Image feature extraction method under variable illumination

Abstract

The invention discloses an image feature extraction method under variable illumination. A double-layer structure model is adopted when a WLSGP characteristic value of a central pixel point is calculated. Different weight coefficients are distributed according to different degrees of influence of different neighborhood radii on central pixel values, the influence of inner and outer neighborhood pixels is not considered for a differential excitation component of an original WLD operator, and illumination sensitivity is avoided by adopting the differential cooperative excitation component. A LOGoperator with variable scale and variable angle is introduced into differential collaborative excitation. An ADEP operator and a WLSGP operator are adopted to replace a differential excitation component and a direction component in an original WLD. The problem that differential excitation of an original WLD operator is not suitable for a direction difference occasion is solved. Differential excitation images of each block image with the respective gradient direction image are fused to obtain a two-dimensional AWSGD histogram. Image classification is performed on the converted one-dimensional histogram by adopting a selective XGBoost classifier to obtain an identification result, so that the invention can show relatively good performance under the condition of changing illumination.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to an image feature extraction method under changing illumination. Background technique [0002] Image feature extraction technology is one of the main research topics in the field of computer vision, and has been widely used in biometrics, image retrieval, object detection and other fields. Under different feature extraction and classification recognition methods, the performance of image recognition technology will change greatly under conditions such as different viewing angles, changing illumination, and partial occlusion. Compared with other conditions, the challenges brought by changing illumination to recognition accuracy are the most significant, mainly including insufficient illumination (too dark), overexposure (too bright), image shadows and other factors. [0003] Scholars at home and abroad have conducted a lot of research on the description techno...

AI Technical Summary

Problems solved by technology

However, the WLD operator has several problems when describing image features: ① WLD only considers the influence of single-layer pixels in the neighborhood of the central pixel, and does not consider the gray information of outer pixels

The reason is that WLD uses an isotropic filter to calculate the differential excitation components, resulting in positive and negative offsets between the central pixel and neighboring pixel differences when summed, so effective gray information cannot be obtained

③ When the WLD operator calculates the gradient component, it only considers 4 points in the parallel and vertical directions, which cannot fully reflect the spatial distribution information of the image

Images