CN105612558B - Device and method for handling image - Google Patents
Device and method for handling image Download PDFInfo
- Publication number
- CN105612558B CN105612558B CN201480055150.0A CN201480055150A CN105612558B CN 105612558 B CN105612558 B CN 105612558B CN 201480055150 A CN201480055150 A CN 201480055150A CN 105612558 B CN105612558 B CN 105612558B
- Authority
- CN
- China
- Prior art keywords
- light
- tree
- sub
- division
- classification
- 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.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/06—Ray-tracing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/005—Tree description, e.g. octree, quadtree
-
- 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
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/50—Lighting effects
- G06T15/80—Shading
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Graphics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
A kind of image processing apparatus, it include: light generator, it is configurable to generate at least one light, multiple tree searchers, it is configured as executing the tree search of at least one light generated, area divider, at least one light generated is divided into multiple sub-districts by the information for being configured as the quantity based on the tree search about previous image frames, and light distributor, it is configured as the multiple sub-districts divided by area divider multiple tree searchers are separately assigned.
Description
Technical field
It is related to image processing apparatus and its image processing method with the consistent device and method of exemplary embodiment, and more
In particular it relates to the image processing apparatus that can render (render) image light (ray) method for tracing.
Background technique
In the related art, figure rendering is to be handled by graphics processing unit (GPU), but really render and 3 dimension wash with watercolours
Dye is impossible.Recently, the true rendering method of such as light projection, ray tracing and global illumination is developed, and is made
For more true and 3 dimensions Rendering algorithms.
Image processing apparatus 100 in the related technology includes light generator 110, tree searcher 120, intersects
(intersection) tester 130 and tinter 140, as shown in fig. 1.The service load for setting searcher 120 is at image
The maximum service load on device 100 is managed, and sets the performance of searcher 120 to the overall performance of image processing apparatus 100
It influences maximum.Therefore, in order to enhance performance, image processing apparatus 100 includes multiple tree searcher 120-1,120-2,120-3 ...
As shown in fig. 1.
Multiple tree searcher 120-1,120-2,120-3 ... are assigned to by each light that light generator 110 generates
One of.The tree searcher that light is assigned influences the cache hit rate (hit ratio) of tree searcher 120.
Because cache hit rate greatly influences to set the performance of searcher 120, efficiently divide for tree searcher 120
Method with light is crucial.
The method of the relevant technologies for distributing light for multiple tree searcher 120-1,120-2,120-3 ... is illustrated in
In Fig. 2A and Fig. 2 B.More specifically, as shown in Figure 2 A, the multiple light generated by light generator 110 are with first in, first out
(FIFO) method is assigned to multiple tree searcher 120-1,120-2,120-3 ....That is, by identifying multiple tree search
The input fifo status of device 120-1,120-2,120-3 ..., empty tree searcher is searched, so that light is sequentially distributed
To the tree searcher of the sky.
Alternatively, by multiple light that light generator 110 generates be divided into multiple tree searcher 120-1,
Corresponding group of the quantity of 120-2,120-3 ..., and every group of light is assigned to corresponding tree searcher, such as institute in Fig. 2 B
Show.
However, these methods have following problems.For example, in the side as shown in Figure 2 A for distributing light in FIFO method
In method, the correlation between light handled by tree searcher is low.Therefore, cache memory is imitated without temporal locality
(temporal locality effect) is answered, and therefore cache hit rate is lower.Therefore, image procossing fills
The performance set may deteriorate.
In addition, the method that light is distributed to group as shown in Figure 2 B is generated higher due to temporal locality effect
Cache hit rate.However, load balancing is asked when needing the light of a large amount of tree searchers to be assigned to a certain group
Topic is likely to occur.Therefore, tree searcher 120 may not be able to be operated efficiently.
Summary of the invention
Technical problem
Exemplary embodiment can overcome disadvantage mentioned above or disadvantage as not described above.Also, exemplary embodiment
It is not required to overcome disadvantages mentioned above, and exemplary embodiment can not overcome any one of above problem.
Exemplary embodiment provides can be by considering cache memory efficiency and load balancing to multiple trees
Searcher distributes light generated to enhance the image processing apparatus and its image processing method of the performance of tree searcher.
According to one exemplary embodiment, image processing apparatus includes: light generator, is configurable to generate at least one light
Line；Multiple tree searchers are configured as executing the tree search of at least one light generated；Area divider is configured as
At least one light generated is divided into multiple sub-districts by the information of the quantity based on the tree search about previous image frames；With
And light distributor, it is configured as the multiple sub-districts divided by area divider multiple tree searchers are separately assigned.
