Carely Guada

Classifying image analysis techniques from their output

In this paper we discuss some main image processing techniques in order to propose a classification based upon the output these methods provide. Because despite a particular image analysis technique can be supervised or unsupervised, and can allow or not the existence of fuzzy information at some stage, each technique has been usually designed to focus on a specific objective, and their outputs are in fact different according to each objective. Thus, they are in fact different methods. But due to the essential relationship between them they are quite often confused. In particular, this paper pursues a clarification of the differences between image segmentation and edge detection, among other image processing techniques.

A methodology for hierarchical image segmentation evaluation

This paper proposes a method to evaluate hierarchical image segmentation procedures, in order to enable comparisons between different hierarchical algorithms and of these with other (non-hierarchical) segmentation techniques (as well as with edge detectors) to be made. The proposed method builds up on the edge-based segmentation evaluation approach by considering a set of reference human segmentations as a sample drawn from the population of different levels of detail that may be used in segmenting an image. Our main point is that, since a hierarchical sequence of segmentations approximates such population, those segmentations in the sequence that best capture each human segmentation level of detail should provide the basis for the evaluation of the hierarchical sequence as a whole. A small computational experiment is carried out to show the feasibility of our approach.

A New Edge Detection Approach Based on Fuzzy Segments Clustering

Edge detection comprises different stages that go from adaptation of the original image -conditioning- to the selection of the definitive edges. This last step, known as scaling, requires the application of a thresholding process over the gradients of luminosity values of the pixels. Traditionally, this is made through a local evaluation process that works pixel by pixel. As the edge candidate pixels are not independent, a wider strategy suggests the use of a more global evaluation. In this sense, this approach resembles more the human vision. This paper further develops ideas related to edge lists, first proposed in 1995 [1]. This paper will refer to edge lists as edge segments. These segments contain visual features similar to the ones that humans use, which might lead to better comparative results. In this paper we propose using clustering techniques to differentiate the appropriate segments or true segments from the false ones, and we introduce an algorithm that uses fuzzy clustering techniques. Finally, this paper shows that this fuzzy clustering over the segments performs at least as well as other standard algorithms used for edge detection.

Automatic Detection of Thistle-Weeds in Cereal Crops from Aerial RGB Images

Capturing aerial images by Unmanned Aerial Vehicles (UAV) allows gathering a general view of an agricultural site together with a detailed exploration of its relevant aspects for operational actions. Here we explore the challenging task of detecting cirsium arvense, a thistle-weed species, from aerial images of barley-cereal crops taken from 50 m above the ground, with the purpose of applying herbicide for site-specific weed treatment. The methods for automatic detection are based on object-based annotations, pointing out the RGB attributes of the Weed or Cereal classes for an entire group of pixels, referring to a crop area which will have to be treated if it is classified as being of the Weed class. In this way, an annotation belongs to the Weed class if more than half of its area is known to be covered by thistle weeds. Hence, based on object and pixel-level analysis, we compare the use of k-Nearest Neighbours (k-NN) and (feed-forward, one-hidden layer) neural networks, obtaining the best results for weed detection based on pixel-level analysis, based on a soft measure given by the proportion of predicted weed pixels per object, with a global accuracy of over 98%.

Graph Approach in Image Segmentation

In this paper we discuss about graph approach in image segmentation. In first place, some main image processing techniques are classified based upon the output these methods provide. Then, a fuzzy image segmentation definition is presented because in the literature review was found that it was not clearly defined. This definition of fuzzy image segmentation is then related to a hierarchical image segmentation procedure, so this concept is also formally defined in this work. As every output of an image processing algorithm has to be evaluated, then a method to evaluate a hierarchical segmentation output is proposed in order to later propose a method to evaluate a fuzzy image segmentation output. Computational experiences point to some advantages of the proposed hierarchical image segmentation procedure over other algorithms.

Fuzzy Image Segmentation Based on the Hierarchical Divide and Link Clustering Algorithm

In this paper, we present a method to obtain a Fuzzy Image Segmentation from the hierarchical clustering algorithm Divide and Link. The Divide and Link algorithm consists on treating a digital image as a graph, then building spanning forest through a Kruskal scheme to successively sort the edges while partitions are obtained. This process is driven until all the pixels of the image are segmented, that is, there are as many regions as pixels.