Method and system for generating high resolution image by fusing low resolution images

An image processing system, the system comprising:

• a processor;
• and
• a memory comprising a plurality of instructions, the plurality of instructions configured to cause the processor to:
• collect, via one or more hardware processors, a plurality of low resolution images as input, by an Input/Output (I/O) module of the image processing system;
• pre-process the plurality of low resolution images, via the one or more hardware processors, wherein the pre-processing comprises changing at least one characteristic of each of the plurality of low resolution images with respect to at least one reference image, and wherein the pre-processing further comprises:
• estimating rotation of each of the low resolution images with respect to the reference image and rotating the low resolution images to align with the reference image, by calculating a rotation matrix,
• scaling the low resolution images to match a resolution of a selected image such that all the low resolution images have the same resolution, wherein the selected image is one of the plurality of low resolution images identified to have a highest resolution, and
• estimating translation using a coarse estimation and a fine estimation, wherein the coarse estimation involves estimating pixel level shift of each image with respect to the reference image at a low resolution scale and selecting a common area between the image and the reference image, and wherein the fine estimation involves sub-pixel level estimation;
• enhance the pre-processed low resolution images based on a plurality of final enhancement factors, via the one or more hardware processors, by an enhancement the image processing system, wherein the plurality of final enhancement factors are dynamically determined for the plurality of low resolution images;
• and
• fuse the plurality of low resolution images using a multi-coset framework to generate a high resolution image, after the enhancement.

One or more non-transitory machine readable information storage media comprising one or more instructions which when executed by one or more hardware processors causes the one or more hardware processors to perform a method comprising:

• collecting, via one or more hardware processors, a plurality of low resolution images as input;
• pre-processing the plurality of low resolution images, via the one or more hardware processors, wherein the pre-processing comprises changing at least one characteristic of each of the plurality of low resolution images with respect to at least one reference image, and wherein the pre-processing further comprises:
• estimating rotation of each of the low resolution images with respect to the reference image and rotating the low resolution images to align with the reference image, by calculating a rotation matrix,
• scaling the low resolution images to match a resolution of a selected image such that all the low resolution images have the same resolution, wherein the selected image is one of the plurality of low resolution images identified to have a highest resolution, and
• estimating translation using a coarse estimation and a fine estimation, wherein the coarse estimation involves estimating pixel level shift of each image with respect to the reference image at a low resolution scale and selecting a common area between the image and the reference image, and wherein the fine estimation involves sub-pixel level estimation;
• enhancing the pre-processed low resolution images based on a plurality of final enhancement factors, via the one or more hardware processors, wherein the plurality of final enhancement factors are dynamically determined for the plurality of low resolution images;
• and
• fusing the plurality of low resolution images using a multi-coset framework to generate a high resolution image, after the enhancement, via the one or more hardware processors.