CN102999885A - Method and device for determining average brightness by Retinex video enhancement algorithm - Google Patents
Method and device for determining average brightness by Retinex video enhancement algorithm Download PDFInfo
- Publication number
- CN102999885A CN102999885A CN2012103621921A CN201210362192A CN102999885A CN 102999885 A CN102999885 A CN 102999885A CN 2012103621921 A CN2012103621921 A CN 2012103621921A CN 201210362192 A CN201210362192 A CN 201210362192A CN 102999885 A CN102999885 A CN 102999885A
- Authority
- CN
- China
- Prior art keywords
- centered
- average brightness
- central pixel
- fritter
- mean value
- 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.)
- Pending
Links
Images
Abstract
The invention discloses a method and a device for determining average brightness by the Retinex video enhancement algorithm. The method includes: partitioning images, calculating the average brightness of an area taking each small block of central pixel point as a center by using conventional methods, for the area average brightness taking noncentral pixel point as a center, sharing the area average brightness taking the central pixel point of the small block as the center, or subjecting the area average brightness taking two adjacent central pixel points as a center to bilinear interpolation to obtain. Consequently, the amount of calculation is greatly reduced, requirements on an internal memory and processing speed are lowered, and cost is saved.
Description
Technical field
The present invention relates to the video enhancement techniques field, particularly relate to a kind of Retinex video enhancement algorithm average brightness and determine method and apparatus.
Background technology
In the Retinex video enhancement algorithm, need to calculate the average brightness of a regional extent centered by each pixel.Take the 1920*1080 image as example, every two field picture calculates a regional luminance mean value according to each point, need to calculate 1920*1080=2073600 time, refresh rate calculating by per second 60 frames needs 1920*1080*60=124416000 time/s, if need to calculate the average brightness of a plurality of yardsticks, calculated amount also needs many times of growths.Can find out that this algorithm all is great expense to calculated amount, data bus bandwidth or memory bandwidth.
The method of traditional calculations regional luminance mean value is: get some scale, calculate with the average brightness of central point to the zone of spreading all around.Equal 6 as example take yardstick, referring to Fig. 1, the calculating pixel point coordinate is the average brightness of (x, y), calculates exactly with coordinate points (x, y) centered by, to around spread 6 pixels, i.e. the pixel region of 13*13 centered by coordinate points (x, y), calculate the average brightness of these all pixels of zone, formula is as follows.This mean value is an important parameter of video enhancement algorithm.
The average brightness computing formula:
Wherein, A
XyBe average brightness, Y
IjBe the brightness value of each pixel in the regional extent, N is pixel quantity in the regional extent.
Need to carry out a large amount of additions because each regional average brightness calculates, data access number of times and operand are all very large, cause above-mentioned computing method to have the great problem of resource consumption.In order to address this problem, produced a kind of regional luminance mean value calculation method of optimization: first entire image is carried out traversing operation one time, store each all pixel of the pixel upper left corner brightness and, formula is as follows.As shown in Figure 2, the brightness value sum of all pixels of coordinate (m, the n) upper left corner is S
Mn:
Wherein, S
MnBe the brightness sum of all pixels of pixel coordinate (m, the n) upper left corner, Y
IjBrightness value for each pixel in the regional extent.
After traversing operation is finished, take calculating regional ABCD average brightness as shown in Figure 3 as example, read first brightness and the S of this position, four angles, zone
A, S
B, S
C, S
D, mean value A
MnCan calculate by following formula:
Wherein, A
MnBe the average brightness of ABCD institute region, S
A, S
B, S
C, S
DRepresent respectively A, B, C, D point upper left corner brightness value sum, N represents that ABCD institute region comprises the number of pixel.
Said method only needs image is once traveled through, and simplified operation has greatly reduced operand than traditional computing method greatly.But each regional luminance mean value calculation still needs more data to process, and causes each regional luminance mean value calculation still to need to consume more system resource.
In sum, because calculated amount is excessive, the internal memory utilized bandwidth is excessive, causes these algorithms to process in real time high-definition picture, can only process low-resolution image; Perhaps must use the internal memory of high speed processing chip and more speed to form disposal system with high cost, could process in real time high-definition picture.
