Research On High Dynamic Range Image Enhancement Algorithm Based On Neural Mechanism Of Visual System

Natural Scenes usually contain information of quite high dynamic range. High dynamic range image(HDRI) is widely used since it is capable of capturing more fine information. However, problems remain in its display. The human eye has nice adaptation for high dynamic range scenes, which motivates us to develop new methods for HDRI displaying. In 2006, Laurence Meylan and Sabine Süsstrunk proposed an algorithm for HDRI’s display. Meylan’s method is characterized by using PCA as a color model and hiring the retinex-based adaptive filter to make the local process. In my article, We firstly prove that the global logarithmic mapping on all of the R,G,B channels results in desaturation. Then we prove mathematically that there are serious problems in using PCA as a color model and encoding all of RGB channels into logarithmic fields. According to this, we then propose improved models for brightness and chromaticity respectively. Keeping the chromaticity not to change, we use a global transformation and the retinex-based adaptive filter only on the brightness channel. We finally change them back to RGB fields after combine the new brightness and old chromaticity together. Our chromaticity model does well in keeping the chromaticity. We make the global mapping on only the brightness channel and this is a good way to avoid desaturation which occurred in Meylan’s method. What’s more, our color model ensures a good independence between brightness and chromaticity. By applying our improved method, the details in both dark areas and bright areas can be well displayed with better hue and saturation.