Finite state DCT with global bitmap image coding
Yung-Gi Wu, Shen‐Chuan Tai · 2002
The ordinary DCT coding techniques always classify the image into 4 classes, then select suitable bitmaps for the 4 classes. This is not enough for complicated images. In order to encode the image in more detail, the authors classify the image into 16 classes. This classification can reduce the artifacts which are caused by the ordinary coding scheme. However, it will take a lot of time to generate the 16 bitmaps for the 16-class classification. Besides, it will take a very large overhead to transmit the 16 bitmaps. In order to solve the deficiency, the authors developed the concept of global bitmaps of 16 classes which can be used for all the images. As to the side information of the classification, the authors use the idea of the finite state machine to predict the current class from the previously classes. Thus, the overhead of the class will be reduced. The experimental results show that the performance of the coding system is very good and the prediction correct ratio can be over 70% in the Lena image.>