Image processing method and system

Abstract

An image processing method comprises inputting data (e.g. image data, motion data, temporal data or context data) relating to one or more first image frames (e.g. frames 54, 56 figure 5) for content to a machine learning model. The machine learning model determines 420 a quality prediction for at least part of a second image frame for the content assuming the second image frame is to be generated based on the one or more first image frames. The quality prediction is compared 432 with a predetermined quality threshold. Dependent upon the quality prediction, the second image frame (e.g. frame 55, figure 5) is generated at least partly based on the one or more first image frames. When the quality prediction is above a predetermined threshold generation of the second image frame (or a portion thereof) may comprise interpolation of extrapolation 434 using the one or more first image frames. When the quality prediction is below the threshold, the second image frame (or a portion thereof) is rendered 436. A corresponding image processing system is also disclosed. The generation of image frames enables a higher frame rate to be achieved, thus improving realism and accuracy of displayed content in video game or movie applications. [Figure 4]