Sensor-assisted motion estimation for efficient video encoding
It is very important to note that motion in a successive video sequence consists of two parts: global motion and local motion. The global motion is generally engendered by camera movement, while the local motion results from object movement in the scene. Most existing motion estimation algorithms do not distinguish them and use a block matching algorithm (BMA) block by block for the encoded picture. Since every block is possible to embed both global and local motions, existing solutions often have to employ a large search window and match all possible candidate blocks. Therefore, they are very compute-intensive. This invention efficiently estimates global motion in video sequences using acceleration data. As a result, the search algorithm only has to find the local motion for each block. As objects in scene typically move very short distance in the time period between two adjacent frames, e.g. 1/30s, the local motion search range can be set relatively small, considerably reducing computation requirements without adversely effecting quality and/or compression.Advantages Sensor assisted video encoding represents a new direction in video encoding. By utilizing data from sensors, video encoding equipment can provide increased video quality and/or reduce their processing requirements, leading to reduced power consumption and increased battery life. The inventors have developed a prototype, Accelerometer-Assisted Video Encoder (AAVE), by bundling a commercial digital camcorder and an in-house sensor board with dual three-axis accelerometers. We have employed a two-stage motion estimation method, consisting of global motion estimation via sensors for predicting camera motion and local motion estimation for macroblock prediction. The findings indicate that AAVE dramatically improves the efficiency of MPEG encoding, by 2-3 times.
Potential Applications This invention represents a new direction in exploiting the synergy between different components in a handheld device to improve the efficiency of multimedia processing. Almost any company, both hardware and software, in the electronics and video industry involved with video recording and/or encoding can benefit from this technology.
Inventor(s): Ye Wang
Type of Offer: Licensing
« More Computer Science Patents