Non-uniform blind deblurring of dynamic scenes has always been a challenging problem in image processing because of the diverse of blurring sources. Traditional methods based on energy minimization cannot make accurate kernel estimation. It leads to that some high frequency details cannot be fully recovered. Recently, many methods based on convolution neural networks (CNNs) have been proposed to improve the overall performance. Followed by this trend, we first propose a two-stage deblurring module to recover the blur images of dynamic scenes based on high frequency residual image learning. The first stage performs initial deburring with the blur kernel estimated by the salient structure. The second stage calculates the difference of input image and initially deblurred image, referred to as residual image, and adopt an encoder-decoder network to refine the residual image. Finally, we can combine the refined residual image with the input blurred image to obtain the latent image. To increase deblurring performance, we further propose a coarse-to-fine framework based on the deblurring module. It performs the deblurring module many times in a multi-scale manner which can gradually restore the sharp edge details of different scales. Experiments conducted on three benchmark datasets demonstrate the proposed method achieves competitive performance of state-of-art methods.

Cancer is one of the common diseases. Quantitative biomarkers extracted from standard-of-care computed tomography (CT) scan can create a robust clinical decision tool for the diagnosis of hepatocellular carcinoma (HCC). According to the current clinical methods, the situation usually accounts for high expenditure of time and resources. To improve the current clinical diagnosis and therapeutic procedure, this paper proposes a deep learning-based approach, called Successive Encoder-Decoder (SED), to assist in the automatic interpretation of liver lesion/tumor segmentation through CT images. The SED framework consists of two different encoder-decoder networks connected in series. The first network aims to remove unwanted voxels and organs and to extract liver locations from CT images. The second network uses the results of the first network to further segment the lesions. For practical purpose, the predicted lesions on individual CTs were extracted and reconstructed on 3D images. The experiments conducted on 4300 CT images and LiTS dataset demonstrate that the liver segmentation and the tumor prediction achieved 0.92 and 0.75 in Dice score, respectively, by as-proposed SED method.

The life cycle of leaves, from sprout to senescence, is the phenomenon of regular changes such as budding, branching, leaf spreading, flowering, fruiting, leaf fall, and dormancy due to seasonal climate changes. It is the effect of temperature and moisture in the life cycle on physiological changes, so the detection of newly grown leaves (NGL) is helpful for the estimation of tree growth and even climate change. This study focused on the detection of NGL based on deep learning convolutional neural network (CNN) models with sparse enhancement (SE). As the NGL areas found in forest images have similar sparse characteristics, we used a sparse image to enhance the signal of the NGL. The difference between the NGL and the background could be further improved. We then proposed hybrid CNN models that combined U-net and SegNet features to perform image segmentation. As the NGL in the image were relatively small and tiny targets, in terms of data characteristics, they also belonged to the problem of imbalanced data. Therefore, this paper further proposed 3-Layer SegNet, 3-Layer U-SegNet, 2-Layer U-SegNet, and 2-Layer Conv-U-SegNet architectures to reduce the pooling degree of traditional semantic segmentation models, and used a loss function to increase the weight of the NGL. According to the experimental results, our proposed algorithms were indeed helpful for the image segmentation of NGL and could achieve better kappa results by 0.743.

Hollow organ perforation can precipitate a life-threatening emergency due to peritonitis followed by fulminant sepsis and fatal circulatory collapse. Pneumoperitoneum is typically detected as subphrenic free air on frontal chest X-ray images; however, treatment is reliant on accurate interpretation of radiographs in a timely manner. Unfortunately, it is not uncommon to have misdiagnoses made by emergency physicians who have insufficient experience or who are too busy and overloaded by multitasking. It is essential to develop an automated method for reviewing frontal chest X-ray images to alert emergency physicians in a timely manner about the life-threatening condition of hollow organ perforation that mandates an immediate second look. In this study, a deep learning-based approach making use of convolutional neural networks for the detection of subphrenic free air is proposed. A total of 667 chest X-ray images were collected at a local hospital, where 587 images (positive/negative: 267/400) were used for training and 80 images (40/40) for testing. This method achieved 0.875, 0.825, and 0.889 in sensitivity, specificity, and AUC score, respectively. It may provide a sensitive adjunctive screening tool to detect pneumoperitoneum on images read by emergency physicians who have insufficient clinical experience or who are too busy and overloaded by multitasking.

Band selection (BS) is one of the important topics in hyperspectral image (HSI) processing. Many types of BS algorithms were proposed in the last decade. However, most of them were designed for off-line use. They can only be used with pre-collected data, and are sometimes ineffective for applications that require timeliness, such as disaster prevention or target detection. This paper proposes an online BS method that allows us obtain instant BS results in a progressive manner during HSI data transmission, which is carried out under band-interleaved-by-sample/pixel (BIS/BIP) format. Such a revolutionary method is called progressive sample processing of band selection (PSP-BS). In PSP-BS, BS can be done recursively pixel by pixel, so that the instantaneous BS can be achieved without waiting for all the pixels of an image. To develop a PSP-BS algorithm, we proposed PSP-OMPBS, which adopted the recursive version of a self-sparse regression BS method (OMPBS) as a native algorithm. The experiments conducted on two real hyperspectral images demonstrate that PSP-OMPBS can progressively output the BS with extremely low computing time. In addition, the convergence of BS results during transmission can be further accelerated by using a pre-defined pixel transmission sequence. Such a significant advantage not only allows BS to be done in a real-time manner for the future satellite data downlink, but also determines the BS results in advance, without waiting to receive every pixel of an image.

Band subset selection (BSS) is one of the ways to implement band selection (BS) for a hyperspectral image (HSI). Different from conventional BS methods, which select bands one by one, BSS selects a band subset each time and preserves the best one from the collection of the band subsets. This paper proposes a BSS method, called band grouping-based sparse self-representation BSS (BG-SSRBSS), for hyperspectral image classification. It formulates BS as a sparse self-representation (SSR) problem in which the entire bands can be represented by a set of informatively complementary bands. The BG-SSRBSS consists of two steps. To tackle the issue of selecting redundant bands, it first applies band grouping (BG) techniques to pre-group the entire bands to form multiple band groups, and then performs band group subset selection (BGSS) to find the optimal band group subset. The corresponding representative bands are taken as the BS result. To efficiently find the nearly global optimal subset among all possible band group subsets, sequential and successive iterative search algorithms are adopted. Land cover classification experiments conducted on three real HSI datasets show that BG-SSRBSS can improve classification accuracy by 4–20% compared to the existing BSS methods and requires less computation time.