Research On Image Enhancement Algorithms Based On Human Visual Characteristics

It is significant for human eyes or machines to study related image enhancement techniques in order to recognize information effectively from low quality images.So far most image enhancement algorithms are based on image features,taking little account of the visual characteristics of human eyes.Applying some features of human vision system to image enhancement can control the degree and improve the quality of image enhancement better.In view of the shortcoming that traditional gray transform algorithms can't make full use of gray level so that visual information loss may be easily caused,this paper proposes an algorithm which merges the gray scale of low brightness and high brightness region based on lightness threshold characteristics of human eyes and then stretches it.After processing,the contrast of the middle gray region of the image is improved and the visible information of low brightness and high brightness region can be reserved well.Traditional nonlinear transform algorithms can enhance the low luminance region of low illumination images well,but often inhibit the bright region.This paper proposes a nonlinear transform algorithm which can stretch or compress gray scale more rationally on account of different luminance.After processing,the contrast of low brightness region of the image can be improved and the contrast of high brightness region can be reserved well considerably.In order to improve the local contrast of images,an algorithm which merges invisible pixels and adjust visible pixels based on adjacent pixels as processing units is proposed.After processing,the contrast of the image is improved hugely and the entropy is close to the original image.Finally,this paper proposes an improved retinex algorithm based on wavelet transform which uses low frequency component of wavelet transform as illumination estimation and fuses high frequency coefficients of enhanced image and original image.In comparison with traditional retinex algorithm,the average brightness and contrast of images processed by the proposed algorithm is higher,the entropy is close,and the execution time is shorter.In addition,the adjacent pixels merging stretching algorithm is used to adjust the reflection component.After processing,the contrast of the image is further improved.