A single-frame image super-resolution reconstruction method based on pre-amplified non-negative neighborhood embedding

Abstract

The invention discloses a single-frame image super-resolution reconstruction method based on pre-amplification non-negative neighborhood embedding, the steps of which are: constructing a high-resolution training image set; blurring and down-sampling it to obtain a temporary low-resolution image set; Pre-enlarge the temporary low-resolution image set by 2 times to obtain a low-resolution training image set; construct a low-resolution training image block set; construct a high-resolution training image block set; pre-amplify the input low-resolution image by 2 times; construct a low-resolution The high-resolution input image block set; the non-negative neighborhood embedding represents the low-resolution input image block set, and obtains the reconstruction coefficient; uses the high-resolution training image block set and the obtained coefficient to obtain a high-resolution output image. The present invention utilizes the non-local similarity of image blocks to propose a new method for constructing training samples, and at the same time uses non-negative neighborhood embedding to effectively solve the problem of selecting the number of neighbors K. Experimental simulations show that the image reconstructed by the invention has clear edges and rich textures, and is closer to real images.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a single-frame image super-resolution reconstruction method based on neighborhood embedding, which can be widely used in the fields of remote sensing reconnaissance, traffic and safety monitoring, pattern recognition and the like. Background technique [0002] The higher the resolution of the image, the more informative it is. High-resolution images play an important role in a wide variety of practical applications. High-resolution satellite images are helpful for target recognition, and high-resolution images are also required in fields such as traffic and safety monitoring, and pattern recognition. Due to the physical conditions of the imaging system and the weather, there are often degradation processes such as motion blur, downsampling, and noise in the imaging process, resulting in low resolution and poor quality of the actual image. To obtain high-res...

AI Technical Summary

Problems solved by technology

However, this method relies on edge detection, and incorrect edge detection can lead to artifacts in the resulting image

Images