The Optimization Algorithm Theory For Displaying Based On Higher Order Statistical Model

Although the current imaging sensors can achieve 12 or higher precision,the current display devices and the commonly used digital image formats are still only 8 bits.This mismatch causes significant waste of the sensor precision and loss of information when storing and displaying the images.Normally,the process of quantification not only affect visual quality but also cause wrong analysis to the image information.From the above discussion,we can come to a conclusion that the protection of the image details is very important,it is necessary to design a better image processing algorithm to protect the information and details in an image.After a long thought,we design the algorithm from two perspectives.First of all,we reexamine the image visualization algorithm from a statistical perspective.The entropy of an image is a important metrics,because the entropy represents the richness of the image information.We build the mathematical model on the image entropy and maximize the entropy in the process of image mapping to protect the image information and details.Considering that human visual system is more sensitive to contrast comparing with absolute illumination,we propose anther model for image visualization on image contrast.In this contrast based model,our objective is to minimize the contrast loss while the tone mapping from the HDR image to the LDR image.The tone-continuity is another aspect that we considered in the visualization.To suppress the large tone distortion artifacts that may caused by traditional visualization technique,we propose the tone-distortion constraint into tone mapping technique.We further discover that another benefacts of the proposed tone-distortion constraint is to reduce the computational complexity of the algorithm.The overall visualization process is finally models as a K-edges shortest path problem in a Directed Acyclic Graph,we propose a dynamic solution to solve the problem in polynomial time.Experimental results show that the proposed models outperforms the traditional approaches by exhibiting more subtle details in the foreground while preserving the smoothness of the background.