Within the frames of Abto Software internal project our Research and Modeling experts studied the possibility of receiving high resolution (HR) images out of a set of low resolution (LR) images. As a result, our specialists developed a truly superb program for getting HR images where it is possible to identify the smallest details which couldn’t be seen on the incoming images. The program can be used in a number of spheres like astronomy, microscopy and healthcare. It can also be applied for the purposes of identification, video surveillance systems, in all areas where high quality images are extremely important.
Abto Software specialists created a MATLAB prototype to implement the algorithm which is based on the model of high resolution images formation. To carry out a successful search for a certain high resolution image our experts offered the solution of the optimization task by means of the conjugated gradients method.
Below you can see the example of an image in twelve low resolution frames of 128×128 pixels which we needed to resize and receive high resolution image. For that there can be used three methods:
1) “Nearest-neighbor interpolation”: interpolation by the nearest neighbor gives unsatisfactory results as small details are still not seen;
2) Bicubic interpolation: the results are also poor as the received images are blurred;
3) “Super-resolution”: by means of super-resolution method most clear and distinct image is received where the smallest details which couldn’t be seen on the incoming images can be identified.
The image below obtaind after using “nearest-neighbor interpolation”:
The image below obtained after applying bicubic interpolation method:
The image below obtained after “super-resolution” method was used:
To get to know more about different techniques that we use in projects that delve into Digital Image Processing field, please view our articles “Introduction to Image Restoration Methods” posted in Abto Software Blog. There we describe general Image Restoration approaches, Convolution, Deconvolution, Inverse Filtering and Wiener filtering and some other topics regarding image restoration.