Pathological conditions in diabetic feet cause surface temperature variations, which can be captured quantitatively using infrared thermography. Thermal images captured during recovery of diabetic feet after active cooling may reveal richer information than those from passive thermography, but diseased foot regions may exhibit very small temperature differences compared with the surrounding area, complicating plantar foot segmentation in such cold-stressed active thermography. In this study, we investigate new plantar foot segmentation methods for thermal images obtained via cold-stressed active thermography without the complementary information from color or depth channels. To better deal with the temporal variations in thermal image contrast when planar feet are recovering from cold immersion, we propose an image pre-processing method using a two-stage adaptive gamma transform to alleviate the impact of such contrast variations. To improve upon existing deep neural networks for segmenting planar feet from cold-stressed infrared thermograms, a new deep neural network, the Plantar Foot Segmentation Network (PFSNet), is proposed to better extract foot contours. It combines the fundamental U-shaped network structure, a multi-scale feature extraction module, and a convolutional block attention module with a feature fusion network. The PFSNet, in combination with the two-stage adaptive gamma transform, outperforms multiple existing deep neural networks in plantar foot segmentation for single-channel infrared images from cold-stressed infrared thermography, achieving an accuracy of 97.3% and 95.4% as measured by Intersection over Union (IOU) and Dice Similarity Coefficient (DSC) respectively.

Multi-focus image fusion consists in the integration of the focus regions of multiple source images into a single image. At present, there are still some common problems in image fusion methods, such as block artifacts, artificial edges, halo effects, and contrast reduction. To address these problems, a novel, to the best of our knowledge, multi-focus image fusion method using energy of Laplacian and a deep neural network (DNN) is proposed in this paper. The DNN is composed of multiple denoising autoencoders and a classifier. The Laplacian energy operator can effectively extract the focus information of source images, and the trained DNN model can establish a valid mapping relationship between source images and a focus map according to the extracted focus information. First, the Laplacian energy operator is used to perform focus measurement for two source images to obtain the corresponding focus information maps. Then, the sliding window technology is used to sequentially obtain the windows from the corresponding focus information map, and all of the windows are fed back to the trained DNN model to obtain a focus map. After binary segmentation and small region filtering, a final decision map with good consistency is obtained. Finally, according to the weights provided by the final decision map, multiple source images are fused to obtain a final fusion image. Experimental results demonstrate that the proposed fusion method is superior to other existing ones in terms of subjective visual effects and objective quantitative evaluation.