Image Interpolation for Motion Estimation using Feedforward MLP Artificial Neural Networks (ANNs)

Motion estimation and comoensation is the key to high quality video coding used in most video codecs, including MPEG-2, MPEG-4 and the like. It t also a key component in the digital restoration of archived video and for post-production and special effects in the movie industry. In this paper, a linear image interpolating method based on artificial neural network is presented to estimate the middle frame between two successive frames. This is done by the application of a j-layer feedforward neural. Experiments reveal that the proposed interpolation algorithm based on neural network provides a reasonable performance. This algorithm may also be used to predict the motion of an object. Such applications can be found in biologt, physics, or any other areas in which motion ofobjects are observed.

کلیدواژه ها:

Image Interpolation ، Motion Estimation ، Multi-Layer Perceptron (MLp) ، Artificial Neural Network