Many low-cost sensors (or cameras) may spatially or electronically under-sample an image. Similarly, cameras taking pictures from great distances, such as aerial photos, may not obtain detailed information about the subject matter. This may result in aliased images in which the high-frequency components are folded into the low-frequency components in the image. Consequently, subtle or detailed information (high-frequency components) is not present in the images.
When under-sampled images have subpixel translations and rotations between successive frames, they contain different information regarding the same scene. SRIR combines information contained in under-sampled images to obtain an aliasfree (high-resolution) image. Super-resolution image reconstruction from multiple snapshots, taken by a detector that has shifted and/or rotated in position, provides far more detailed information than any interpolated image from a single snapshot.
Generally, SRIR involves acquiring a sequence of images from the same scene with sub-pixel translations and rotations among the images. The methods for acquiring low-resolution images with subpixel translations and rotations include acquiring images affected by translation and/or rotation.
There is a need to produce high-resolution images from a sequence of low-resolution images captured by a low-cost imaging device that has experienced translations and/or rotations. The amount of sub-pixel translation and/or rotation may be unknown, and it creates a need to estimate sub-pixel translation and rotation. In order to produce high-resolution images from a sequence of translated and rotated under-sampled (low-resolution) images, there exists a need to eliminate aliasing, while taking into account natural jitter.
An algorithm was developed that effectively increases the resolution of the reconstructed output image using the existing imaging device from a sequence of low-resolution, under-sampled imagery. By removing aliasing due to under-sampling, SRIR can also improve range performance of the sensor.
SRIR exploits the sub-pixel movement of an image sequence. When low-resolution images have sub-pixel shifts be tween successive frames, they convey different information about the same scene. SRIR takes advantage of this distinct information, and fuses the information from the low-resolution frames during reconstruction to generate a high-quality, high-resolution image of the true scene. Enhanced performance can be obtained using existing hardware.
The method can accommodate subpixel movement of an unknown nature such as natural jitter. It also requires very few input images — 4 frames for 2× linear (4× pixel) resolution improvement. This provides low computational overhead, fast processing, and realistic mimicking of fast-moving targets. Applications include infrared and flash LADAR, cellphones, medical imaging, remote sensing, target recognition, biometric recognition, and industrial inspection.