ObserverSensor Eye Response Camera Response Reflected light spectrum
Observer/Sensor Eye Response Camera Response Reflected light spectrum is represented by a 3 element vector ・人間の目とカメラセンサーの色知覚が異なる ・反射光のスペクトルがRGB 3次元のベクトルより表現
CIE色空間 u CIEは“imaginary”光を三つ(X, Y, Z)定義 X Y Z = 0. 490 0. 310 0. 200 0. 177 0. 813 0. 011 0. 000 0. 010 0. 990 R G B
明るさと独立した色度平面内のRGB Same Color, different brightnesses Colour Cube Chromaticity Plane 色度平面
HSV色空間 Saturation Hue Value
Color Spaces Models Applications Colorimetirc XYZ(基準空間) Colorimetric calculations Deviceoriented Storage, processing, analysis, coding, color TV Useroriented Munsell u. Non-uniform spaces RGB, YIQ, YCC, . . . Color difference u. Uniform spaces L* a* b*, L* u* v*, . . . evaluation, analysis, color HSI, HSV, HSL, I 1 I 2 I 3, . . . management systems Human color perception, computer graphics Human visual system
RGB to YIQ・YUV u YIQ色空間(カラーテレビ、Used in NTSC: National Television Systems Committee) u Y:明るさ, I & Q:色 (I=red/green, Q=blue/yellow) • Y = 0. 299 R + 0. 587 G + 0. 114 B • I = 0. 596 R - 0. 275 G - 0. 321 B • Q = 0. 212 R - 0. 528 G + 0. 311 u YUV色空間(デジタルビデオカメラ、1982 • Y = 0. 299 R + 0. 587 G + 0. 114 B • U = 0. 492(B – Y) • V = 0. 877(R – Y) standard)
Dichromatic Reflection Model (2色性反射モデル)ー絶縁体材料 Incident Light Body Reflected Light (object color) Material surface Surface Reflected Light (light source color) Absorption(吸収) Scattering(散布) Colorant(着色剤) 物体からの反射光が2つの反射成分の組合わせで記述できることを仮定している
Example Reflectance of Skin From [Anderson and Parrish]
同じ素材のRGBは同じ平面に存在 ●For small a the camera RGB will be in the ‘body’ direction B RGBbody+a RGBsurface R RGBsurface G ●As a becomes large so the RGB moves towards the ‘surfface’ direction
2色平面(Dichromatic Plane) B RGBbody 反射されたRGBは RGBbody+a RGBsurface 必ず“body”RGBと “surface”RGBの間にある R 2 面 平 色 RGBsurface G RGBs outside this plausible (striped) region can only occur if we have negative body or surface contributions. We cannot since light is a positive quantity
Shape From Shading : source brightness : surface albedo (reflectance) : constant (optical system)
画像の色が光源色に依存 Macbeth color checker タングステン電球 Tungsten 白色蛍光灯 Daylight Redder
光源色の推定 Illuminant Estimation ? Color Correction Macbeth color checkerと比較
Estimating the Illuminant Color || Computational Color Constancy 光源色の推定 || 色恒常性の計算
Example of Application Color Correction Estimation error 1. 6° von Kreis Transformation Image taken under Re-rendered image Canonical image to canonical cond. at CCT = 3200 K CCT = 4700 K 標準条件の画像
Color Image Segmentation u Pixel-based techniques(画素) u Region-based techniques(領域) u Edge-based techniques(エッジ) u Stochastical Model-based techniques(確 率的) u Physics-based techniques(物理) u Hybrid techniques(混成) u Example modelling skin and segmenting
Color Clustering (or Segmentation) by K-means Algorithm Image Clusters on intensity Clusters on color
Color Clustering (Segmentation) by K-means Form K-means clusters from a set of n-dimensional vectors 1. Set i (iteration count) to 1 2. Choose randomly a set of K means m 1(1), …, m. K(1). 3. For each vector xi, compute D(xi, mk(ic)), k=1, …K and assign xi to the cluster Cj with nearest mean. 4. Increment i by 1, update the means to get m 1(i), …, m. K(i). 5. Repeat steps 3 and 4 until Ck(i) = Ck(i+1) for all k. Original RGB Image Color Clusters by K-Means
- Slides: 58