Summary of the invention
For above situation, the present invention proposes a kind of Retinex video enhancement algorithm average brightness and determine method and apparatus, with the operand in the further Dimming mean value deterministic process.
A kind of Retinex algorithm for image enhancement average brightness is determined method, comprises step:
Entire image is divided into a plurality of fritters, the concrete number of the fritter weighing apparatus of between the calculated amount of average brightness and computational accuracy, making even;
The regional luminance mean value of calculating centered by each fritter central pixel point;
Regional luminance mean value centered by non-central pixel shares the regional luminance value centered by this fritter central pixel point;
Or
Regional luminance mean value centered by non-central pixel is tried to achieve by the average brightness centered by this fritter central pixel point and centered by the adjacent isles central pixel point is carried out bilinear interpolation.
A kind of Retinex algorithm for image enhancement average brightness is determined device, comprising:
Divide module, be used for entire image is divided into a plurality of fritters, the concrete number of the fritter weighing apparatus of between the calculated amount of average brightness and computational accuracy, making even;
Central point mean value determination module is used for calculating the regional luminance mean value centered by each fritter central pixel point;
Non-central some mean value determination module is used for the regional luminance value centered by central pixel point as the regional luminance mean value centered by the non-central pixel of this fritter; Or, be used for calculating the regional luminance mean value centered by the non-central pixel of this fritter by the average brightness centered by this fritter central pixel point and centered by the adjacent isles central pixel point is carried out bilinear interpolation.
Retinex algorithm for image enhancement average brightness of the present invention is determined method and apparatus, image is carried out piecemeal to be processed, regional luminance mean value centered by every fritter central pixel point calculates with classic method, regional luminance mean value centered by non-central pixel, share the regional luminance mean value centered by this fritter central pixel point, perhaps the regional luminance mean value centered by two adjacent center pixels being carried out bilinear interpolation tries to achieve, thereby greatly reduced operand, reduced the requirement to internal memory and processing speed, provided cost savings.
Description of drawings
Fig. 1 is centered by coordinate (x, y), and 6 is the synoptic diagram of yardstick institute region;
Fig. 2 is all pixel synoptic diagram in coordinate (m, the n) upper left corner;
Fig. 3 is the ABCD area schematic;
Fig. 4 is the zone and area schematic centered by point (x+1, y) centered by point (x, y);
Fig. 5 is the synoptic diagram that image is divided into the fritter of 3*3;
Fig. 6 is the distribution schematic diagram of the regional luminance mean value centered by each fritter central pixel point;
Fig. 7 is the distribution schematic diagram of the regional luminance mean value centered by each pixel;
Fig. 8 is another distribution schematic diagram of the regional luminance mean value centered by each pixel;
Fig. 9 is the schematic flow sheet that Retinex video enhancement algorithm average brightness of the present invention is determined method;
Figure 10 is the structural representation that Retinex video enhancement algorithm average brightness of the present invention is determined device.
Embodiment
When calculating average brightness centered by adjacent two pixels, it is identical greatly that selected zone has.As shown in Figure 4, centre coordinate is (x, y) average brightness calculates selection range and centre coordinate is (x+1, y) average brightness calculates selection range and only differs (x-6) and (x+7) two be listed as, be that centre coordinate is (x+1, y) only have 13 pixel differences with the average brightness computer capacity of (x, y), account for 1/13 of overall pixel point quantity.Yardstick is chosen larger, and difference pixel proportion is less, if choose yardstick 21, the difference pixel of the regional luminance mean value calculation chosen area centered by neighbor pixel only accounts for 1/43 of overall pixel point quantity.Hence one can see that, the regional luminance mean value difference of calculating centered by neighbor pixel is very little, the present invention utilizes this characteristics, image is divided, the pixel that is divided into the same area shares a regional luminance mean value, the perhaps regional luminance mean value corresponding according to fixed pixel, the regional luminance mean value that approximate treatment adjacent pixels point is corresponding reduces operand with this.Explain in detail the present invention below in conjunction with accompanying drawing and embodiment.
