A CMOS imager with pixel prediction for image compression

A predictive approach to on-chip focal plane compression is presented. This approach, unlike the transform coding techniques, is based on pixel prediction. In this technique, the value of each pixel is predicted by considering the values of its closest neighbor pixels. An analog circuit has been designed to accomplish a simple prediction function based on the values of two neighbor pixels. To demonstrate the approach, a prototype chip containing an imager and the prediction circuits has been designed, implemented, and tested.