Method and apparatus to correct digital video to counteract effect of camera shake

What is this patent about?

’450 is related to the field of digital image processing, specifically addressing the problem of motion blur in video and still photography. In conventional imaging, relative movement between the camera and the subject during the time the shutter is open causes light to smear across the sensor, degrading image quality. While mechanical stabilization exists, it often requires heavy, expensive lens components that can compromise optical performance.

The underlying idea behind ’450 is to replace or augment mechanical stabilization with a computational alignment strategy that treats motion as a mathematical transfer function. Instead of relying on a single long exposure that is prone to smearing, the system captures a rapid succession of images and uses precise motion data to shift and overlay them. By calculating the exact displacement of the device during the capture interval, the invention can digitally re-align the constituent frames to reconstruct a sharp, stabilized final output.

The claims of ’450 focus on a method and apparatus for stabilizing video by capturing a sequence of images and utilizing synchronous motion information from onboard sensors. The system records the physical movement of the device at the exact moment each frame is captured, allowing a processor to determine specific vertical and horizontal shift values. These values are used to modify individual frames before they are combined and compressed into a final, stabilized video stream.

In practice, the invention utilizes accelerometers or similar sensors to track the device's trajectory in real-time. This sensor data is mapped to the image coordinates, allowing the processor to perform a pixel-by-pixel re-registration of the sequence. This approach effectively mimics a high-speed shutter while maintaining the light-gathering benefits of a longer total exposure time, as the coherent addition of multiple frames improves the signal-to-noise ratio without the penalty of motion-induced smearing.

This digital approach differs from prior art by moving the stabilization logic from the optical path to the digital backend. Unlike traditional methods that simply sharpen a blurred image—often losing data in the process—this invention uses deconvolution filters and iterative algorithms like Least Mean-Square to reverse the blurring effect. By integrating motion sensing directly with the video compression pipeline, the system provides a lightweight, software-driven solution that corrects for both user instability and subject movement.