Super-resolution reconstruction method of single-width infrared image based on scale analogy

Abstract

The invention discloses a super-resolution reconstruction method of a single-width infrared image based on scale analogy for mainly solving the problems of fuzzy edge, weak texture, low signal to noise ratio (SNR) and bad real-time performance. The method comprises the following steps: (1) performing bi-cubic interpolation to an infrared image s with to-be-reconstructed super-resolution to obtain an image B after the bi-cubic interpolation; (2) performing non-subsampled contourlet decomposition to an infrared image H with high resolution, using a low-pass sub-band thereof as a low-resolution image for studying, and using N band-pass directional sub-bands thereof as a high-frequency detail image for studying, and studying mapping relations between the low-resolution image and the high-frequency detail image through the scale analogy in order; (3) orderly acting the mapping relations on the image B after the bi-cubic interpolation to obtain the N band-pass directional sub-bands with high resolution; and (4) performing the non-subsampled contourlet inverse transformation to the infrared image B after the bi-cubic interpolation and the N band-pass directional sub-bands with high resolution to obtain the infrared image S with reconstructed super-resolution. The infrared image with reconstructed super-resolution has good visual effect, prominent texture and edge and high SNR.

Description

technical field [0001] The invention belongs to the technical field of image signal processing, and relates to a super-resolution image reconstruction method, in particular to a super-resolution reconstruction method of a single infrared image based on scale analogy. Background technique [0002] Infrared imaging equipment has a long range, good concealment, and ideal environmental adaptability. Compared with visible light, infrared images have the characteristics of less interference and concise image information, and can also reflect the thermal radiation characteristics of objects. Therefore, it is widely used in military and civilian fields. However, infrared images generally have shortcomings such as overall darkness, low contrast between the target and the background, weak texture, large noise, and relatively concentrated grayscale distribution of the image, making it difficult to observe the details of the image. Due to the above characteristics, super-resolution rec...

AI Technical Summary

Problems solved by technology

In order to obtain more accurate detailed information in the learning-based super-resolution method, the number of samples in the training set is often large, resulting in a slow running speed of the algorithm.

In 2001, Herzmann et al. of New York University proposed image analogy, and super-resolution reconstruction can be realized through image analogy, but because the gray value of the image varies in a relatively large range, it will cause a large error in the analogy process

This method can produce more reasonable details to enhance the image, but because the learning sample is only itself, the amount of information that can be obtained for learning is low, so the effect of the obtained super-resolution image is not very ideal, and certain artificial traces and artifacts will be generated. Visually hard edges

For infrared images with low resolution and blurred edge textures, most commonly used super-resolution reconstruction methods have problems such as edge blurring, weak texture details, low signal-to-noise ratio, and weak real-time performance.

Images