Color illusion and histogram equalization

第13回錯覚ワークショップ　明治大学中野キャンパス６階セミナー室３
2019年2月26日（火）16:30-17:00

北岡明佳（Akiyoshi KITAOKA） email
立命館大学総合心理学部（Ritsumeikan University, Psychology）

2019/2/20 より配布物　抄録

"Akiyoshi Kitaoka and Kyoro-chan (mascot of a chocolate snack)"

The head of the bird appears to be reddish, though the pixels are not.

森永チョコボール

"Akiyoshi Kitaoka and Kyoro-chan 2"

My face appears to be reddish, though the pixels are greenish.

"Akiyoshi Kitaoka and Kyoro-chan 3"

The beak of the bird appears to be yellowish, though the pixels are greenish.

A tomato appears to be reddish, though the pixels are not.

Signboards appear to be reddish, though the pixels are grayish in the upper image or yellowish in the lower one.

「RGBで白と黒」

Eyes appear to be reddish, yellowish, greenish and bluish, though they are gray as shown below.

左から、赤・黄・緑・青の目に見えるが、ほぼ同じ灰色である。

加算的色変換による色の恒常性
Additive color change and color constancy

(γ = 1.2, α = .48)

Strawberries appear to be reddish, though each pixel is cyan hue.

(sRGB, α = .48)

「赤く見えるいちご」

Strawberries appear to be reddish, though each pixel is cyan hue (luminance-based transformation).

すべてのピクセルはシアン色近辺の色相であるが、イチゴは赤く見える。加法色はシアンで、透明度（アルファ値）は48%の加法的色変換。

How to make a transformed image with alpha blending

(with a CIE diagram)

(Transformed images with a variety of alpha values)

Strawberries appear to be bluish, though each pixel is yellow hue.

(γ = 1.2, α = .48)

Strawberries appear to be bluish, though each pixel is yellow hue.

(sRGB, α = .48)

「青く見えるいちご」

Strawberries appear to be bluish, though each pixel is yellow hue (luminance-based transformation).

すべてのピクセルは黄色の色相であるが、イチゴは青く見える。加法色は黄色で透明度は48%の加法的色変換。

(with a CIE diagram)

Any color works.

(Transformed images with a variety of alpha values for the six color images)

３種類の色の恒常性画像
Three different color changes to show color constancy

The hypothesis of "histogram equalization" proposed by Shapiro et al. (2018) can explain these phenomena.

Shapiro, A., Hedjar, L., Dixon, E., and Kitaoka, A. (2018). Kitaoka's tomato: Two simple explanations based on information in the stimulus. i-Perception, 9(1), January-February, 1-9. PDF (open access)

additive color change↓　　↑histogram equalization

multiplicative color change↓　　↑histogram equalization

color change with the two-color method↓

histogram equalization↓　　↑multiplicative color change

How to make an image with the two-color method

In image E, the tram appears to be bluish, though all the pixels are reddish.

静脈が青く見える錯視
Vein color illusion

Veins appear to be bluish, though the pixels are not.

「静脈錯視：腕」

この手の青く浮き出て見える静脈は物理的には青くなく、彩度の低いオレンジ色であった。

Thanks to MK for your tremendous contribution!

「静脈錯視：腕２」

青白く撮れた写真の場合でも、この手の青く浮き出て見える静脈は物理的には青くなく、彩度の低い黄色かオレンジ色であった。

histogram equalization↓　　↑additive and multiplicative color changes

静脈画像化アルゴリズム（「ヒストグラム圧縮」）
Color change to vein color or "histogram compression"

「青く見える電車」

The train cars appear to be bluish, though the pixels are yellowish.

すべてのピクセルは黄色の色相（一部灰色）であるが、電車の車体は青く見える。

histogram equalization↓　　↑histogram compression

Abstract

The color illusion produced by additive color change can be explained with the histogram equalization model proposed by Shapiro, Hedjar, Dixon and Kitaoka (2018, i-Perception). Here I apply this model to the two different types of spatial color mixture (additive color mixture and subtractive one) and related color illusions (Munker illusion, neon color spreading, etc.) and discuss color perception.

おしまい　

北岡明佳の錯視のページ