Retinex algorithm for image enhancement average brightness of the present invention is determined method, as shown in Figure 9, comprises step:
Step S1, entire image is divided into a plurality of fritters, the concrete number of the fritter weighing apparatus of between the calculated amount of average brightness and computational accuracy, making even;
Step S2, the regional luminance mean value of calculating centered by each fritter central pixel point;
Step S3,
Regional luminance mean value centered by non-central pixel shares the regional luminance value centered by this fritter central pixel point;
Or
Regional luminance mean value centered by non-central pixel is tried to achieve by the average brightness centered by this fritter central pixel point and centered by the adjacent isles central pixel point is carried out bilinear interpolation.
By above step as can be known, the present invention carries out piecemeal to process on the basis of traditional calculations average brightness method: first entire image is divided into a plurality of smaller fritters, such as 2*2,3*3 or 4*4 etc., the fritter number is more, the result of calculation of average brightness is more accurate, but corresponding operand is also larger, therefore needs to make even between operand and degree of accuracy weighing apparatus.When the ranks number of fritter is even number after dividing, from getting one as central pixel point by paracentral 4 points.Take the 3*3 piecemeal as example, as shown in Figure 5, per 9 neighbor pixels of original image are divided into the piece of a 3*3.Behind the piecemeal, adopt first the regional luminance mean value of classic method calculating centered by each fritter central pixel point.The computing method of the regional luminance mean value centered by non-central pixel have two schemes, as described in step S3, the first scheme is regional luminance mean value corresponding to all pixel common center pixels in the same fritter, first scheme all has the rule of increasing or decreasing when being most according to the regional luminance mean value centered by adjacent pixels point, adopt bilinear interpolation, calculate according to the average brightness centered by this fritter central pixel point and centered by the adjacent isles central pixel point and get.
For example per 9 neighbors are divided into the piece of a 3*3, as shown in Figure 6.Calculate regional luminance mean value centered by each fritter central pixel point as A
MnAccording to the first scheme, as shown in Figure 7, the regional luminance mean value centered by non-central pixel is thought and is equaled regional luminance mean value A centered by its place fritter central pixel point
Mn, namely use A
MnFill all the other non-central pixels.According to first scheme, in the same fritter, in the time of respectively centered by each pixel, gained regional luminance mean value is different.As shown in Figure 9, in the fritter in the upper left corner, the regional luminance mean value centered by central pixel point is A11, and the regional luminance mean value in this fritter centered by non-central pixel comprises A111, A112 ..., A118.In the time of respectively centered by each fritter central pixel point, corresponding regional luminance mean value comprises A11, A12, A21, A22 ... so, adopt bilinear interpolation, A114=A11-(A12-A11)/3, A115=A11+(A12-A11)/3, A124=A11+(A12-A11) * 2/3.Similarly, all the other A
MnxValue also can try to achieve by bilinear interpolation.The effect of first scheme is better than the first scheme, but the calculated amount of the first scheme still less, determines to adopt which kind of scheme according to actual conditions during implementation.
Behind step S3, obtain the regional luminance sequence of average number of entire image centered by each pixel, this sequence number can be used as the parameter of Retinex video enhancement algorithm and calculates.
Retinex algorithm for image enhancement average brightness of the present invention is determined device, as shown in figure 10, comprising:
Divide module, be used for entire image is divided into a plurality of fritters, the concrete number of the fritter weighing apparatus of between the calculated amount of average brightness and computational accuracy, making even;
Central point mean value determination module is used for calculating the regional luminance mean value centered by each fritter central pixel point;
Non-central some mean value determination module is used for the regional luminance value centered by central pixel point as the regional luminance mean value centered by the non-central pixel of this fritter; Or, be used for calculating the regional luminance mean value centered by the non-central pixel of this fritter by the average brightness centered by this fritter central pixel point and centered by the adjacent isles central pixel point is carried out bilinear interpolation.
Division module, central point mean value determination module link to each other successively with non-central some mean value determination module and have consisted of this device, and the other technologies feature of this device is identical with said method, does not repeat them here.
