Normalization Modifications for Fast Self-Quotient Image Method

Fast Self-Quotient Image (FSQI) method, which was introduced by the present authors in 2018, is a modification of Self-Quotient Image (SQI) method introduced by Haitao Wang, Stan Z. Li et al. in 2004. The main difference between these methods is that FSQI method uses circularly shifted Gaussian kernel instead of locally weighted Gaussian filter kernel used in the original SQI method. This allows increasing computation performance of FSQI and improving lighting normalization quality, especially in very dark shadow regions. However, the recognition quality after the lighting normalization using FSQI is not perfect and still require improvements. In this paper, we introduce two modifications of the normalization step for FSQI and compare their quality with the original FSQI method.