| 主权项 |
1. A computer method for prioritizing image tiles comprising:
automatically creating by a computer a discretized representation of an image, the image including tiles, each of the tiles being mapped to at least one element of the discretized representation; automatically defining by the computer a set of action functions describing each action of navigation of the image, the action functions including at least one of a pan, a zoom, and a seek function; automatically describing by the computer a trajectory through the image based on the set of action functions; automatically computing by the computer probabilities of transition from one of the actions to another of the actions in the trajectory; automatically generating by the computer a prioritized list of the tiles based on the probabilities; maintaining a state set including a list of states, each of the states being created based on steady state probabilities and one of the actions from a previous of the states, the previous of the states based on the trajectory and containing a set of the tiles corresponding to a viewport and a set of possible of the actions for the state; calculating references to the discretized representation of the image based on the state, the set of tiles, and the action from the previous state; and preparing the prioritized list based on the state set; wherein the step of calculating the references comprises:
(a) setting the list of states to an initial state, the initial state including a tile set having references to the discretized representation of the image of the tiles in the viewport;(b) setting the probabilities of the list of possible actions for the state equal to the steady state probabilities;(c) traversing each of the list of possible actions of each of the states in the state set to locate an active state, the active state having the highest of the probabilities;(d) removing an action from the list of possible actions, the action corresponding to the highest of the probabilities;(e) adding a new state to the list of states; and(f) calculating a new state tile set and new state action set stochastically based on a video model, a user-interaction model, and an operating mode, wherein the step of calculating a new state the set comprises:
(1) computing the mean and the standard deviation of a probability mass function;(2) computing a first discretized representation of the image offset reference based on the discretized representation of the image, a viewport, and a set of panned pixels;(3) computing a second discretized representation of the image offset reference based on the discretized representation of the image, the viewport, and a view ratio zoomed;(4) marking each of the tiles in the tile list in the active states based on the actions corresponding to the tiles, the corresponding actions being based on the first discretized representation of the image offset reference, the second discretized representation of the image offset reference, the mean, and the standard deviation;(5) creating a new state based on the marked tiles; and(6) repeating steps (1)-(5) until the number of the marked tiles meets a pre-selected threshold. |