According to the scheme of the invention described above, each fritter only need to calculate average brightness one time, can greatly reduce calculated amount.As press the 3*3 piecemeal, adopt the present invention program's calculated amount to only have 1/9 of original algorithm.Adopt and add the first scheme that the bilinear interpolation algorithm calculates, because the bilinear interpolation algorithm lacks than mean flow rate algorithm calculated amount, and do not need to carry out again the read operation of area pixel brightness value, can significantly save equally data bus bandwidth and take with memory bandwidth and take.
The above embodiment has only expressed several embodiment of the present invention, and it describes comparatively concrete and detailed, but can not therefore be interpreted as the restriction to claim of the present invention.Should be pointed out that for the person of ordinary skill of the art without departing from the inventive concept of the premise, can also make some distortion and improvement, these all belong to protection scope of the present invention.Therefore, the protection domain of patent of the present invention should be as the criterion with claims.
Claims (4)
1. a Retinex algorithm for image enhancement average brightness is determined method, it is characterized in that, comprises step:
Entire image is divided into a plurality of fritters, the concrete number of the fritter weighing apparatus of between the calculated amount of regional luminance mean value and computational accuracy, making even;
The regional luminance mean value of calculating centered by each fritter central pixel point;
Regional luminance mean value centered by non-central pixel shares the regional luminance value centered by this fritter central pixel point;
Or
Regional luminance mean value centered by non-central pixel is tried to achieve by the average brightness centered by this fritter central pixel point and centered by the adjacent isles central pixel point is carried out bilinear interpolation.
2. Retinex algorithm for image enhancement average brightness according to claim 1 is determined method, it is characterized in that, and be even number if described fritter comprises the ranks number of pixel, then to lean on a paracentral pixel as central pixel point.
3. a Retinex algorithm for image enhancement average brightness is determined device, it is characterized in that, comprising:
Divide module, be used for entire image is divided into a plurality of fritters, the concrete number of the fritter weighing apparatus of between the calculated amount of average brightness and computational accuracy, making even;
Central point mean value determination module is used for calculating the regional luminance mean value centered by each fritter central pixel point;
Non-central some mean value determination module is used for the regional luminance value centered by central pixel point as the regional luminance mean value centered by the non-central pixel of this fritter; Or, be used for calculating the regional luminance mean value centered by the non-central pixel of this fritter by the average brightness centered by this fritter central pixel point and centered by the adjacent isles central pixel point is carried out bilinear interpolation.
4. Retinex algorithm for image enhancement average brightness according to claim 3 is determined device, it is characterized in that, and be even number if described fritter comprises the ranks number of pixel, then to lean on a paracentral pixel as central pixel point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012103621921A CN102999885A (en) | 2012-09-25 | 2012-09-25 | Method and device for determining average brightness by Retinex video enhancement algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012103621921A CN102999885A (en) | 2012-09-25 | 2012-09-25 | Method and device for determining average brightness by Retinex video enhancement algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102999885A true CN102999885A (en) | 2013-03-27 |
Family
ID=47928422
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012103621921A Pending CN102999885A (en) | 2012-09-25 | 2012-09-25 | Method and device for determining average brightness by Retinex video enhancement algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102999885A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104796582A (en) * | 2015-04-20 | 2015-07-22 | 中国矿业大学 | Video image denoising and enhancing method and device based on random ejection retinex |
| CN105210360A (en) * | 2013-04-03 | 2015-12-30 | 日立麦克赛尔株式会社 | Video display device |
| CN105243671A (en) * | 2015-10-27 | 2016-01-13 | 浙江理工大学 | Testing and evaluating method for flatness of clothing in wear |
