日本人白平衡AWB算法

所需积分/C币:31 2018-12-02 23:08:02 357KB PDF

这是日本著名相机公司的白平衡算法awb,全是English,希望在专业人士有所帮助,看得懂绝对有用,没用说明你们根本看不懂。
9.3 Modification and Reliability of tungsten Average 36 9.3.1 Typical Col 36 9.3.2 mixture of fluorescent and tungsten .37 9.3.3 Brightness 37 9.3.4 Strobe 37 10 Modification and Reliability of Unidentified Light Source Av erage 40 10.1 Classification 10.1.1 Fluorescent casc and Daylight Fluorescent case 4 42 10.1.2 Tungsten case .42 10.1.3 H-Daylight case and L-Daylight case 42 10.1.4 other case 42 11 Special Modification for averages and reliabilities 44 11.2 The number of Fluorescent, Tungsten, and ULS Paxels are small yy 11.1 Leaves detection 45 11.3 Bright Daylight scene 12 Weighted Average 47 13 Calculation of white balance gain 47 a Lincar to paxcl nonlincar function for Matt. bR and mi3 419 B Axel Nonlinear to Linear function for Matt. bR and mi3 51 List of figures 1 CFA Bayer Patte 9 2 Paxels in cfa ·· 3 Bayer Unit 10 4 RGB of axel 10 5 Flow Chart of axel calibrat 11 6 Flow Chart of white balance evaluation 14 7 Mix Tungsten area and mix daylight area 16 8 Distribution area of white under a light source 18 9 Daylight 20 10 relativity between brightness and rgtHDy 21 11 Fluorescent areas 22 12 Tungste 24 13 Unidentified light source area 25 14 mIDy 31 15 wgt Lv Dy in the case that mlDilDy> thr Dil Dy0 32 16 wgt Lv Dy in the other case 32 gtLvFi 35 18 tLⅴTu 38 19 classification for unidentified light source average 40 20 thrUn DyL D 41 21 Change Reliabilities along with the number of paxels 45 List of fables 1 Axel size and Number of paxels 10 pling ratios of panels fEval 12 saturation degree and uniformity degree 15 5 Luminance threshold facto 15 6 Mixed light Source area 17 7 thr Satlum 19 8 Parameters of Daylight Area 21 9 Parameters of fluorescent area ,23 10 Parameters of Tungsten Area 24 11 Parameters of unidentified area 26 12 Relativity between uniformity degree and weight 13 Parameters for Bad axel 27 14 parameters for stro be virtual axel 27 15 Typical colors of Three Daylights 30 16 Parameters of typical values of Daylight 30 17 Criteria of Low-Color-Tem perature Daylight 31 18 Parameters for Low-Color-Temperature Daylight ........ 31 19 Parameters for Daylight modification with Lv Value 33 20 Decision for Fluorescent or Fluorescent Tungste 33 21 Paraineters for Typical Fluorescent color 34 22 Parameters for modification by LV 35 23 Parameters for modification by Strobe 36 4 24 Typical colors of tungsten 36 25D fwgtFiT by 37 26 Parameters for wgt FITu .37 27 Parameters for modification by Lv 38 28 Parameters for modification by strobe 38 29 Paraeters of Unidentified average classification 41 30 decision whet her it is fuorescent scene or not 31 Parameters of leave detection 44 32 parameters for few paxel case 45 33 decision whether to cancel uls or not 46 34 coefDef 48 1 Introduction This document says the Simplified Auto White Balance Algorithm(sAWB) by weighted Average Method (WA-SAWB This version of AwB was introduced for DX3700(Shrike) first[2]. But, some part of this algorithm were modified and some new algorithm were introduced for matt. Blue ridge and mi3 cameras In this document. WA-awb for matt Blue ridge and mi3 cameras are shown 1.1 Features of WA-sAwB algorithm WA-SAWB algorithm is based on the following ideas When a color for a bright area is close to the pre-determined color for one f the expected light sources i.e. Daylight, Fluorescent, Tungsten), th color of the area must represent the color of its illuminant The brightest area must represent the color of the illuminant, even if the color of the area is not so close to the expected color for any light sources Fluorescent scene and Tungsten scene is usually darker than Dayligl scene The average of color for the dark area is not usually very different from the color of the light source WA-sAWB algorithm has the following features e In order to estimate the light source, an image is divided into 6x4 regions ocks are calle ed "Axel" in terms of Kodak. And if the pixel is estimated as Fluorescent. it is called fluorescent Pixel. Same for other light sources Even if the bightest paxel is not estimated as any light sources, the paxel is supposed to be illuminated by the unidentified light source" whose color lose to the color of th o Reliability for each light source is calculated from the number of each light source paxel, the brightness of the scene, and several rules which depends on each light source. The average of all paxel is also regarded as one o the light ht sources ' The color of the light source illuminating the scene is estimated as a eighted average of daylight, Fluorescent, Tungsten, the unidentified light Source as for the weights their reliabilities are used 1.2 White balance calibration In order to remove the variance for rgb sensitivities of cameras the white bal ance calibration should be done. It should be measured and calculated for each camera. And then, the white balance