This technology relates to an image restoration apparatus and method. Specifically, it is an image super-resolution restoration technology using patch feature-based similarity search.
Conventional methods needed to address the challenge of achieving stable performance in sample-based super-resolution image restoration by determining the degree of energy based on patch gradients and using it as a function for similarity search between patches. To address this, the present technology proposes a configuration that includes an image decoder with a resolution conversion unit, a patch partition section, a similarity measurement unit, and an image restoration part.
Accordingly, this technology can improve the performance of sample-based monoscopic image super-resolution restoration methods by determining the degree of energy based on patch gradients and using it as a function for similarity search between patches. It has practical value in the fields of software and electronic devices.
N/A
