Brief analysis of the classic Retinex algorithm

Retinex is a commonly used image enhancement method based on scientific experiments and scientific analysis. It was proposed by Edwin.H.Land in 1963 . Just as Matlab is composed of Matrix and Laboratory , Retinex is also a word composed of two words, retina  and cortex , namely: retina and cortex. Land 's retinex model is based on the following three assumptions:

(1) The real world is colorless, and the color we perceive is the result of the interaction of light and matter. The water we see is colorless, but the water film - soap film is colorful, which is the result of light interference on the surface of the film.

(2) Each color area is composed of three primary colors of red, green and blue with a given wavelength;

(3) The three primary colors determine the color of each unit area.

The basic theory of Retinex theory is that the color of an object is determined by the reflection ability of the object to long-wave (red), medium-wave (green), and short-wave (blue) light, rather than the absolute value of the reflected light intensity. The color of the object is not affected by the non-uniformity of illumination, and has consistency, that is, the retinex is based on the consistency of color perception (color constancy). Different from the traditional linear and nonlinear methods that can only enhance a certain type of image features, Retinex can achieve a balance in three aspects: dynamic range compression, edge enhancement and color constancy, so it can automatically perform various types of images. Enhanced adaptation.

For more than 40 years, researchers have imitated the human visual system to develop the Retinex algorithm, from a single-scale Retinex algorithm to a multi-scale weighted average Retinex algorithm, and then to a color restoration multi-scale Retinex algorithm.

1. Single Scale SSR (Single Scale Retinex)

A given image S(x,y) can be decomposed into two different images: reflection image R(x,y) and luminance image (also called incident image) L(x,y) , the principle is as follows As shown in the figure:




Guess you like

Origin http://10.200.1.11:23101/article/api/json?id=326786373&siteId=291194637