Search

We present in this paper an approach based on multiple strategies to tackle the problem of invariance in the framework of texture-based image retrieval. This approach, rather than to consider invariance at the representation level, considers it at the query level. We use two models to represent the visual content of textures, namely, the autoregressive model and a perceptual model based on a set of perceptual features. The perceptual model is used with two viewpoints: the original image's viewpoint and the autocovariance function viewpoint. After a brief presentation and discussion of these multiple representation models/viewpoints, we present the invariant texture retrieval algorithm. This algorithm considers results fusion (merging) at two different levels: (1) the first level consists in merging results returned by different models/viewpoints (representations) for the same query in one results list using a linear results fusion model; (2) the second level consists in merging each fused list of different queries into a unique fused list using a round robin fusion scheme. Experimental retrieval results and benchmarking based on the precision/recall measures applied on a large image database show interesting results. [ABSTRACT FROM AUTHOR]

The main contribution of this work is a novel set of image features called the virtual circles and their use in the registration of images under similarity transformations. A virtual circle is a circle with maximal radius encompassing a background area that does not contain edge points. It has many useful properties such as its radius, and its dominant edge direction for example, which can be utilized for efficient registration. Furthermore, virtual circles are frequent and can be extracted efficiently with the help of the distance transform from many types of images. We have tested the new virtual circles method in the registration of 66 pairs of images, half of which are printed labels and the other half are indoor scenes. Experimental results have shown that this method has a linear complexity in terms of the number of pixels. It is also highly automatic, because it has a small number of parameters, which almost never need to be changed throughout the experiments. [ABSTRACT FROM AUTHOR]