The area divider, which can be configured as, stores multiple strokes that wherein sub-district is divided according at least one method
It is sub-category, it determines and is directed to the division classification that previous image frames have the shortest tree search time in the multiple division classification, and
And at least one light generated is divided into multiple sub-districts corresponding with identified division classification.
The area divider, which can be configured as, is divided into multiple sub-districts for previous image frames according to the division classification,
By by by with include the quantity of the corresponding tree search of pixel in each sub-district multiplied by the size with each sub-district
Corresponding weight and the value obtained adds up and calculates tree search value, and division is determined using tree search value calculated
The tree search time of classification.
The area divider, which can be configured as, calculates multiple tree search values for dividing classification, and determines to have and be counted
The division classification of minimum tree search value in the tree search value of calculation is that have the division classification of the shortest tree search time.
The area divider, which can be configured as, to be stored in memory about the identified information for dividing classification,
And provided when subsequent picture frame is divided into sub-district to the light distributor stored about identified division class
Other information.
The light distributor can be included at least one light in the sub-district that wherein tree search has not been completed
Extraly distribute to having been completed including at least one light in corresponding sub-district in the multiple tree searcher
Set the tree searcher of search.
Image processing apparatus can also include: intersection search device, be configured as executing the intersection of the light by tree search
Test and tinter are configured as executing the coloring of the light by cross-beta.
According to another exemplary embodiment, the image processing method of image processing apparatus includes: to generate at least one light,
At least one light generated is divided into multiple sub-districts by the information of the quantity based on the tree search about previous image frames, will
Multiple tree searchers are separately assigned in the sub-district of the multiple division, and, it is generated by the execution of the multiple tree searcher
At least one light tree search.
In the division operation, wherein sub-district can be deposited by the multiple division classifications for using at least one method to divide
Storage, the multiple division classification is directed to previous image frames in the middle can be determined with the division classification of the shortest tree search time,
And at least one light generated can be divided into multiple sub-districts corresponding with identified division classification.
In the division operation, previous image frames can be divided by multiple sub-districts according to the division classification, it can be with
By by by with include the quantity of the corresponding tree search of pixel in each sub-district multiplied by the size with each sub-district
Corresponding weight and the value obtained adds up and calculates tree search value, and tree search value calculated can be used to determine
The tree search time for dividing classification.
In the division operation, multiple tree search values for dividing classification can be calculated, and have tree calculated
The division classification of minimum tree search value in search value can be determined that the division classification with the shortest tree search time.
The division operation may include: will about it is identified divide classification information storage in memory, and
There is provided to the light distributor when subsequent picture frame is divided into sub-district stored divide classification about identified
Information.
In the batch operation, being included therein at least one light in the sub-district that has not been completed of tree search can be with
Extraly being distributed to having been completed in the multiple tree searcher includes at least one light in corresponding sub-district
Tree search tree searcher.
The method can also include executing the cross-beta of the light by tree search and executing to pass through cross-beta
Light coloring.
According to another exemplary embodiment, the method for handling image include: generate at least one light, by it is generated extremely
Few light is divided at least one sub-district, is distributed at least one light generated at least according to sub-area division information
One tree searcher, and searched for by the tree that at least one described tree searcher executes at least one light, wherein described
At least one light was distributed according to the tree search time of previous image frames.
The method can also include: the cross-beta for executing at least one light searched for by tree, and execute quilt
Set the coloring at least one light searched for.
At least one described light can be divided at least one sub-district corresponding with sub-area division information.
According to another exemplary embodiment, image processing apparatus includes: light generator, is configurable to generate at least one
Light, area divider are configured as picture frame being divided at least one sub-district, and light distributor is configured as to give birth to
At at least one light distribute at least one sub-district and at least one tree searcher, be configured as to distributed to
Few light executes tree search, wherein at least one described light is distributed according to sub-area division information.
Described image processing unit can also include cross-beta device, be configured as executing having already been through and set search extremely
The cross-beta of few light.
Tinter, which can be configured as, colours at least one light for having already been through cross-beta.
The additional and/or other aspect and advantage of exemplary embodiment will be explained partly in the following description
It states, and is partly become apparent from the description, or practical exemplary embodiment acquistion can be passed through.
Advantageous effect of the invention
As mentioned above one group of light with high correlation is distributed to by single tree searcher according to sub-district, so tree
The cache hit rate of searcher can be enhanced.Therefore, the performance and image processing apparatus 300 of searcher 350 are set
Overall performance can be enhanced, and external memory bandwidth can reduce.In addition, because one group of light is taken into account previous frame
And be adaptively divided into subgroup (subgroup) and be assigned to tree searcher, so good load balancing can be tieed up
It holds.
