The usage of convolutional neural networks (CNNs) for unsupervised image segmentation was investigated in this study. Similar to supervised image segmentation, the proposed CNN assigns labels to pixels that denote the cluster to which the pixel belongs. In unsupervised image segmentation, however, no training images or ground truth labels of pixels are specified beforehand. Therefore, once a target image is input, the pixel labels and feature representations are jointly optimized, and their parameters are updated by the gradient descent. In the proposed approach, label prediction and network parameter learning are alternately iterated to meet the following criteria: (a) pixels of similar features should be assigned the same label, (b) spatially continuous pixels should be assigned the same label, and (c) the number of unique labels should be large. Although these criteria are incompatible, the proposed approach minimizes the combination of similarity loss and spatial continuity loss to find a plausible solution of label assignment that balances the aforementioned criteria well. The contributions of this study are four-fold. First, we propose a novel end-to-end network of unsupervised image segmentation that consists of normalization and an argmax function for differentiable clustering. Second, we introduce a spatial continuity loss function that mitigates the limitations of fixed segment boundaries possessed by previous work. Third, we present an extension of the proposed method for segmentation with scribbles as user input, which showed better accuracy than existing methods while maintaining efficiency. Finally, we introduce another extension of the proposed method: unseen image segmentation by using networks pre-trained with a few reference images without re-training the networks. The effectiveness of the proposed approach was examined on several benchmark datasets of image segmentation.
21 May 2015 in IEEE Transactions on Image Processing
Single-sensor imaging using the Bayer color filter array (CFA) and demosaicking is well established for current compact and low-cost color digital cameras. An extension from the CFA to a multispectral filter array (MSFA) enables us to acquire a multispectral image in one shot without increased size or cost. However, multispectral demosaicking for the MSFA has been a challenging problem because of very sparse sampling of each spectral band in the MSFA. In this paper, we propose a high-performance multispectral demosaicking algorithm, and at the same time, a novel MSFA pattern that is suitable for our proposed algorithm. Our key idea is the use of the guided filter to interpolate each spectral band. To generate an effective guide image, in our proposed MSFA pattern, we maintain the sampling density of the G-band as high as the Bayer CFA, and we array each spectral band so that an adaptive kernel can be estimated directly from raw MSFA data. Given these two advantages, we effectively generate the guide image from the most densely sampled G-band using the adaptive kernel. In the experiments, we demonstrate that our proposed algorithm with our proposed MSFA pattern outperforms existing algorithms and provides better color fidelity compared with a conventional color imaging system with the Bayer CFA. We also show some real applications using a multispectral camera prototype we built.
A color difference interpolation technique is widely used for color image demosaicking. In this paper, we propose residual interpolation as an alternative to the color difference interpolation, where the residual is a difference between an observed and a tentatively estimated pixel value. We incorporate the proposed residual interpolation into the gradient based threshold free (GBTF) algorithm, which is one of current state-of-the-art demosaicking algorithms. Experimental results demonstrate that our proposed demosaicking algorithm using the residual interpolation can give state-of-the-art performance for the 30 images of Kodak and IMAX datasets.
Color image demosaicking for the Bayer color filter array is an essential image processing operation for acquiring high-quality color images. Recently, residual interpolation (RI)-based algorithms have demonstrated superior demosaicking performance over conventional color difference interpolation-based algorithms. In this paper, we propose adaptive residual interpolation (ARI) that improves existing RI-based algorithms by adaptively combining two RI-based algorithms and selecting a suitable iteration number at each pixel. These are performed based on a unified criterion that evaluates the validity of an RI-based algorithm. Experimental comparisons using standard color image datasets demonstrate that ARI can improve existing RI-based algorithms by more than 0.6 dB in the color peak signal-to-noise ratio and can outperform state-of-the-art algorithms based on training images. We further extend ARI for a multispectral filter array, in which more than three spectral bands are arrayed, and demonstrate that ARI can achieve state-of-the-art performance also for the task of multispectral image demosaicking.
In recent years, many applications using a set of red-green-blue (RGB) and near-infrared (NIR) images, also called an RGB-NIR image, have been proposed. However, RGB-NIR imaging, i.e., simultaneous acquisition of RGB and NIR images, is still a laborious task because existing acquisition systems typically require two sensors or shots. In contrast, single-sensor RGB-NIR imaging using an RGB-NIR sensor, which is composed of a mosaic of RGB and NIR pixels, provides a practical and low-cost way of one-shot RGB-NIR image acquisition. In this paper, we investigate high-quality system designs for single-sensor RGB-NIR imaging. We first present a system evaluation framework using a new hyperspectral image dataset we constructed. Different from existing work, our framework takes both the RGB-NIR sensor characteristics and the RGB-NIR imaging pipeline into account. Based on the evaluation framework, we then design each imaging factor that affects the RGB-NIR imaging quality and propose the best-performed system design. We finally present the configuration of our developed prototype RGB-NIR camera, which was implemented based on the best system design, and demonstrate several potential applications using the prototype.
A division-of-focal-plane or microgrid image polarimeter enables us to acquire a set of polarization images in one shot. Since the polarimeter consists of an image sensor equipped with a monochrome or color polarization filter array (MPFA or CPFA), the demosaicking process to interpolate missing pixel values plays a crucial role in obtaining high-quality polarization images. In this study, we proposed a novel MPFA demosaicking method based on intensity-guided residual interpolation (IGRI) and extended it to CPFA demosaicking. The key of IGRI is generating an effective intensity guide image, for which we proposed two methods considering four-directional intensity and polarization edge information. We also constructed a new full color-polarization image dataset captured using a 3-CCD RGB camera and a rotating polarizer. By using the constructed dataset, we experimentally validated that our IGRI-based methods outperform existing methods in MPFA and CPFA demosaicking.