An Improving Infrared Image Resolution Method via Guided Image Filtering

Abstract

Image resolution is of importance to image processing. In super resolution methods, images obtained from same sensor are adopted to improve the resolution of image. However these methods do not take advantage of the correlation of the images from different sensors to improve the resolution of image. To solve this problem, in this paper a method is proposed to enhance infrared image by using the correlation of an infrared (IR) image and its corresponding visible image. Firstly, phase congruency is used to generate the edge maps of the infrared and visible images. Then, correlated edge regions and uncorrelated regions are calculated according to the edge maps. Finally, different strategies are applied to those regions. Specifically, for the correlated regions, the visible image is considered as the guidance image while for uncorrelated regions, the infrared image itself is considered as the guidance image. Hereafter guided image filter is applied to the infrared image. Finally, the filtered result of correlated edge regions and uncorrelated regions are combined to obtain the final result. The resultant image inferred by the proposed method is with better subjective and objective quality compared with other methods.

Reproducing kernel hilbert space based single infrared image super resolution

“…The human vision processing mechanism together with vision lateral inhibition was utilized to enhance the image contrast between object and background [28]. Moreover, the correlation of an infrared image and its corresponding visible image were suggested to improve resolution, in which the visible image generated edge maps [29]. In overall, most approaches were directly borrowed from visible image processing which didn't consider the features…”