International Journal of Pattern Recognition and Artificial Intelligence

When several classifiers are brought to contribute to the same task of recognition, various strategies of decisions, implying these classifiers in different ways, are possible. A first strategy consists in deciding using different opinions: it corresponds to the combination of classifiers. A second strategy consists in using one or more opinions for better guiding other classifiers in their training stages, and/or to improve the decision-making of other classifiers in the classification stage: it corresponds to the cooperation of classifiers. The third and last strategy consists in giving more importance to one or more classifiers according to various criteria or situations: it corresponds to the selection of classifiers. The temporal aspect of Pattern Recognition (PR), i.e. the possible evolution of the classes to be recognized, can be treated by the strategy of selection.

In this paper, a novel image projection analysis method (UIPDA) is first developed for image feature extraction. In contrast to Liu's projection discriminant method, UIPDA has the desirable property that the projected feature vectors are mutually uncorrelated. Also, a new LDA technique called EULDA is presented for further feature extraction. The proposed methods are tested on the ORL and the NUST603 face databases. The experimental results demonstrate that: (i) UIPDA is superior to Liu's projection discriminant method and more efficient than Eigenfaces and Fisherfaces; (ii) EULDA outperforms the existing PCA plus LDA strategy; (iii) UIPDA plus EULDA is a very effective two-stage strategy for image feature extraction.

In this paper, we propose a new method for extracting salient regions and learning their importance scores in region-based image retrieval. In Region-Based Image Retrieval (RBIR), not all the regions are important for retrieving similar images and rather, in retrieval, the user is often interested in performing a query on only one or a few regions rather than the whole image. Therefore, for a successful retrieval system, it is an important issue to specify which regions are important for retrieving an image. To extract salient regions from images automatically, we make three assumptions and determine salient regions with their importance scores. In this paper, we apply the relevance feedback algorithm to the matching process as two different purposes: one is for updating importance scores of salient regions and the other is for updating weights of feature vectors. By using our relevance feedback method, the matching process can improve retrieval performance interactively and allow progressive refinement of query results according to the user's feedback action. Through experiments and comparison with other methods, our proposed method shows good performance as well as easy and semantic interface for region-based image retrieval. The efficacy of our method is validated using a set of 3000 images from Corel-photo CD.

In this paper, we propose a thorough investigation of a nearest neighbor rule which we call the "Symmetric Nearest Neighbor (sNN) rule". Basically, it symmetrises the classical nearest neighbor relationship from which are computed the points voting for some instances. Experiments on 29 datasets, most of which are readily available, show that the method significantly outperforms the traditional Nearest Neighbors methods. Experiments on a domain of interest related to tropical pollution normalization also show the greater potential of this method. We finally discuss the reasons for the rule's efficiency, provide methods for speeding-up the classification time, and derive from the sNN rule a reliable and fast algorithm to fix the parameter k in the k-NN rule, a longstanding problem in this field.

In this paper, a coarse-to-fine hierarchical classification method based on the features derived from adaptive cellular color decomposition is proposed. The proposed method is general and can be applied to all kinds of color image databases as long as a sample set of images have been classified. In addition, the number of classes can be as versatile as required. To achieve the goal mentioned above, our method consists of two phases: color quantization and classification. In the color quantization step, cellular decomposition is used to adaptively quantize color images in the HSV color space since H and S components construct a hexagon structure that is same as the cellular pattern. In the classification step, a coarse-to-fine strategy is employed. In the coarse stage, five image-based features extracted directly from the quantization results of the query images are used to prune irrelevant database images. In the fine stage, two cluster-based features are extracted from a small set of candidate images using closest-cluster matching. On the other hand, according to feature evaluation, one image-based and two cluster-based features are selected to derive individual-based similarity measure, which, in turn, is used to measure image-to-image similarity. In addition, class-based similarity measure using class characteristics is proposed to evaluate image-to-class similarity. Candidate images are then sorted according to the similarity measure, which is a combination of individual-based and class-based similarity measures. Finally, k-NN rule is used to assign the query image to a single class according to the sorting results. The effectiveness and practicability of the proposed method have been demonstrated by various experimental results.

In this paper, an image feature of color differences on edges in spiral scan order (CDESSO) is presented. This proposed CDESSO feature can characterize the principal pixel colors, color complexity and color differences among adjacent objects in an image. In addition, this paper employs the CDESSO feature to develop an image retrieval system. The CDESSO-based image retrieval system can provide a high accuracy rate in finding the database images that satisfy the users' requirement. Besides, it can also resist the scale variants of images as well as the shift and rotation variants of objects in images.

Naïve models of V1 simple cells, based on principal components analysis (PCA), have poor performance and do not form a distributed population code, limiting their scientific value and practical application in pattern recognition of natural scenes. This paper evaluates two strategies for enhancing the validity of PCA models that have been applied to comparable nonorthogonal models: pre-cortical "whitening", and local decomposition of visual features. The goal is to form a distributed population code. Four sets (landscapes, trees and plants, people and animals, and books and buildings) of sixteen natural images (256 × 256 pixels) were whitened and decomposed into 160 randomly selected local segments of either 8 × 8 or 16 × 16 pixels. The variance distribution of each representation was evaluated using a "distributed coding efficiency index" (DCE). Highly significant increases in DCE were observed for the principal components of local, whitened image segments compared to whole, nonwhitened images. This suggests that whitening and local decomposition may improve PCA performance for natural image processing, potentially providing more accurate simple cell models. In addition, the comparatively good performance of more complex, nonorthogonal models may be partially explained by their use of whitening and local decomposition.

We consider the difficult problem of identification of independent causes from a mixture of them when these causes interfere with one another in a particular manner: those considered are visual inputs to a neural network system which are created by independent underlying causes which may occlude each other. The prototypical problem in this area is a mixture of horizontal and vertical bars in which each horizontal bar interferes with the representation of each vertical bar and vice versa. Previous researchers have developed artificial neural networks which can identify the individual causes; we seek to go further in that we create artificial neural networks which identify all the horizontal bars from only such a mixture. This task is a necessary precursor to the development of the concept of "horizontal" or "vertical".

In this paper, we will present a method to extract slow-motion replays in a soccer game video. According to our observation, a slow-motion replay always happens as a single shot and appears between two gradual transitions. Based on this fact, a video will first be segmented into individual shots and only those shots appearing between two gradual transitions are considered as candidate shots. There are two kinds of slow-motion replays. One comes from a standard camera and consists of some repeating or inserted frames. The other is from a high-speed camera with larger variation between two consecutive frames. Based on these features, an extractor is then provided to check if a candidate shot is a slow motion replay. Some experimental results are given to show the effectiveness of the proposed method.

This paper proposes a new approach to extracting natural strokes from the skeletons of loosely-constrained, off-line handwritten Chinese characters. It admits the output substrokes from a previously proposed fuzzy substroke extractor as its inputs. By identifying a number of expected ambiguities which include mutual similarities, unstable touches and joint/cross distortions, fuzzy stroke models are constructed and a "hit-all" fuzzy stroke matching strategy is pursued. Fuzzy partitioning technique is used to generate a ranked list of consistent stroke sets from the set of fuzzy strokes being identified. With this approach, a maximum of 20 distinct natural stroke classes can be extracted from each input character, together with an estimate on the actual count of strokes which compose the character. Our system offers a number of performance tuning capabilities such as the computation of the fuzzy scores of each extracted stroke, the adjustment on the fuzzy stroke model parameters, and the potential of incorporating one's personal writing styles into our methodology.