Detailed description of the invention
With reference to attached drawing, above and/or other aspect of exemplary embodiment be will be apparent from, in which:
Fig. 1 is the block diagram of the configuration of image processing apparatus in the related technology；
The method that Fig. 2A and Fig. 2 B shows distribution light in the related technology；
Fig. 3 is the block diagram of the configuration of image processing apparatus according to one exemplary embodiment；
Fig. 4 shows multiple division classifications according to one exemplary embodiment；
The tree that Fig. 5 shows each of the multiple division classifications of calculating for picture frame according to one exemplary embodiment is searched
The method of rope value；And
Fig. 6 is the flow chart of image processing method according to one exemplary embodiment.
Specific embodiment
Exemplary embodiment can be modified diversely.Therefore, exemplary embodiment is shown in the accompanying drawings and detailed
It is described in detail in thin description.Item defined in description, such as detailed construction and element are provided to help to of the invention
Comprehensive understanding.It is obvious, therefore, that may not need the item that those are specifically defined to execute exemplary embodiment.Also, it is well known
Function or structure are not described in detail, because they may obscure the present invention with unnecessary details.
Term " first ", " second " ... can be used to describe multiple assembly, but component is not limited by the term.Institute
Term is stated only to be used only for distinguishing a component with other components.
The term is only used only for description exemplary embodiment, it is not intended that the model of limitation exemplary embodiment
It encloses.The expression of odd number further includes the meaning of plural number, as long as it is not stated differently.Term " includes " and " Consists of " instruction
The presence of feature in the description, number, step, operation, component, element or their combination is write, but is not excluded for one
A or a number of other features, number, step, operation, component, element or their combination presence or addition possibility
Property.
According to one exemplary embodiment, " module " or " unit " executes at least one functions or operations, and can be with hard
The combination of part, software or hardware and software is implemented.In addition, " module " implemented in addition to specific hardware must be utilized or
Except person's " unit ", multiple " modules " or multiple " units " are desirably integrated at least one module, and can use to
Lack a processor (not shown) to implement.
Exemplary embodiment is more fully described with reference to the drawings.In the following description, identical attached drawing reference
Label is used for identical element, even if being in different drawings also such.Therefore, the description of identical element is not weighed
It is multiple.
Fig. 3 is the block diagram of the configuration of image processing apparatus 300 according to one exemplary embodiment.With reference to Fig. 3, image procossing dress
Setting 300 may include light generator 310, light distributor 320, area divider 330, memory 340, multiple tree searchers
350-1,350-2,350-3 ... cross-beta device 360 and tinter 370.Image processing apparatus 300 can use such as intelligent
A variety of image processing apparatus of phone, intelligence TV, tablet personal computer (PC), notebook PC etc. are implemented.
Light generator 310 generates light according to ray trace method to export image data.Light generator 310 can
To generate the light of each of multiple pixels for forming described image data.
More specifically, light generator 310 can generate information based on light to generate at least one light.Specifically,
Light generator 310 can generate the screen coordinate value of acquisition of information light based on light.In addition, light generator 310 can be with
It is generated based on the position (that is, the position of the eyes of user relative to screen) of acquired screen coordinate value and camera
Light.
Light distributor 320 can will be by light generator based on the sub-area division information provided from area divider 330
310 light generated distribute to multiple tree searcher 350-1,350-2,350-3 ....Sub-area division information may include about right
The information for multiple sub-districts that picture frame is divided.In other words, light distributor 320 can be included within sub-district --- by area
Domain divider 330 in each frame from adaptively dividing --- in light be separately assigned multiple tree searcher 350-1,
350-2、350-3...。
The light that light distributor 320 can be included in the sub-district that tree search is completed not yet is extraly distributed to
Multiple tree searcher 350-1,350-2,350-3 ... having been completed to the tree for including light in corresponding sub-district in the middle
The tree searcher of search.For example, when for including setting to search for the light in the sub-district for distributing to the first tree searcher 350-1
When rope has been completed, and when for include in the light in the sub-district for distributing to the second tree searcher 350-2 for set search
When having not been completed, light distributor 320 can be included within the light in the sub-district for distributing to the second tree searcher 350-2
A part is reassigned to the first tree searcher 350-1.As mentioned above light distributor 320 will can adaptively come from
The light for the tree searcher that wherein tree search has not been completed distributes to the tree searcher for having completed tree search, so faster
Tree search can be possible.