gain should be stored into the EEPROm of the camera The following is how to measure and calculate the white balalice gain, calR calg and cal B 6 1. Take a picture of the light box, whose color temerature is 5500K, with the pre-determined brightness 2. Calculate the average of linear RGB value for the pre-determined center area(magey i ath x imageleight Let aver. aveg and aveb denote them And t he maximum of them is denoted by aveM 3. Calcultate the white balance calibration with the following equatiOn ave calx= CoefF X X (X=R, G,B) Coefx is the coefficient in order to remove the variance of color for light boxes 1.3 White Balance mode This AWB has Auto-mode only 1.4 CFA Type The white balance gain calculated by this aWb algorithm is used only for Still Capture. CFA size for still Capture is the same regardless of bequarity mode or orientation 1.5 White balance system The Matt, BR, and MI3 AWb algorithm consists of the following four processes 1. Paxcl Generation 2. White balance evaluation 3. White Balance Compensation 4. Tone scale processin Definitions of thel are the followings. The details are shown froi the next sectlon 1.5.1 Definition of "paxel generation? Paxel means one of the blocks, which divides an image. For each paxel, RGB of the paxel is calculated by averaging pixels in the paxel with the manner explained later. Not only RGB value, but also two kinds of statistics are cal- ulated. And then the rgb of paxels are calibrated with the white balance calibration and the color-correction lllatrix. Finally, the color-corrected RGB of paxels is converted to LGI value, scaled T-space value. T-space is the 3 dimensional space introduced by the Kodak color printing system This paxel generation process is done at first. And then the generated paxel data(i.e. the LGi of pixels, and two kinds of statistics) pass the next process White balance evaluation 1.5.2卫 efinition of“ White balance evaluation” The process, White Balance EvaluatiOn, calculates the white balance gain frOll the paxel data. The white balance gain is used for the Image Processing Chain for Still Image 1.5.3 Definition of "White Balance Compensation" This process compensates white balance of an image with the white balance gain calculated by the white balance evaluation process. This process is done as a part of the Image Processing Chain 1.5.41 Dcfinition of " onc Scalc Proccssing This process compensates Tone Scale of an image with Kodak 120% Tone Scale This process is done as a part of the Image Processing Chain. I WB compensation and Tone Scale Processing are done simultaneously on Matt, BR and 2 Axel Generation 2.1 CFA Originally, Cfa stands for >Color Filter Al In this document, h CFA means an image, which consists of pixels with one of rGB corresponding to the color filter array. In this document, the color filter array is supposed to be a bayer pattern Figure 1: CFA Bayer Pa 2.2 Axel "Axel"is one of the blocks, which divide an image as Figure 1. The height and the width of a panel are denoted by pax height and pax width. and the number of pixels horizontally and vertically are denoted by paxNumh and paxNum V respectively. A paxel is indicated by [i, coordinate( Figure 2). The origin 10,0], indicates the upper and left corner paxel Figure 2. Paxels in cFa For Matt, BR, MI3, the parameters paxHeight, pax Width, paxNumH, and paxNum v have the following values on table arameter Value paxHeight 154(Matt,BR),193(M3) pax width 272( Matt. bR),346(MI3 paxNulmH panUm V 4 Table 1: Axel size and number of paxels 2.3 RGB of axel The rgb of a paxel means the average for RGb of pixels, which are sampled from the paxel at a certain sampling ratio unless the pixels are in saturated Bayer units. The Bayer unit means the three pixels as figure 3 shows GR Figure 3: Bayer Unit If any one of the three pixels in a Bayer unit is saturated the Bayer unit is called the“ saturated Bayer unit” For example, if all Bayer units, shown as gray in the Figure 4, are not saturated, the average of the following gray Red-pixels in the paxel is"R of the paxel. Likewise, Green and Blue of the paxel are calculated Figure 4: rGB of paxel Let paz ri,], paz G[i, j], and pae Bli, j mean the RGB of the paxel at the coordinate [i, j] respectively As for Matt, BR and mi3 the rGB of a pixel in a CFa is stored in 8 bit Non-Linear value. and the exact cfa value is converted from " Non -linear to

...展开详情
试读 52P 日本人白平衡AWB算法

评论 下载该资源后可以进行评论 1

yixiayiqiu 这个是2002年的文档,确实英文不行看不懂也不清楚这个算法有没有用
2020-07-16
回复
img
hddzxyhyj
  • 分享宗师

    成功上传21个资源即可获取

关注 私信 TA的资源

上传资源赚积分,得勋章
    最新推荐
    日本人白平衡AWB算法 31积分/C币 立即下载
    1/52
    日本人白平衡AWB算法第1页
    日本人白平衡AWB算法第2页
    日本人白平衡AWB算法第3页
    日本人白平衡AWB算法第4页
    日本人白平衡AWB算法第5页
    日本人白平衡AWB算法第6页
    日本人白平衡AWB算法第7页
    日本人白平衡AWB算法第8页
    日本人白平衡AWB算法第9页
    日本人白平衡AWB算法第10页
    日本人白平衡AWB算法第11页
    日本人白平衡AWB算法第12页
    日本人白平衡AWB算法第13页
    日本人白平衡AWB算法第14页
    日本人白平衡AWB算法第15页
    日本人白平衡AWB算法第16页

    试读已结束,剩余36页未读...

    31积分/C币 立即下载 >