Method and apparatus for capturing digital video

What is this patent about?

'254 is related to the field of digital image processing, specifically addressing the problem of image distortion caused by relative motion between the imaging device (e.g., a camera) and the subject being photographed. This distortion, often manifested as blur, arises from camera shake or subject movement during image capture, especially with slow shutter speeds. Existing techniques to sharpen images after capture often result in data loss and do not truly correct the underlying blur. Furthermore, mechanical image stabilization systems add cost, weight, and complexity to lenses.

The underlying idea behind '254 is to compensate for image blur by either counteracting the motion during capture or computationally removing the blur after capture. This is achieved by determining a transfer function that mathematically describes the distortion caused by the motion. This transfer function can be derived either from motion sensors that track the camera's movement or through blind estimation techniques that analyze the blurred image itself. Once the transfer function is known, it can be inverted and used to deconvolve the blurred image, effectively recovering the original, undistorted image.

The claims of '254 focus on an imaging apparatus (e.g., a digital camera) equipped with a movable image sensor, motion sensors, and a processor. The processor executes instructions to receive motion information from the sensors and then uses this information to either adjust the position of the image sensor during image capture to counteract motion blur, or to process a sequence of captured images to calculate pixel values for a corrected image, taking into account the motion-induced offsets between the images. The corrected image is then stored and displayed.

In practice, the invention uses motion sensors (like accelerometers) to detect camera shake during image capture. This motion data is then used to calculate the transfer function, which represents how each point in the original image was displaced during the exposure. The inverse of this transfer function is then applied as a deconvolution filter to the blurred image, effectively undoing the distortion. Alternatively, the image sensor itself can be moved during capture, guided by the motion sensor data, to physically stabilize the image and prevent blur from occurring in the first place.

A key differentiation from prior approaches is the use of computational methods to correct blur, rather than relying solely on optical or mechanical stabilization. The patent also describes a method of capturing multiple images at a fast shutter speed and then combining them, using motion sensor data to align the images and reduce blur. This approach avoids the trade-off between shutter speed and image brightness, allowing for sharper images even in low-light conditions. The ability to designate a subject and selectively deblur it against a blurred background is another novel feature.