The Research Of Low Illumination Array Images Processing Technology

It is a research hotspot in the field of computer vision and image processing to directly obtain clear and useful information from low illumination image.Traditional methods be divided into two categories:spatial domain and frequency domain,RETINEX algorithm is one of the most popular algorithms in spatial domain.But the RETINEX algorithm needs to assume uniform illumination as the premise,and usually,to the different areas of an image with contrast brightness differences,the illumination image is not smooth and causes halo artifacts.So,the method that uses a single image to enhance the low illumination image still has a lot of defects.It is necessary to make use of fusing multiple images to enhance.The research of this subject is based on a 3*3 micro camera to study processing technology of low illumination image.The main work of this paper are as follows:(1)This paper introduces the present situation of the enhancement of low illumination image and the classical processing method,especially introduces the common methods of enhancement,sharpening,noise reduction and the evaluation standard of image quality.(2)Array images registration and fusion methods.Mainly introduces the image calibration method based on multi-scale feature,this paper see the fifth images as the center image to calibrate other images;describes the commonly used image fusion algorithm and using the weighted average fusion rules on the array images.(3)The method of image enhancement based on the micro camera and the comparison of the experimental results is carried out.Multiple scales RETINEX algorithm is used to achieve preliminary enhancement after getting array images,image sharpening and calibration image fusion are applied,then the noise reduction is applied and then enhancement by RETINEX algorithm on V channel is applied again.Finally,we obtain a clear image.Computer simulation experimental results show that the values of objective evaluation indexes ofSD,ENTR,AVEGRAD have been greatly improved compared with the values of multi scale RETINEX algorithm,it shows that the method in this paper can well compensate the defects of uneven illumination,can greatly inhibit the halo phenomenon,bitter retain the details of the original image,makes the enhanced image is more clearly.