A Potential Method for the Nonuniformity Correction and Noise Removal of Infrared Thermal Image

Nonuniformity and noise are the two crucial factors that reduce the sharpness of infrared thermal images. It is essential to correct the nonuniformity and remove the noise of thermal images. A method for the nonuniformity correction and noise removal of the infrared thermal images that combines convolutional neural networks with a double-Gaussian filter was proposed. To demonstrate the advantages of this method, the values of roughness and nonuniformity of 300 infrared thermal images with different degrees of nonuniformity captured from various focal distances and fields of view were analyzed based by combined convolutional neural networks with a double-Gaussian filter. Furthermore, the results of that were compared in the form of a line chart with other commonly used algorithms. As a result, combined convolutional neural networks with a double-Gaussian filter was neither the best nor the worst one from the point of roughness and nouniformity while it was the best from a comprehensive performance of view. So, combined convolutional neural networks with a double-Gaussian filter may be a potential method for the nonuniformity correction and denoising of the infrared thermal image.