Because area divider 330 divides sub-district, light based on the information of the quantity of the tree search about previous frame
The light that line distributor 320 can be included in initial frame with preordering method distribute to multiple tree searcher 350-1,350-2,
350-3…。
Subsequent frame is divided into multiple by the information of quantity of the area divider 330 based on the tree search about previous image frames
Sub-district.
More specifically, area divider 330 can store multiple division classifications, in the multiple division classification sub-district with
Multiple methods are divided.For example, when picture frame is made of 8 × 8 regions and is divided into 4 sub-districts, area divider 330
It can store the first division classification 410 and divide classification 490 to the 9th, as shown in Figure 4.In more detail, first classification 410 is divided
Be located at upper left 2 × 2 the first sub-district, positioned at 2 × 6 the second sub-district of upper right portion, positioned at the 6 of bottom left section
× 2 third sub-district and 6 × 6 the 4th sub-district positioned at lower right-most portion.Second, which divides classification 420, has positioned at upper left
2 × 4 the first sub-district, 2 × 4 the second sub-district positioned at upper right portion, 6 × 4 third sub-district positioned at bottom left section and
Positioned at 6 × 4 the 4th sub-district of lower right-most portion.Third divide classification 430 have positioned at upper left 2 × 6 the first sub-district,
2 × 2 the second sub-district positioned at upper right portion, 6 × 6 third sub-district positioned at bottom left section and positioned at lower right-most portion 6 ×
2 the 4th sub-district.4th divide classification 440 have positioned at upper left 4 × 2 the first sub-district, positioned at upper right portion 4 ×
6 the second sub-district, 4 × 2 third sub-district positioned at bottom left section and 4 × 6 the 4th sub-district positioned at lower right-most portion.5th
Divide the first to the 4th sub-district that classification 450 has 4 × 4.6th divide classification 460 have positioned at upper left 4 × 6 the
One sub-district, 4 × 2 the second sub-district positioned at upper right portion, 4 × 6 third sub-district positioned at bottom left section and be located at right lower quadrant
4 × 2 the 4th sub-district divided.7th, which divides classification 470, has 6 × 2 the first sub-district positioned at upper left, is located at upper right quarter
Point 6 × 6 the second sub-district, 2 × 2 third sub-district positioned at bottom left section and 2 × 6 positioned at lower right-most portion the 4th son
Area.8th divides the second son that classification 480 has 6 × 4 the first sub-district positioned at upper left, 6 × 4 positioned at upper right portion
Area, 2 × 4 third sub-district positioned at bottom left section and 2 × 4 the 4th sub-district positioned at lower right-most portion.9th divides classification
490 have 6 × 6 the first sub-district positioned at upper left, 6 × 2 the second sub-district positioned at upper right portion, are located at bottom left section
2 × 6 third sub-district and 2 × 2 the 4th sub-district positioned at lower right-most portion.
It is only as described above exemplary embodiment at four sub-districts by each division category division.Being also possible to will be every
A category division that divides is at the multiple sub-districts for being different from four sub-districts.The quantity of sub-district in each division classification can correspond to
Set the quantity of searcher 350.For example, each quantity for dividing the sub-district in classification can when the quantity for setting searcher 350 is eight
To be eight.
In addition, area divider 330 can determine in the division classification 410 to 490 of multiple storages for prior images
Frame has the division classification of the shortest tree search time, and previous image frames is divided into corresponding with identified division classification
Multiple sub-districts.More specifically, area divider 330 can calculate the tree search value proportional to search time is set, and make
Each tree search time for dividing classification is determined with tree search value calculated.
At this moment, area divider 330 can calculate tree search value by following steps: divide classification according to specific
Previous image frames are divided into multiple sub-districts, and by by with include that the corresponding tree of pixel in each sub-district is searched for
Quantity multiplied by weight (weighting) corresponding with the size of each sub-district and obtain value add up.It can be used down
The mathematic(al) representation 1 in face calculates tree search value.
[mathematic(al) representation 1]
In mathematic(al) representation 1, TRV_CRT (i) indicates the quantity of the tree search of ith pixel, and w1 indicates the first sub-district
Weight, w2 indicate the second sub-district weight, w3 indicate third sub-district weight, and w4 indicate the 4th sub-district weight.It is described
Weight w1, w2, w3 and w4 respectively correspond to the size of corresponding sub-district.
