Local feature methods suitable for image feature based object recognition and for the estimation of motion and structure are composed of two steps, namely the 'where ' and 'what' steps. The 'where' step (e.g., interest point detector)

No abstract

A key step for the effective use of local image features (i.e., highly distinctive and robust features) for recognition or image matching is the appropriate grouping of feature matches. Spatial constraints are important in this grouping because, during a recognition process, they allow for the reduction of the number of hypotheses that must be verified and also reduce the number of false positives present in each of these hypotheses. A common choice for this grouping task is to use the Hough transform on the global spatial transformation parameters of the hypothesized matches.Here, instead, we use semi-local spatial constraints which allow for a greater range of shape deformations. A comparison with Hough transform shows that our method is more robust to both rigid and non-rigid deformations. Its functionality is demonstrated in an exemplar-based object recognition system that deals well with severe non-rigid deformations. We also show the efficacy of our flexible spatial grouping for long range motion problems. . IntroductionThe complexity of the image descriptor (also called indexing primitive) used for image representation in an object recognition system has a great impact on the design of a recognition system (for a thorough discussion, see [7]). Complex global/semi-local image descriptors (e.g., generalized cylinders [3], geons [2], superquadrics [17], among others) reduce the complexity of the model by decreasing the number of descriptors necessary for the representation. This allows for a sparsely populated database of model features, which causes a reduction in the complexity of the search and verification steps. However, these image descriptors are difficult to extract and sensitive to partial occlusion. Alternatively, simple local image descriptors (e.g., 2D points [13]) are easy to extract, robust to rigid deformations and partial occlusion, but sensitive to background clutter and non-rigid deformation. Unfortunately, their low distinctiveness typically results in an overpopulated database of model descriptors due to the large number of descriptors necessary to form a model. Therefore, systems based on simple local descriptors have complex search and verification steps, where the latter step depends strongly on global pose determination.In this context, there is a recent surge of interest in more complex local descriptors that aim at finding a good balance between detectability, robustness to image deformation, and distinctiveness. The goal is to increase the robustness to background clutter and to reduce the complexity of the search and verification steps. For example, in the literature we find descriptors based on: principal components analysis of image patches [8,16], Gabor filter responses [12], wavelet coefficients [23], differential invariants [22], local phase features [4], and histograms of local filter responses [14,20].Nevertheless, as the size of the database of object models grows, the false detection rates for correspondences between test image features and database features also increases...

When analysing screening mammograms, radiologists can naturally process information across two ipsilateral views of each breast, namely the cranio-caudal (CC) and mediolateral-oblique (MLO) views. These multiple related images provide complementary diagnostic information and can improve the radiologist's classification accuracy. Unfortunately, most existing deep learning systems, trained with globallylabelled images, lack the ability to jointly analyse and integrate global and local information from these multiple views. By ignoring the potentially valuable information present in multiple images of a screening episode, one limits the potential accuracy of these systems.Here, we propose a new multi-view global-local analysis method that mimics the radiologist's reading procedure, based on a global consistency learning and local co-occurrence learning of ipsilateral views in mammograms. Extensive experiments show that our model outperforms competing methods, in terms of classification accuracy and generalisation, on a large-scale private dataset and two publicly available datasets, where models are exclusively trained and tested with global labels.