On Finding Gray Pixels

01/09/2019
by   Yanlin Qian, et al.
16

We propose a novel grayness index for finding gray pixels and demonstrate its effectiveness and efficiency in illumination estimation. The grayness index, GI in short, is derived using the Dichromatic Reflection Model and is learning-free. The proposed GI allows estimating one or multiple illumination sources in color-biased images. On standard single-illumination and multiple-illumination estimation benchmarks, GI outperforms state-of-the-art statistical methods and many recent deep net methods. GI is simple and fast, written in a few dozen lines, processing a 1080p image in about 0.4 seconds with a non-optimized Matlab code.

READ FULL TEXT

page 1

page 4

page 5

page 8

research
03/22/2018

Dichromatic Gray Pixel for Camera-agnostic Color Constancy

We propose a novel statistical color constancy method, especially suitab...
research
06/30/2014

Pixel-wise Orthogonal Decomposition for Color Illumination Invariant and Shadow-free Image

In this paper, we propose a novel, effective and fast method to obtain a...
research
02/27/2019

Flash Lightens Gray Pixels

In the real world, a scene is usually cast by multiple illuminants and h...
research
03/22/2021

Polarization Guided Specular Reflection Separation

Since specular reflection often exists in the real captured images and c...
research
02/06/2021

Sill-Net: Feature Augmentation with Separated Illumination Representation

For visual object recognition tasks, the illumination variations can cau...
research
08/06/2019

Fast Fourier Color Constancy and Grayness Index for ISPA Illumination Estimation Challenge

We briefly introduce two submissions to the Illumination Estimation Chal...
research
06/13/2019

Illuminant Chromaticity Estimation from Interreflections

Reliable estimation of illuminant chromaticity is crucial for simulating...

Please sign up or login with your details

Forgot password? Click here to reset