Image Enhancement Algorithm Based On Improved Threshold Function And Bilateral Filtering

Retinex algorithm is a widely used algorithm in the field of image enhancement.Through the estimation of the light image,a reflective image reflecting the nature of the image is obtained,which avoids the influence of light on the image content.Wavelet threshold function denoising method plays a main role in wavelet denoising.Because of the constant error of soft threshold function and the discontinuity of hard threshold function at the discontinuity point,the image is easy to vibrate after denoising,which affects the image denoising effect.To solve these problems,this paper proposes an image enhancement algorithm based on improved threshold function and bilateral filtering.Firstly,the image is decomposed by wavelet to obtain the low frequency and high frequency coefficients.Secondly,the modified two-side filtering Retinex algorithm is adopted to process the image's low-frequency coefficient,which can better remove the halo phenomenon by eliminating the influence of the grayscale value of the pixel far from the center point.The improved threshold function is adopted to deal with the high frequency coefficient,and according to the continuity,progressivity and deviation of the improved threshold function,the oscillation phenomenon and constant error are avoided.The enhanced reconstructed image is obtained by inverse discrete wavelet transform.Finally,the reconstructed image is transformed into three segments of piecewise linear transformation to enhance the image contrast and avoid the phenomenon of gray.In the experiment,the validity of the proposed algorithm is verified from three categories of images,namely foggy images,inhomogeneous images and low-illumination images.The experimental results show that the proposed algorithm avoids the phenomenon of gray,halo and oscillation,and effectively removes the noise.The enhanced image has rich details and strong contrast,which lays a foundation for the subsequent image processing.This paper contains 30 pictures,11 tables and 55 references.