Research On Image Enhancement Algorithm Under Different Light Conditions

In an environment where the lighting conditions are not ideal,the quality of the captured image may be degraded,resulting in a decrease in the performance of the related image processing algorithm,so image enhancement is a very necessary research direction.For underexposed images,most of the existing algorithms cannot enhance the intensity according to the adaptive selection of image quality,resulting in excessive image enhancement or insufficient enhancement.In addition,in the case of severely insufficient lighting at night,how to make the image enhanced while reducing the distortion of the image as much as possible is a difficult task.In response to the existing problems,we carried out the following two aspects of research work.(1)For the underexposed image,this paper proposed a low illumination image enhancement algorithm based on image fusion technology,the purpose of which is to make better use of the useful information of each part and reduce the loss of feature information.Due to the effect of the fusion processing,the key lies in the choice of the weight matrix.Therefore,in order to better preserve the details of the clear part of the input graph,our weight selection is related to the luminance map of the input graph.Next is the enhancement of the image exposure rate.The adaptive threshold method is used to screen out the pixels with insufficient brightness in the input image,which is used as an estimate of the optimal exposure rate.The optimal exposure is then applied to the camera response model to obtain an image with enhanced exposure.In addition,in order to make the enhancement algorithm have better robustness,we have constructed different fusion schemes for images with different exposure conditions.For images with less exposure,we use the original image to blend with the enhanced exposure image.For images with severely underexposed images,we use a reflection map based on Retinex to blend with an image with enhanced exposure.Experiments show that this method reduces the distortion of image brightness while enhancing the image.(2)In view of the serious lack of illumination at night,a multi-layer model image enhancement algorithm was proposed in this paper.A structure-aware filter is used to obtain a luminance map(high-frequency component)of the image,thereby obtaining a reflection map(low-frequency component).Since the brightness map of the image has more image brightness order information and partial structure information.Therefore,in order to preserve the luminance information,we obtain the first layer luminance map after being corrected and merged into the multi-layer model.Each layer of luminance map can be decomposed into two parts:a high frequency component and a low frequency component.Finally,the final brightness map and the corrected brightness map are combined with the multi-layer reflection map to obtain the final enhancement result.The model is far superior to the classic multi-layer model in terms of runtime,and it is also superior to many night image enhancement algorithms in terms of enhancement.The image enhancement methods in different environments proposed in this paper were tested with the public data with the comparison map and the data collected in different environments.The experimental results show that the proposed method has better robustness for images under different conditions,and the visual effect of the enhanced image is better.