Search

Deep convolutional neural networks (CNNs) have recalibrated the state-of-the-art for a plethora of applications in medical image analyzing such as segmentation and classification. Large receptive field is crucial for modeling long-range spatial dependency in medical images. In this paper, we propose a novel architectural network design for accomplishing a full-image global receptive field, which we call spectral residual block (SRB). Specifically, we propose to utilize a unitary transform that essentially conducts a local-to-global transform. All elements are mapped to spectral domain and thus globally depend on each other. A variety of global operators are carefully devised and efficiently enforce a full-image receptive field, including spectral ReLU for frequency-sensitive filtering and spectral convolutions. The output in spectral domain is eventually converted back global-to-local via a reverse unitary transform. The proposed framework is generic and flexible, and could be applied to various network structures and tasks. Comprehensive evaluations on skin lesion segmentation and Chest X-Ray classification show that our method achieves the state-of-the-art performance, demonstrating both effectiveness and efficiency.

Embedded watermarks in printed halftone images, which can subsequently be detected using an visual aid or using a watermark detection algorithm on a scan of the image, are of interest in wide range of applications. For black and white halftone printing using stochastic screens, digital watermarks that are embedded as correlations in the halftone screen have been previously proposed. Here we present a novel extension of these watermarks to color that produces a high contrast watermark by using the colorant separations coherently with a single watermarked stochastic screen and performing detection coherently across the color separations. Compared with independent watermarking of the halftone separations, the resulting watermark offers significantly higher contrast in the detected image.