Invariant Gabor-based interest points detector under geometric transformation

This paper proposes a novel approach for detecting interest points invariant to rotation and scale using Gabor wavelet. Our scale and rotational invariant interest points detector is produced based on a multi-scale representation by selecting points at different scales from a combination of multi-orientations response maps using a non-maximum suppressing technique. We present a comparative evaluation of our proposed detector with the existing detectors. Experimental results show that our proposed method outperforms other methods under different viewpoint angles and provides a comparable results to the existing methods under scale changes for a set of test images with different geometric transformations.     

基于Gabor多尺度空间的不变兴趣点检测

针对以往仿射不变兴趣点的特征尺度不能直接断定的问题, 提出一种基于Gabor多尺度空间的不变兴趣点检测算法。该算法主要包括三个步骤：应用Gabor滤波器组与图像卷积建立图像Gabor多尺度空间; 通过极大值准则检测兴趣点并直接断定特征尺度; 采用二阶矩矩阵描述兴趣点局部结构。实验结果表明, 相比较其他Hessian-Affine、MSER等算法, 该算法在图像模糊和JPEG压缩情况下可重复率和可匹配率均取得最好结果, 是一种能有效直接提取特征尺度的兴趣点检测算法。     

一种新的局部不变特征检测和描述算法

局部不变特征已经被成功地用来解决计算机视觉领域诸多实际问题.文中提出一种新的局部不变特征检测和描述算法,提取出的特征能够对旋转、尺度缩放、光照等变化,甚至弱仿射变换保持不变.一般说来,局部特征的提取分为特征检测和描述两个关键步骤.在特征检测阶段,首先在每一层尺度图像上提取Harris角点,然后在以Harris角点为中心的固定大小的搜索窗内搜索三维尺度空间的极值点作为局部特征点的位置和特征尺度,最后为每个特征点计算主方向.文中的特征检测算法具有良好的可重复率性能.在特征描述阶段,建立了梯度的距离和方向直方图来描述局部特征,文中的特征描述子不但具有良好的匹配性能,而且维数更低,十分有利于提高图像特征的匹配速度.大量的图像匹配与图像检索实验结果验证了文中算法的有效性.     

A fast and fully automatic registration approach based on point features for multi-source remote-sensing images

Automatic registration of multi-source remote-sensing images is a difficult task as it must deal with the varying illuminations and resolutions of the images, different perspectives and the local deformations within the images. This paper proposes a fully automatic and fast non-rigid image registration technique that addresses those issues. The proposed technique performs a pre-registration process that coarsely aligns the input image to the reference image by automatically detecting their matching points by using the scale invariant feature transform (SIFT) method and an affine transformation model. Once the coarse registration is completed, it performs a fine-scale registration process based on a piecewise linear transformation technique using feature points that are detected by the Harris corner detector. The registration process firstly finds in succession, tie point pairs between the input and the reference image by detecting Harris corners and applying a cross-matching strategy based on a wavelet pyramid for a fast search speed. Tie point pairs with large errors are pruned by an error-checking step. The input image is then rectified by using triangulated irregular networks (TINs) to deal with irregular local deformations caused by the fluctuation of the terrain. For each triangular facet of the TIN, affine transformations are estimated and applied for rectification. Experiments with Quickbird, SPOT5, SPOT4, TM remote-sensing images of the Hangzhou area in China demonstrate the efficiency and the accuracy of the proposed technique for multi-source remote-sensing image registration.     

Localized image watermarking based on feature points of scale-space representation

This paper proposes a novel method for content-based watermarking based on feature points of an image. At each feature point, the watermark is embedded after scale normalization according to the local characteristic scale. Characteristic scale is the maximum scale of the scale-space representation of an image at the feature point. By binding watermarking with the local characteristics of an image, resilience against affine transformations can be obtained easily. Experimental results show that the proposed method is robust against various image processing steps including affine transformations, cropping, filtering and JPEG compression.     

Scale Selection Properties of Generalized Scale-Space Interest Point Detectors

Scale-invariant interest points have found several highly successful applications in computer vision, in particular for image-based matching and recognition. This paper presents a theoretical analysis of the scale selection properties of a generalized framework for detecting interest points from scale-space features presented in Lindeberg (Int. J. Comput. Vis. 2010, under revision) and comprising:

an enriched set of differential interest operators at a fixed scale including the Laplacian operator, the determinant of the Hessian, the new Hessian feature strength measures I and II and the rescaled level curve curvature operator, as well as

an enriched set of scale selection mechanisms including scale selection based on local extrema over scale, complementary post-smoothing after the computation of non-linear differential invariants and scale selection based on weighted averaging of scale values along feature trajectories over scale.

It is shown how the selected scales of different linear and non-linear interest point detectors can be analyzed for Gaussian blob models. Specifically it is shown that for a rotationally symmetric Gaussian blob model, the scale estimates obtained by weighted scale selection will be similar to the scale estimates obtained from local extrema over scale of scale normalized derivatives for each one of the pure second-order operators. In this respect, no scale compensation is needed between the two types of scale selection approaches. When using post-smoothing, the scale estimates may, however, be different between different types of interest point operators, and it is shown how relative calibration factors can be derived to enable comparable scale estimates for each purely second-order operator and for different amounts of self-similar post-smoothing. A theoretical analysis of the sensitivity to affine image deformations is presented, and it is shown that the scale estimates obtained from the determinant of the Hessian operator are affine covariant for an anisotropic Gaussian blob model. Among the other purely second-order operators, the Hessian feature strength measure I has the lowest sensitivity to non-uniform scaling transformations, followed by the Laplacian operator and the Hessian feature strength measure II. The predictions from this theoretical analysis agree with experimental results of the repeatability properties of the different interest point detectors under affine and perspective transformations of real image data. A number of less complete results are derived for the level curve curvature operator.     

Key Points-based Indexing for Pre-attentive Similarities: The KIWI System

In this paper, we introduce the general architecture of an image-search engine based on pre-attentive similarities. Local features are computed in key points to represent local properties of the images. The location of key points, where local features are computed, is discussed. We present two new key point detectors designed for image retrieval, both based on multi-resolution: the contrast-based point detector, and the wavelet-based point detector. Four different local features are used in our system: differential invariants, texture, shape and colour. The local information computed in each key point is stored in 2D histograms to allow fast querying. We study the choice of the key points detector depending on the feature used, for different test sets. The Harris corner detector is used for benchmarking. Uniformly distributed points are also used, and we conclude for which applications they are effective. Finally, we show that point detector and feature efficiency depend upon the test set studied.     

基于轮廓的旋转和尺度不变区域的检测

为了获取鲁棒的特征区域,提出了一种基于轮廓的旋转和尺度不变区域的检测算法。算法应用多尺度乘积LoG(Laplacian of Gaussian)提取轮廓上稳定的角点作为特征点。根据角平分线的旋转和尺度不变性提取特征方向,利用特征方向求得特征半径。由角点、特征方向和特征半径构造不变区域。进行了特征匹配的实验,图像集包含旋转、尺度、仿射、光照和压缩五种变换,算法获得了很好的匹配结果。结果表明算法简单快速,具有较强的鲁棒性和广泛的应用性。