CNN shape segmentation advantages in MPEG-4 simple profile encoding
Lauri Koskinen, A. Paasio, Kari A. I. Halonen · 2003
The MPEG-4 core profile offers content-based functionalities. It is shown here that content-based encoders also have advantages in simple profile coding. An intermediate phase of a cellular nonlinear network (CNN) shape segmentation algorithm indicates the area containing motion within a video frame. This information can be used to reduce the required computational power requirements of motion estimation. Depending on the amount of motion in an sequence significant savings in coded bits can be achieved without loss in subjective video quality. The number of computed SADs also shows considerable decrease. The method is simulated with full search and MVFAST but is motion estimation algorithm independent and has increased advantages in MPEG-4 AVC.