For example, when the quantity for the tree search for including each pixel in previous image frames is illustrated in Fig. 5, region
Divider 330 can calculate each tree search for dividing classification 410 to 490 as shown in following mathematic(al) representation 2
Value.
[mathematic(al) representation 2]
Pcat1=(0*4)+(13*12)+(0*12)+(23*36)=984
Pcat2=(1*8)+(12*8)+(1*24)+(22*24)=656
Pcat3=(1*12)+(12*4)+(1*36)+(22*12)=360
Pcat4=(0*8)+(19*24)+(0*8)+(17*24)=864
Pcat5=(1*16)+(18*16)+(1*16)+(16*16)=576
Pcat6=(1*24)+(18*8)+(1*24)+(16*8)=320
Pcat7=(0*12)+(28*36)+(0*4)+(8*12)=1114
Pcat8=(2*24)+(26*24)+(0*8)+(8*8)=736
Pcat9=(2*36)+(26*12)+(0*12)+(8*4)=416
Work as in addition, area divider 330 can be determined with the tree search value calculated for multiple division classifications 410 to 490
In the division classification of minimum tree search value be that there is the division classification of the shortest tree search time.For example, when the first division classification
When the tree search value of 410 to the 9th division classifications 490 is computed as described above, area divider 330 can be determined with minimum
Setting the 6th of search value and dividing classification 460 is to have the division classification of the shortest tree search time.
In addition, area divider 330 can be provided to light distributor 320 has the shortest tree about for preceding frame image
The information of the division classification of search time, as the sub-area division information for subsequent picture frame.For example, when such as mathematic(al) representation
Such determination calculated has the division classification of the shortest tree search time in 2 when being the 6th division classification 460, area divider
330 can provide the information for dividing classification 460 about the 6th to light distributor 320.
Light can be distributed to multiple tree searcher 350- based on the sub-area division information received by light distributor 320
1,350-2,350-3….For example, being drawn when light distributor 320 receives the information for dividing classification 460 about the 6th as sub-district
When dividing information, light generated can be distributed to multiple tree searchers based on the 6th division classification 460 by light distributor 320
350-1,350-2,350-3….More specifically, light distributor 320 can be included within the first son of the 6th division classification 460
Light in area distributes to the first tree searcher 350-1, the light point being included in the second sub-district of the 6th division classification 460
Dispensing the second tree searcher 350-2, the light being included in the third sub-district of the 6th division classification 460 are distributed to third tree and are searched
Rope device 350-3, and the light being included in the 4th sub-district of the 6th division classification 460 distributes to the 4th tree searcher 350-
4。
In addition, the information about the division classification determined for previous frame can be stored in storage by area divider 330
In device 340, and provides and described deposit about being stored in light distributor 320 when subsequent picture frame is divided into multiple sub-districts
The information of division classification in reservoir 340.
It is each tree searcher 350 execute institute received light tree search for.Kd-Tree can be used in each tree searcher 350,
It is a kind of space segmentation tree method.Kd- tree may include frame node (box node), internal node (inner node) and
Leaf node (leaf node).Leaf node may include for being directed toward including at least one 3 angle informations in geometric data
Triangle chained list (triangle list).In an exemplary embodiment of the invention, when including three angle informations in geometric data
It is when implementing, to correspond to array indexing including the triangle chained list in leaf node using array.
However, the use of kd-Tree being only as described above exemplary embodiment.Such as enclosure body level (bounding
Volume hierarchy) and/or it is other tree searching method a variety of tree searching methods can be used.
Specifically, because image processing apparatus 300 may include there may be heavy duty for tree search
Multiple tree searcher 350-1,350-2,350-3 ... as shown in Figure 3, to set searcher 350-1,350-2,350-3 ...
Each of can independently execute light tree search.
Cross-beta device 360 executes cross-beta to the light for wherein performing tree search.More specifically, cross-beta device
The leaf node that 360 search intersect with light, and reading includes the triangle chained list in the leaf node.In addition, cross-beta
Device 360 reads the coordinate information of the triangle chained list, and executes cross-beta to given light.In addition, cross-beta device 360
It is hit a little using the vector value of distance and the given light away from the triangle for being hit (hit) by the light to calculate light-triangle
Coordinate value.
The light that tinter 370 is performed wherein cross-beta executes coloring.More specifically, tinter 370 can make
Light-triangle is calculated with the information being stored in material memories and triangle information cache hits color-values a little, and
And coloring is executed based on color-values calculated.
Hereinafter, it is described with reference to Figure 6 the image processing method of image processing apparatus 300.
Firstly, image processing apparatus 300 generates light (S610).Image processing apparatus 300 can be generated for picture frame
Each pixel light.
