Enhanced LBP algorithm and its application in the rotation invariant image retrieval

Disclaimer: This article is a blogger original article, follow the CC 4.0 BY-SA copyright agreement, reproduced, please attach the original source link and this statement.
This link: https://blog.csdn.net/qq_42111606/article/details/102689104

1, a so-called gray scale invariance refers to the illumination change whether the impact will be described, so there is such a strong light above a certain pixel, this value should be 9 will increase, but the relative size relationship is still It will be encoded as a binary image such as the right.
2, rotation invariance: rotation of the image will get different of LBP values
. 3, LBP rotationally invariant pattern, texture pattern portion discarded ? ? ?
4, centrosymmetric LBP ???

LBP equivalent mode

Investigation LBP operator defines understood, a LBP operator may produce multiple binary pattern (p sampling points) as: 3x3 neighborhood have p = 8 sampling points can be obtained 2 ^ 8 = 256, 5x5 neighborhood have P = 24 sampling points, can be obtained 2 ^ 24 = 16777216 kinds binary mode, and so on .... Obviously, too much binary mode both for extraction or texture texture identification, classification and information access are disadvantageous in practical applications requires not only the operator uses as simple as possible, while also taking into account the computing speed, storage capacity size and other issues. It is necessary to reduce the dimensions of the original LBP mode.

Ojala proposed a "equivalent mode" (Uniform Pattern) to dimensionality reduction, Ojala et LBP operator of that image, repeating binary number corresponding to a local binary pattern from 0-> 1 or 1-> 0 , up to jump twice, the local binary pattern corresponding to an equivalent binary pattern becomes. The 00000000,00111000,10001111,11111111 other pattern classes are equivalent.

Determining whether a simple binary pattern mode is equivalent approach is to LBP value and its value after moving a cycle number for the bitwise binary number, calculated in 1, if the number is less than or equal to 2, is equivalent mode; otherwise, no. In addition to the property of the equivalent model of eleven classes mode, referred to as mixed mode type, e.g. 10010111 (a total of four jumps).

Calculation hopping: such as 10010111, firstly the first two 10, from 1-> 0 becomes a jump; second and third 00, no jump; third and fourth 01, a 0-> 1 transition a fourth, five 10, from 1-> 0 transition time; fifty six 01, a 0- & g

Guess you like

Origin blog.csdn.net/qq_42111606/article/details/102689104