Research On Image Enhancement Algorithm Based On Global And Local Fusion

Images collected under complex illumination environments tend to have insufficient local contrast,and there are large dark areas and some bright areas.The effective information is masked,resulting in no practical application value.At this time,it is necessary to perfonn operations such as correcting brightness,improving contrast,and highlighting details of the low-quality image,which is also called image enhancement.In the image enhancement algorithm,whether it is the typical gradation transformation method or the Retinex algorithm based on the human retina theory,it is necessary to enhance the local detail to improve the clarity while improving the original enhancement algorithm.Therefore,image enhancement algorithm based on the fusion of global enhancement and local detail enhancement is proposed.The main research contents are as follows:(1)The typical gray-scale transformation theory and its improvement are studied,including gamma correction,logarithnic transformation and histogram correction,and the incomplete Beta function which can fit various nonlinear transformation curves is derived.On the basis of the more flexible non-complete Beta transform,aiming at the shortcomings of the original control parameter estimation method,an adaptive estimation parameter using image histogram statistical information is proposed.Thus,a hybrid histogram modification function(ABTWD)combining combined incomplete Beta transform and histogram equalization method is derived.The mixed histogram modification function(ABTWD)achieves the purpose of correcting image brightness and enhancing contrast.At the same time,for the deficiencies of the gamma transformation,the detail layer information extracted by the unsharp mask is obtained based on the Beta transform.Naturally clear enhanced image.Compared with other typical grayscale transformation algorithms,this algorithm can effectively correct brightness,stretch contrast and enhance sharpness for low quality images.(2)Aiming at the problems of color distortion,halo artifact and lack of detail in the typical Retinex algorithm,an image enhancement model based on illuminance estimation with rolling guidance filter and multi-scale detail ethancement is proposed.First,the independence of the luminance and color components in the HSV color space is firstly utilized,and the Retinex algorithl is globally enhanced for the luminance component to avoid color distortion.Second,The problem is the illumination estimation of the current Retinex algorithm is inaccurate,resulting in insufficient detail and blooming artifacts.Thus,the rolling guidance filter(RGF)illuminance estimation algorithm is proposed,which can accurately estimate the illumination by using the ability of RGF to accurately preserve the edges of large structures and remove small textures,and then avoid over-enhancement with gamma correction to obtain local contrast enhancement.Brightness channel enhancement map without halo artifacts;Third,the final conversion back to the RGB space for multi-scale detail enhancement,multi-scale details including rough detail,medium detail and fine detail to further improve image clarity.Compared with other advanced Retinex algorithms,this algorithm effectively eliminates image whitening and halo artifacts,and has rich details and natural distortion.