Then, light generated is divided into multiple by the quantity of tree search of the image processing apparatus 300 based on previous frame
Sub-district (S620).More specifically, image processing apparatus 300 can store multiple division classifications (wherein using in multiple methods
One divide sub-district), determine in the division classification of multiple storages when there is the shortest tree search for previous image frames
Between division classification, and light generated is divided into multiple sub-districts corresponding with identified division classification.More in detail
Carefully, image processing apparatus 300 can calculate tree search value by following steps: will previously be schemed according to the specific classification that divides
As frame is divided into multiple sub-districts, and by will multiply with the quantity for including the corresponding tree search of pixel in each sub-district
The value obtained with weight corresponding with the size of each sub-district adds up.In addition, image processing apparatus 300 can calculate it is more
A tree search value for dividing classification, and determine the division classification with the minimum tree search value in tree search value calculated
It is the division classification with the shortest tree search time.
Then, the multiple sub-district is distributed to multiple tree searcher 350-1,350-2,350- by image processing apparatus 300
3…(S630).More specifically, image processing apparatus 300 can be based on according to the division classification with the shortest tree search time
Information distributes to each corresponding tree searcher come the light for dividing multiple sub-districts, and being included in each sub-district.
Then, image processing apparatus 300 is to including that light in the sub-district distributed executes tree search (S640).Image
The tree search that multiple tree searcher 350-1,350-2,350-3 ... independently execute light can be used in processing unit 300.
Then, 300 pairs of the image processing apparatus light by tree search executes cross-beta and coloring (S650 and S660).
As mentioned above one group of light with high correlation is assigned to single tree searcher according to sub-district, so tree
The cache hit rate of searcher can be enhanced.Therefore, the performance and image processing apparatus 300 of searcher 350 are set
Overall performance can be enhanced, and external memory bandwidth can reduce.In addition, because one group of light is taken into account previous frame
And be adaptively divided into subgroup (subgroup) and be assigned to tree searcher, so good load balancing can be tieed up
It holds.
Device according to one exemplary embodiment may include: processor；The memory of storage and operation program data；Permanently
Property reservoir, such as disk drive；Communication port, with communication with external apparatus；And user's interface device, such as touch tablet,
Key and button.The method implemented using software module or algorithm is can to run and can be read by calculator on a processor
The code or program command taken, and can be stored in the recording medium that can be readable by a computer.It can be calculated machine-readable
The recording medium taken may include magnetic storage medium (for example, read-only memory, random access memory, floppy disk, hard disk etc.)
With optically-readable media (for example, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) etc.).It can be calculated
The machine-readable recording medium taken can be dispersed in the computer system by network connection, so as to the code being readable by a computer
It can store and run with dispersing method.That the medium can be is being readable by a computer, storage in memory and
It is run by processor.
Exemplary embodiment can be indicated by functional block components and a variety of processing steps.The functional block can be to execute
The hardware and or software components of a variety of quantity of specific function is implemented.For example, exemplary embodiment can be using by one
Or the control of multi-microprocessor or other control devices executes the direct circuit components of multiple functions, such as memory,
Handle logic and look-up table.When the component be as software program or software element to run when, demonstration of the invention
Embodiment can be realized with programming language or scripting language (such as C, C++, Java and assembler), including the use of number
The many algorithms implemented according to the combination of structure, process, routine or other program components.Functional factor can use by one
The algorithm of a or multiple processors operation is implemented.In addition, exemplary embodiment can be used at electrical arrangement, signal
Reason, and/or the relevant technologies of data processing.Term " mechanism ", " element ", " component " and " component " can be used broadly, and
And it is not limited to mechanical component or physical assemblies.The term may include the set of the routine of software connected to the processor.
Above-mentioned exemplary embodiment and advantage are only exemplary, and will not be interpreted to be limited.In order to
Succinctly, in the related technology electronic building brick, control system, software and other functional factors can be omitted.In addition, as showing
Example, the connection of the line or connector (connection member) between component in attached drawing show function connects and/or
Physically or electrically road connects, and therefore, and in actual device, multiple functions connection alternatively or additionally, physics connect
It connects or circuit connection can be provided.
