技发展有限公司 | Video shot switching detection method and device based on frame difference cluster |
Non-Patent Citations (2)
| Title |
|---|
| Photo Defect Detection for Image Inpainting;Rong-Chi Chang等;《Proceedings of the Seventh IEEE International Symposium on Multimedia(ISM"05)》;20051214;403-407 * |
| 用于视频镜头突变切换检测的二次差分法和像素点匹配二次差分法;刘谦雷 等;《中国图象图形学报》;20030228;第8(A)卷(第2期);161-168 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106447656A (en) | 2017-02-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106447656B (en) | Rendering flaw image detecting method based on image recognition | |
| CN107833220B (en) | Fabric defect detection method based on deep convolutional neural network and visual saliency | |
| CN105184763B (en) | Image processing method and device | |
| US9042662B2 (en) | Method and system for segmenting an image | |
| US8605795B2 (en) | Video editing methods and systems | |
| CN111126115B (en) | Violent sorting behavior identification method and device | |
| AU2012258421A1 (en) | Superpixel-based refinement of low-resolution foreground segmentation | |
| CN101777180A (en) | Complex background real-time alternating method based on background modeling and energy minimization | |
| CN102194443A (en) | Display method and system for window of video picture in picture and video processing equipment | |
| CN102202224A (en) | Caption flutter-free method and apparatus used for plane video stereo transition | |
| CN101510304B (en) | Method, device and pick-up head for dividing and obtaining foreground image | |
| CN113449606B (en) | Target object identification method and device, computer equipment and storage medium | |
| CN112364865B (en) | Method for detecting small moving target in complex scene | |
| Sakpal et al. | Adaptive background subtraction in images | |
| CN111383244A (en) | Target detection tracking method | |
| CN114419006A (en) | Method and system for removing watermark of gray level video characters changing along with background | |
| Lu et al. | Real-time video stylization using object flows | |
| CN106056575B (en) | A kind of image matching method based on like physical property proposed algorithm | |
| CN115761580A (en) | Detection method of water surface floater, edge calculation terminal and storage medium | |
| Kim et al. | Robust foreground segmentation from color video sequences using background subtraction with multiple thresholds | |
| CN110580706A (en) | Method and device for extracting video background model | |
| CN111160255B (en) | Fishing behavior identification method and system based on three-dimensional convolution network | |
| CN113936030A (en) | Moving target detection method and system based on convolutional coding | |
| CN113627342A (en) | Method, system, device and storage medium for video depth feature extraction optimization | |
| CN104732558B (en) | moving object detection device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20191113 Address after: 200030 room 249, zone a, 2f, no.420 Fenglin Road, Xuhui District, Shanghai Patentee after: Shanghai Zanqi Culture Technology Co., Ltd Address before: 23, building 9-2, 213022 East Taihu Road, Jiangsu, Changzhou Patentee before: JIANGSU CUDATEC CO., LTD. |