The effects of attention mechanisms on low light image enhancement

The clarity of the input images, which will significantly increase the success of computer vision studies, cannot be captured because the cameras cannot access the deseired points of natural scenes. Just as the scenes they capture when the exposure times of the cameras are kept short are dark, when they are kept long, the image clarity cannot reach the desired high dynamic range due to too much light. The dark scene obtained by keeping the exposure time short is both unclear and has unwanted noisy. The too bright scene obtained by keeping the exposure time long becomes problems caused by shaking in the hand. Low-light image enhancement algorithms are used to solve the problem of limited dynamic range of cameras. The main purpose of this thesis is to measure the effects of attention mechanisms to improve low-light images. For this purpose, a low-light image enhancement method using different attention mechanisms has been proposed based on the SSIENET method, which has shown successful results in the literature. It has been observed that more successful results are obtained when different attention mechanisms are included in the low-light image enhancement study existing in the literature.