Research On Denoising And Enhancement Algorithm For Low Illumination Image

In the digital,intelligent information age,the image has become a main way to get,message and exchange information.However,under the condition of low or uneven illumination,such as greasy weather,nighttime,indoor and so on,because of a lack of light,and inadequate exposure,the collected still images or video images appear much noise and low luminance or contrast.Meanwhile,the identification of image detail is very difficult.So the low-light image denoising and enhancement have significance of application.The traditional image denoising and enhancement algorithm uses mostly single frequency and space domain,the image denoising and enhancement results are often unsatisfactory.In this paper,the existing frequency domain and space domain algorithm is modified.Then the low-light image denoising and enhancement improved algorithm is proposed.The concrete contents and results are as follows:First of all,it exposes the research background,significance and situation of low-light images denoising and enhancement algorithm.Meanwhile overseas and domestic research status on low-light images denoising and enhancement algorithm are analyzed.The key technology and technical route involved in this paper are proposed.The related theory on low-light image denoising and enhancement algorithm is detailedly analyzed and discussed.The advantages and disadvantages of traditional denoising and enhancement algorithm are compared.Then it puts forward to related parameters that evaluate the quality of the image denoising and enhancement.Secondly,the paper analyzes the hard or soft threshold low-light image denoising algorithm based on wavelet transform : because the hard threshold function is discontinuous in the threshold,it gives rise to image signal oscillation.Compared with the hard threshold function,soft threshold function has a better continuity in the overall threshold and processing of reconstructed image signal is relatively smooth.However,in the reconstructed image signal,the soft threshold decreases the approximation degree of the image,and results in distortion such as extra edge blur.At present,the hard or soft threshold function is put forward by the researchers.Although the compromise threshold between the hard threshold and soft threshold is used,compared with the traditional thresholding function,it overcomes s the disadvantages of hard or soft threshold low-light image enhancement algorithm,butthe image noise and processing has certain dependence,there was some fuzzy edge in image denoising.This paper puts forward a kind of bilateral filter and wavelet transform to the low-light image enhancement algorithm.The bilateral filter is adopted to the input of low-light image enhancement algorithm,and remove the vast majority of the high frequency part of the noise.The wavelet transform is adopted denoising image to denoise further.The image is decomposed into myopia part,vertical,horizontal and diagonal parts.Then the bilateral filter is adopted to myopia part of image,and the wavelet transform denoising is applied to vertical,horizontal and diagonal parts.Finally the denoised image is obtained by inverse wavelet transform.The MATLAB simulation show that the algorithm of gaussian noise,salt and pepper noise and mixed noise has better denoising effect.Secondly,The Retinex theory model based on human visual luminance and color perception is elaborated.The model has high color constancy,color fidelity,large dynamic range compression and so on.The algorithm can have a good enhancement effect to low-light images.Although low-light images by a single scale Retinex algorithm have good color fidelity,but the enhancement effect is unsatisfactory for dark area.The processing result has a certain color distortion.Aiming at the shortcomings of these algorithms,a improved multiscale Retine low-light image enhancement algorithm is proposed.This algorithm can adaptively adjust brightness and saturation for low-light image.It has a bigger superiority to enhances the darker part for image,and image enhancement effect is obvious.The Simulation experiments showed that the algorithm can have a good enhancement effect by improving the brightness of dark areas.In the meantime,it overcomes the color distortion,low contrast,details fuzzy,etc.Finally,we summarize the major research contents and achievements,point out the deficiency and the research direction.