In the present specification (especially in the claims), the use of term " described " and similar reference term can be with
Refer to both odd number and plural number.In addition, when writing range, the single value within the scope of this is included, unless there is no opposite
Narration.It therefore, is also the same for the case where writing the single value for forming the range in detailed description.Finally, group
It can execute at the step of method, if the sequence is not write clearly or if deposit in the proper sequence
If opposite narration.This exemplary embodiment is not always to be limited to write the sequence of step.Above-mentioned exemplary embodiment and
Advantage is only exemplary, and will not be interpreted to be limited.Also, to the description of exemplary embodiment be intended at
To be illustrative, rather than it is used to limit the scope of the claims, and many replacements, modification and variation are for this field skill
It will be apparent for art personnel.
Claims (12)
1. a kind of image processing apparatus, comprising:
Light generator is configurable to generate at least one light；
Multiple tree searchers are configured as executing the tree search of at least one light generated；
Area divider is configured as the information based on the quantity for setting search about previous image frames for generated at least one
A light is divided into multiple sub-districts, wherein multiple sub-districts are divided category division according at least one of multiple division classifications；
And
Light distributor is configured as the multiple sub-districts divided by area divider multiple tree searchers are separately assigned,
Wherein, the area divider is configured as storing the multiple division classification, determines in the multiple division classification
For previous image frames have the shortest tree search time division classification, and by least one light generated be divided into
The corresponding multiple sub-districts of identified division classification.
2. image processing apparatus as described in claim 1, wherein the area divider is configured as according to the division class
Previous image frames are not divided into multiple sub-districts, by by by with include that the corresponding tree of pixel in each sub-district is searched
The Calais Zhi Xiang that the quantity of rope is obtained multiplied by weight corresponding with the size of each sub-district calculates tree search value, and uses
It is calculated to set search value to determine the tree search time for dividing classification.
3. image processing apparatus as described in claim 1, wherein the area divider is configured as calculating multiple division classes
Other tree search value, and determine that the division classification with the minimum tree search value in tree search value calculated is that have most
The division classification of short tree search time.
4. image processing apparatus as described in claim 1, wherein the area divider is configured as will be about identified
The information for dividing classification stores in memory, and mentions when subsequent picture frame is divided into sub-district to the light distributor
For being stored about the identified information for dividing classification.
5. image processing apparatus as described in claim 1, wherein the light distributor, which is included within wherein tree search, not to be had also
Have at least one light in the sub-district of completion extraly distribute to it is the multiple tree searcher in have been completed including
The tree searcher of the tree search of at least one light in corresponding sub-district.
6. image processing apparatus as described in claim 1, further includes:
Intersection search device is configured as executing the cross-beta of the light by tree search；And
Tinter is configured as executing the coloring of the light by cross-beta.
7. a kind of image processing method of image processing apparatus, which comprises
Generate at least one light；
At least one light generated is divided into multiple sons by the information of the quantity based on the tree search about previous image frames
Area, wherein multiple sub-districts are used at least one of multiple division classifications to divide category division；
Multiple tree searchers are separately assigned in the sub-district of the multiple division；And
The tree search of at least one light generated is executed by the multiple tree searcher,
Wherein, it in the division operation, wherein the multiple division classification is stored, is directed in the multiple division classification
There is previous image frames the division classification of the shortest tree search time to be determined, and at least one light generated is divided into
Multiple sub-districts corresponding with identified division classification.
8. the method for claim 7, wherein in the division operation, according to the division classification by prior images
Frame is divided into multiple sub-districts, by by by with include the corresponding tree search of pixel in each sub-district quantity multiplied by
Weight corresponding with the size of each sub-district and the Calais Zhi Xiang that obtains calculates tree search value, and searched using tree calculated
Rope value come determine it is described divide classification tree search time.
9. the method for claim 7, wherein in the division operation, multiple tree search values for dividing classification are counted
It calculates, and the division classification with the minimum tree search value in tree search value calculated is confirmed as with the shortest tree search
The division classification of time.
10. the method for claim 7, wherein the division operation includes:
It will be stored in memory about the identified information for dividing classification；And
There is provided when subsequent picture frame is divided into sub-district to the light distributor stored about identified division class
Other information.
11. the method for claim 7, wherein in the batch operation, be included therein tree search and have not been completed
Sub-district at least one light extraly distributed to having been completed and be included in phase in the multiple tree searcher
Answer the tree searcher of the tree search of at least one light in sub-district.
12. the method for claim 7, further includes:
Execute the cross-beta of the light by tree search；And
Execute the coloring of the light by cross-beta.
