A monocular vision low-light image calibration method based on improved image enhancement

Abstract

The present application is directed to the problem that the feature extraction fails due to insufficient image brightness when monocular camera maps in low light environment. A kind of low light image calibration method based on improved image enhancement and monocular vision is disclosed, the method comprises: a camera vignetting light compensation strategy is proposed, the vignetting compensation function is analogous to response function, expressed as high-order polynomial, set the halo value of the center point, the non-linear compensation of vignetting effect is carried out, the image output brightness is uniform;On the basis of traditional image enhancement optimization function, initial brightness estimation value and fitting smoothness are used as constraint factor, pixel weight is added as error correction term, considering balancing two residual terms, an improved image enhancement calibration method is proposed, the corrected image is output, finally pose estimation is carried out. The present application can improve the image brightness of monocular camera when mapping in low light environment, so as to ensure the stability and accuracy of monocular vision SLAM front-end visual odometry.

Images