Research Of Image Enhancement Based On Retinex Theory

With the development of image acquisition equipment and image processing technology,digital image has come to pervade every aspect of human activities. However, due to the limitation of image acquisition equipment and other factors, the collected images are generally unable to meet the requirements of eye observation or machine analysis. Therefore,to enhance the image is particularly important. In numerous image enhancement algorithms,the Retinex algorithm because of its color constancy and dynamic range compression characteristics is gradually becoming a research focus of image enhancement. In order to effectively improve quality and visual perception of the image, a comprehensive and deep research is conducted on Retinex algorithm, and the following two major research results are achieved.1.A new Retinex algorithm based on the YUV color space is proposed. Firstly, the image is converted from RGB color space to YUV color space. Then the brightness component Y is enhanced by using multi-scale Retinex algorithm. At the same time, chrominance components U, V are corrected based on Y component after enhancement. Finally, the image is converted to RGB color space to output. Results show that while image details are enhanced the color has been maintained. By comparison with other Retinex algorithms, experimental results show that the algorithm has achieved good results in image color fidelity and detail information enhancement respects.2.A Retinex low illumination image enhancement algorithm based on hue bilateral filter is proposed. Firstly, the illumination component of the image is accurately evaluated by a bilateral filter based on hue, thus obtaining the reflection component of the image. Then the reflection component is stretched and compressed by a non-linear adaptive correction function.Finally, the color of the image is restored by a color restoring method which could maintain the hue invariant. The experimental results show that the algorithm not only effectively eliminates the halo, but also achieves good effects in detail enhancement and color restoration,compared with other algorithms.