Applications Claiming Priority (5)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US201361864877P | 2013-08-12 | 2013-08-12 | |
| US61/864,877 | 2013-08-12 | ||
| KR10-2014-0022489 | 2014-02-26 | ||
| KR1020140022489A KR102176511B1 (en) | 2013-08-12 | 2014-02-26 | Apparatus and Method for processing image |
| PCT/KR2014/007484 WO2015023106A1 (en) | 2013-08-12 | 2014-08-12 | Apparatus and method for processing image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105612558A CN105612558A (en) | 2016-05-25 |
| CN105612558B true CN105612558B (en) | 2019-02-01 |
Family
ID=52578999
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201480055150.0A Expired - Fee Related CN105612558B (en) | 2013-08-12 | 2014-08-12 | Device and method for handling image |
Country Status (2)
| Country | Link |
|---|---|
| KR (1) | KR102176511B1 (en) |
| CN (1) | CN105612558B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR102493461B1 (en) | 2015-08-31 | 2023-01-30 | 삼성전자 주식회사 | System and Method of rendering |
| US10262456B2 (en) | 2015-12-19 | 2019-04-16 | Intel Corporation | Method and apparatus for extracting and using path shading coherence in a ray tracing architecture |
| US9892544B2 (en) * | 2015-12-22 | 2018-02-13 | Intel Corporation | Method and apparatus for load balancing in a ray tracing architecture |
| KR102670714B1 (en) * | 2018-12-17 | 2024-06-03 | 서강대학교산학협력단 | Method for matching correspondence by using decision forest based on GPU |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103049927A (en) * | 2013-01-17 | 2013-04-17 | 浙江大学 | Real-time ray tracing and rendering method based on GPU (Graphics Processing Unit) aggregate |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100924122B1 (en) | 2007-12-17 | 2009-10-29 | 한국전자통신연구원 | Ray tracing device based on pixel processing element and method thereof |
| US8189001B2 (en) | 2010-01-04 | 2012-05-29 | Adshir Ltd. | Method and apparatus for parallel ray-tracing employing modular space division |
-
2014
- 2014-02-26 KR KR1020140022489A patent/KR102176511B1/en active IP Right Grant
- 2014-08-12 CN CN201480055150.0A patent/CN105612558B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103049927A (en) * | 2013-01-17 | 2013-04-17 | 浙江大学 | Real-time ray tracing and rendering method based on GPU (Graphics Processing Unit) aggregate |
Also Published As
| Publication number | Publication date |
|---|---|
| KR102176511B1 (en) | 2020-11-09 |
| CN105612558A (en) | 2016-05-25 |
| KR20150020012A (en) | 2015-02-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP4198909A1 (en) | Image rendering method and apparatus, and computer device and storage medium | |
| US10776688B2 (en) | Multi-frame video interpolation using optical flow | |
| US10621762B2 (en) | Systems and methods for high dimensional 3D data visualization | |
| Akinci et al. | Parallel surface reconstruction for particle‐based fluids | |
| US11106261B2 (en) | Optimal operating point estimator for hardware operating under a shared power/thermal constraint | |
| CN105684037B (en) | Graphics processing unit | |
| US9153035B2 (en) | Depth map movement tracking via optical flow and velocity prediction | |
| CN103503030B (en) | Image processing device for specifying depth of object present in real space by performing image processing, stereoscopic viewing device, and integrated circuit | |
| CN105612558B (en) | Device and method for handling image | |
| US10810784B1 (en) | Techniques for preloading textures in rendering graphics | |
| TW201108152A (en) | Displaying a visual representation of performance metrics for rendered graphics elements | |
| CN108280455A (en) | Human body critical point detection method and apparatus, electronic equipment, program and medium | |
| US20190122421A1 (en) | Batch rendering method, device, and apparatus | |
| CN106125938A (en) | A kind of information processing method and electronic equipment | |
| US11727632B2 (en) | Shader binding management in ray tracing | |
| US7999808B1 (en) | Parallel processing system, method, and computer program product for executing node traversal or primitive intersection | |
| US20240203030A1 (en) | 3d model rendering method and apparatus, electronic device, and storage medium | |
| US20230368432A1 (en) | Synthesized Camera Arrays for Rendering Novel Viewpoints | |
| US20210366176A1 (en) | Shadow culling | |
| US8072454B1 (en) | Parallel processing system, method, and computer program product for selecting a ray tracing entity from a group of ray tracing entities for processing | |
| CN113822975B (en) | Techniques for efficient sampling of images | |
| US9519992B2 (en) | Apparatus and method for processing image | |
| US11372756B2 (en) | Memory pool management | |
| US8059123B1 (en) | Parallel processing system, method, and computer program product for postponing the execution of primitive intersection | |
| US10747394B2 (en) | Providing data visualization |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190201 Termination date: 20200812 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |