Image Mosaic
Mosaic
With a single sensor, there is a necessary trade-off between resolution and field of view. Higher resolution images are usually desired because they provide more information, but the use of high magnification optics results in a narrow field of view and loss of context and situational awareness.
A mosaic image, digitally generated from many individual overlapping high-resolution images, can simultaneously provide both very wide field of view and maintain full sensor resolution.
Mosaic Formation Process
A digital mosaic is formed by seamlessly joining multiple images from a single sensor into a larger image. In order to do this, a number of sophisticated and highly optimised algorithms are required, especially for video rate operation.
The core of the mosaic formation process is an image registration engine that must be able to intelligently cope with sensor noise, variable image content, large displacements, rotations and other image deformations, as well as independent moving targets. Once the relative displacements between two images are known, they must then be warped into alignment and any changes in sensor gain, offset, or colour must be corrected. Finally, the overlapping images must be blended together, taking care over any moving targets, to form the mosaic. The process is then repeated for every new video frame, which represents a considerable amount of image manipulation. The mosaic formation process is illustrated below.
The mosaic formation process digitally registers and blends images into a single larger image. This results in a wide field of view whilst maintaining high resolution.
Technology
Due to its demanding nature, Mosaic formation is often performed off-line after all the images are collected and optimises the global registration of all the available images simultaneously. However, we construct the mosaic in real-time, employing a sequential registration approach, whereby each new image is added to the mosaic as it becomes available.
Scenelock, Octec's real-time scene tracking module, runs continuously as a background task that reports the global scene motion and provides first-guess registration parameters, even when the motion is fast. The registration is then refined as the image is blended into the mosaic, resulting in very high quality, seamless mosaic.