An improved self calibration image enhancement algorithm based on attention mechanism and its application in low illumination scenes at sea

Objectives In order to solve the problems of insufficient brightness enhancement, low sharpness, and color distortion of existing maritime low-light image enhancement algorithms, a maritime low-light image enhancement algorithm based on improved self-calibrated light learning is proposed. Methods On the basis of the self-calibrating light learning algorithm, the unevenly illuminated regions in low illumination images are enhanced to different degrees by introducing an attention mechanism; the illumination adjustment module is constructed to secondary the intermediate outputs of the light learning process; the denoising module is introduced to ameliorate the problem that the noise in dark regions will be amplified with the enhancement of the brightness; and the batch normalization (BN) is changed to batch channel normalization (BCN) , which normalization method utilizes channel and batch dimensions to adaptively combine the normalized outputs. Image quality is evaluated by both subjective and objective aspects. Results Experiments were conducted under three test sets, and the results show that the improved algorithm not only improves the image brightness, but also enhances the results with higher color richness and no color distortion, and improves the standard deviation by an average of 20.01%, reduces the natural image quality evaluation by 9.16%, and improves the average gradient and information entropy by 23.68% and 6.46%, respectively, when compared with the original unimproved algorithm. Conclusions The improved algorithm has made a breakthrough in image visual quality, enabling better enhancement of maritime low-light images in different environments.