The Research On Low Illumination Image Enhancement And Denoising

In the era of digital information, the images have become an important medium for acquiring and transmitting information. However it is often because of the insufficient exposure, the backlight shooting and the other reasons that make the images captured by the electronic equipments in reality appear to be of low brightness, low contrast, narrow dynamic range and lots of noise. This seriously affects the recognition of the images. Therefore, it is very important to study the enhancement and denoising of the low illumination images. In this paper, for the characteristics of low brightness, low contrast and lots of noise existing in the low illumination images, we propose the corresponding algorithms to enhance the low illumination images and remove noise. The main work of this article is as follows:(1)The existing low illumination image enhancement algorithms are studied. Then, a new HDR(High Dynamic Range) image enhancement algorithm based on gray scale transform and detail enhancement is proposed by analyzing the low illumination images from the point of global information and local information. Using adaptive gray level mapping, we rapidly increase the brightness in dark image areas. At the same time, the dynamic range of the image is compressed in the bright regions. Furthermore, we enhance the local contrast of the images by the high frequency details. The enhanced image details are clear with natural color.(2)We discuss the image enhancement algorithms based on retinex and analyze the reasons such as the drawbacks of partial color, over enhancement and halo phenomenon that the traditional retinex algorithms exist. A new retinex image enhancement algorithm based on domain filtering is proposed. According to the accurate estimation of the illumination image by domain filtering, the brightness of the illumination image is improved adaptively. Thus, the enhanced brightness image is obtained. The local contrast of the image is enhanced by the CLAHE(Contrast Limited Adaptive Histogram Equalization) algorithm. The proposed algorithm overcomes the shortcomings of partial color, over enhancement in bright regions and halo effect existing in the traditional retinex algorithms. The proposed algorithm is also applied to night haze image enhancement. The enhanced image details are clear with natural color and contain less noise. In addition, the algorithm can effectively overcome the partial color phenomenon.(3)We analyze the characteristics of the noise under the low illumination condition. A denoising method based on spatial and frequency filtering is proposed. The guided filtering is used to spatially remove most of the high frequency noise. Then the image is decomposed by wavelet transform. We use the guided filtering to remove the noise existing in the approximation of the wavelet coefficients. The vertical partial, the diagonal partial and the horizontal partial of the wavelet coefficients are handled by adaptive wavelet threshold denoising to remove some high frequency noise and most of the low frequency noise. The filtered image is clear and can retain the more edge and details of the image.Finally, the low illumination image enhancement algorithm is implemented on the android platform. We design the software to enhance the low illumination image. After optimized, the app software can almost achieve real-time.