Apple's patent application, titled "Video transmission using content-based frame search," looks to deal with dropped frames in a FaceTime video call, a problem some iPhone or iPad users may encounter when communicating over a low-bandwidth connection.
Currently, video communication over cellular data is spotty in many areas due to bandwidth restrictions and existing wireless technology. In some cases, features like Apple's FaceTime are nearly unusable due to dropped frames, extremely low-resolution images and poor audio quality.
Detailed in the filing is a method in which previously transmitted video frames and "video frame content information" are used to reduce the amount of data being sent between two devices during a video call.
According to one method, a set of parameters or coefficients can be generated from, and associated with, a given video frame. For example, object landmarks points like facial features, orientation and scale may be called coefficients and can be tracked and stored alongside transmitted video. This data is processed into a searchable database from which single frames can be retrieved.
During low-bandwidth scenarios, a device can compute and send a set of coefficients for a given live video feed instead of a video frame. On the other hand, when bandwidth is available, the full image with associated coefficients may be sent and stored.
When a connection is poor and frames begin to drop, the system is able to use the coefficients to search a database of previously stored video for similar objects. In the example given, the coefficients of a live feed are compared to the video frame database to find a best match. When a frame with similar object characteristics is found, it is retrieved and displayed on the receiving device.
In another embodiment, the generated coefficients are used to intelligently morph an image in a previously recorded video frame. The result is inserted into the live feed in place of dropped frames.
Finally, objects deemed to be background content can be transmitted at a lower resolution than subject content. For example, the system will present a person's face in a higher resolution than the wall behind them.
Aside from the obvious benefits in call quality, the technology may also be used to cut back on data consumption as coefficients can be used to dynamically reproduce a scene from already transmitted video. Even if one artificially generated video frame were to be inserted for every three live frames, the savings would be huge.
As with any Apple patent application, it is unknown if the invention will one day be implemented into a consumer product. With rising data costs and the use of high-res camera modules, however, the feature would be useful for those who frequently make FaceTime video calls.
Apple's morphing and matching FaceTime video frame patent application was first filed for in 2012 and credits Alex Tremain Nelson and Richard E. Crandall as its inventors.