Superpixel segmentation has become an important research problem in image processing. In this paper, we propose an Iterative Spanning Forest (ISF) framework, based on sequences of Image Foresting Transforms, where one can choose i) a seed sampling strategy, ii) a connectivity function, iii) an adjacency relation, and iv) a seed pixel recomputation procedure to generate improved sets of connected superpixels (supervoxels in 3D) per iteration. The superpixels in ISF structurally correspond to spanning trees rooted at those seeds. We present five ISF methods to illustrate different choices of its components. These methods are compared with approaches from the state-of-the-art in effectiveness and efficiency. The experiments involve 2D and 3D datasets with distinct characteristics, and a high level application, named sky image segmentation. The theoretical properties of ISF are demonstrated in the supplementary material and the results show that some of its methods are competitive with or superior to the best baselines in effectiveness and efficiency.

No abstract

We investigate the intersection between hierarchical and superpixel image segmentation. Two strategies are considered: (i) the classical region merging, that creates a dense hierarchy with a higher number of levels, and (ii) the recursive execution of some superpixel algorithm, which generates a sparse hierarchy with fewer levels. We show that, while dense methods can capture more intermediate or higher-level object information, sparse methods are considerably faster and usually with higher boundary adherence at finer levels. We first formalize the two strategies and present a sparse method, which is faster than its superpixel algorithm and with similar boundary adherence.We then propose a new dense method to be used as post-processing from the intermediate level, as obtained by our sparse method, upwards. This combination results in a unique strategy and the most effective hierarchical segmentation method among the compared state-of-the-art approaches, with efficiency comparable to the fastest superpixel algorithms.

Interpretation errors may still represent a limiting factor for diagnosing Cryptosporidium spp. oocysts with the conventional staining techniques. Humans and machines can interact to solve this problem. We developed a new temporary staining protocol associated with a computer program for the diagnosis of Cryptosporidium spp. oocysts in fecal samples. We established 62 different temporary staining conditions by studying 20 experimental protocols. Cryptosporidium spp. oocysts were concentrated using the Three Fecal Test (TF-Test®) technique and confirmed by the Kinyoun method. Next, we built a bank with 299 images containing oocysts. We used segmentation in superpixels to cluster the patches in the images; then, we filtered the objects based on their typical size. Finally, we applied a convolutional neural network as a classifier. The trichrome modified by Melvin and Brooke, at a concentration use of 25%, was the most efficient dye for use in the computerized diagnosis. The algorithms of this new program showed a positive predictive value of 81.3 and 94.1% sensitivity for the detection of Cryptosporidium spp. oocysts. With the combination of the chosen staining protocol and the precision of the computational algorithm, we improved the Ova and Parasite exam (O&P) by contributing in advance toward the automated diagnosis.

como parte dos requisitos exigidos para a obtenção do título de Mestra em Ciências na área de concentração em Saúde da Criança e do