| CN109003582A (en) * | 2018-08-15 | 2018-12-14 | 四川长虹电器股份有限公司 | A method of improving super more backlight zoning control system image display effects |
| CN109509161A (en) * | 2018-12-20 | 2019-03-22 | 深圳市华星光电半导体显示技术有限公司 | Image intensifier device and image enchancing method |
| CN113452967A (en) * | 2020-03-24 | 2021-09-28 | 合肥君正科技有限公司 | Lens shadow correction method for removing blocking effect |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101325646A (en) * | 2007-06-13 | 2008-12-17 | 三星电子株式会社 | Method and apparatus for contrast enhancement |
-
2012
- 2012-09-25 CN CN2012103621921A patent/CN102999885A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101325646A (en) * | 2007-06-13 | 2008-12-17 | 三星电子株式会社 | Method and apparatus for contrast enhancement |
Non-Patent Citations (1)
| Title |
|---|
| 徐美芳等: "《基于统计特性的表面裂纹提取》", 《测试技术学报》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105210360A (en) * | 2013-04-03 | 2015-12-30 | 日立麦克赛尔株式会社 | Video display device |
| CN105210360B (en) * | 2013-04-03 | 2018-05-18 | 麦克赛尔株式会社 | Image display |
| CN104796582A (en) * | 2015-04-20 | 2015-07-22 | 中国矿业大学 | Video image denoising and enhancing method and device based on random ejection retinex |
| CN104796582B (en) * | 2015-04-20 | 2018-05-04 | 中国矿业大学 | Video image denoising and Enhancement Method and device based on random injection retinex |
| CN105243671A (en) * | 2015-10-27 | 2016-01-13 | 浙江理工大学 | Testing and evaluating method for flatness of clothing in wear |
| CN105243671B (en) * | 2015-10-27 | 2017-10-24 | 浙江理工大学 | A kind of clothes wear test and the evaluation method of flatness |
| CN109003582A (en) * | 2018-08-15 | 2018-12-14 | 四川长虹电器股份有限公司 | A method of improving super more backlight zoning control system image display effects |
| CN109509161A (en) * | 2018-12-20 | 2019-03-22 | 深圳市华星光电半导体显示技术有限公司 | Image intensifier device and image enchancing method |
| CN109509161B (en) * | 2018-12-20 | 2021-03-16 | 深圳市华星光电半导体显示技术有限公司 | Image enhancement device and image enhancement method |
| CN113452967A (en) * | 2020-03-24 | 2021-09-28 | 合肥君正科技有限公司 | Lens shadow correction method for removing blocking effect |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102999885A (en) | Method and device for determining average brightness by Retinex video enhancement algorithm | |
| WO2018040463A1 (en) | Data compression and decompression methods for demura table, and mura compensation method | |
| US10356385B2 (en) | Method and device for stereo images processing | |
| EP3161793B1 (en) | Adaptive partition mechanism with arbitrary tile shape for tile based rendering gpu architecture | |
| CN110390382B (en) | Convolutional neural network hardware accelerator with novel feature map caching module | |
| US20190318461A1 (en) | Histogram Statistics Circuit and Multimedia Processing System | |
| Xie et al. | A binocular vision application in IoT: Realtime trustworthy road condition detection system in passable area | |
| Li et al. | High throughput hardware architecture for accurate semi-global matching | |
| Zha et al. | A real-time global stereo-matching on FPGA | |
| US11682212B2 (en) | Hierarchical data organization for dense optical flow processing in a computer vision system | |
| CN104254868A (en) | Method and apparatus for increasing frame rate of an image stream using at least one higher frame rate image stream | |
| CN112419183A (en) | Method and device for reducing zoomed image, computer equipment and storage medium | |
| CN101483771B (en) | Method and apparatus for promoting frame rate | |
| CN104751441A (en) | Method and device for estimating image noise | |
| US20200327638A1 (en) | Connected component detection method, circuit, device and computer-readable storage medium | |
| CN102831633B (en) | Convolutional filtering optimizing method based on linear texture filtering | |
| CN111369435A (en) | Color image depth up-sampling method and system based on self-adaptive stable model | |
| CN108510441B (en) | Image multi-resolution display method based on staggered grids | |
| CN110648356A (en) | Multi-target tracking processing optimization method based on visual digital signal processing | |
| CN108875733A (en) | A kind of infrared small target quick extraction system | |
| CN105260526A (en) | Image based adaptive finite element mesh division method | |
| US20220121551A1 (en) | Method and device for calculating runtime of neural network on processor | |
| CN112146834B (en) | Method and device for measuring structural vibration displacement | |
| CN111191309A (en) | Real-time shielding and rejecting method, device and equipment suitable for BIM light weight | |
| US9338353B2 (en) | Method of and system for detecting motion in real time |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20130327 |