Tensor Based Feature Detection for Color Images

2010

This paper proposes a new tensorial color representation, obtained by making a correspondence between color models (HSL, IHSL, HSV, RGB and CIELUV) and tensors. Based on this representation, a proposed tensorial morphological gradient (TMG), defined as the maximum dissimilarity over the neighborhood, was tested using several tensor similarity measures. Experimental results illustrate which color models are more suitable to the proposed tensorial representation and which measures give best results in the TMG computation. The watershed transform was used to demonstrate that the proposed representation and the TMG can be applied to segment color images. A quantitative analysis of segmentation results was also conducted.

2005

In this paper, we aim at the classification of the physical nature of local image structures in color images on the basis of geometrical and photometrical information. To this end, a framework is proposed to combine the local differential structure (i.e. geometrical information such as edges, corners, T-junctions etc) and color (i.e. photometrical information such as shadows, shading, illumination, highlights) in a multi-dimensional feature space. This framework is used to yield a proper classifier to classify salient image structures on the basis of their physical nature. The proposed framework is empirically verified on a set of images. From the theoretical and experimental results it is concluded that the proposed classification scheme successfully classifies local image structures robust to image translation and rotation, illumination intensity variations, and noise.

Lecture Notes in Computer Science, 2009

Colour image representation using real quaternions has shown to be very useful for linear and morphological colour filtering. This paper deals with the extension of first derivatives-based structure tensor for various quaternionic colour image representations. Classical corner and edge features are obtained from eigenvalues of the quaternionic colour structure tensors. We study the properties of invariance of the quaternion colour spatial derivatives and their robustness for feature extraction on practical examples.

Proceedings. Fourteenth International Conference on Pattern Recognition (Cat. No.98EX170)

We present in this paper a new method for matching points in stereoscopic color images, based on color di erential invariants involving only rst order derivatives of images. Our method is able to match robustly the images even if they present important transformations like rotation, range of viewpoint and change of intensity between each other. We present here a generalization of a gray level corner detector to the case of color images. This detector is robust and allows us to extract point primitives in stereoscopic images to be matched together, only with rst order derivatives. We then describe these points with our set of local color invariants, and we propose a simple and e cient scheme for matching them. The robustness of the matching against local deformations is shown using deformations of single color images, then our stereo matching scheme is evaluated using true stereo color images with viewpoint variations. The results obtained on complex scenes clearly show the pertinence of our approach. We show that color information greatly improves the matching method, according to the fact that the additional features allow us to use only derivatives till rst order.

1989

A new, computationally inexpensive, approach to texture recognition and segmentation is proposed based on a discriminating quantity called the texture energy tensor (TET). The TET is defined as the sum over a window of the products of pixel pairs at fixed vector displacement. For mathematical textures the TET gives easy to interpret results. Texture classification and segmentation using Laws masks or "tuned masks" and "texture energy" is shown to be a special case of a TET approach.

Biomedical Signal Processing and Control, 2013

We propose a method for feature extraction from clinical color images, with application in classification of skin lesions. Proposed feature extraction method is based on tensor decomposition of the clinical color image of skin lesion. Since color image is naturally represented as a three-way tensor, it is reasonable to use multi-way techniques to capture the underlying information contained in the image. Extracted features are elements of the core tensor in the corresponding multi-way decomposition, and represent spatial-spectral profile of the lesion. In contrast to common methods that exploit either texture or spectral diversity of the tumor only, the proposed approach simultaneously captures spatial and spectral characteristics. The procedure is tested on a problem of noninvasive diagnosis of melanoma from the clinical color images of skin lesions, with overall sensitivity 82.1% and specificity 86.9%. Our method compares favorably with the state of the art results reported in the literature and provides an interesting alternative to the existing approaches.

1997

Our aim is to analyze and evaluate di erent color models to be used for the purpose of 3-D object recognition by color-metric histogram matching according to the following criteria: invariance to the geometry of the object and illumination circumstances, high discriminative power, and noise robustness.

The choice of a colour space is of great importance for many computer vision algorithms (e.g. edge detection and object recognition). It induces the equivalence classes to the actual algorithms. Since there are many colour spaces available, the problem is how to automatically select the weighting to integrate the colour spaces in order to produce the best result for a particular task. In this paper we propose a method to learn these weights, while exploiting the non-perfect correlation between colour spaces of features through the principle of diversification. As a result an optimal trade-off is achieved between repeatability and distinctiveness. The resulting weighting scheme will ensure maximal feature discrimination. The method is experimentally verified for three feature detection tasks: Skin colour detection, edge detection and corner detection. In all three tasks the method achieved an optimal trade-off between (colour) invariance (repeatability) and discriminative power (distinctiveness).

The 2005 IEEE/WIC/ACM International Conference on Web Intelligence (WI'05)

General tensors can represent colour images more naturally than conventional features; however the general tensors' stability properties are not reported and remain to be a key problem. In this paper, we use the tensor minimax probability (TMPM) to prove that the tensor representation is stable. The proof is based on the random subspace method through a large number of experiments.