Boosting color saliency in image feature detection

Abstract

Abstract-The aim of salient feature detection is to find distinctive local events in images. Salient features are generally determined from the local differential structure of images. They focus on the shape-saliency of the local neighborhood. The majority of these detectors are luminance-based, which has the disadvantage that the distinctiveness of the local color information is completely ignored in determining salient image features. To fully exploit the possibilities of salient point detection in color images, color distinctiveness should be taken into account in addition to shape distinctiveness. In this paper, color distinctiveness is explicitly incorporated into the design of saliency detection. The algorithm, called color saliency boosting, is based on an analysis of the statistics of color image derivatives. Color saliency boosting is designed as a generic method easily adaptable to existing feature detectors. Results show that substantial improvements in information content are acquired by targeting color salient features.

Evaluation of color descriptors for object and scene recognition

“…Hue histogram In the HSV color space, it is known that the hue becomes unstable around the grey axis. To this end, Van de Weijer et al [21] apply an error analysis to the hue. The analysis shows that the certainty of the hue is inversely proportional to the saturation.…”

“…Therefore, the descriptor is only partially invariant to light color changes. HueSIFT Van de Weijer et al [21] introduce a concatenation of the hue histogram (see section 3.1) with the SIFT descriptor. When compared to HSV-SIFT, the usage of the weighed hue histogram addresses the instability of the hue around the grey axis.…”

“…In addition, changes in the illumination of a scene can greatly affect the performance of object recognition if the descriptors used are not robust to these changes. To increase illumination invariance and discriminative power, color descriptors have been proposed [2,8,21]. However, as there are many different methods, a taxonomy is required based on principles of illumination changes to arrange color invariant descriptors in the context of image category recognition.…”

Focus of Attention and Region Segregation by Low-Level Geometry

“…Their model was later extended by Itti and Koch (2001), who integrated more features, for instance intensity, edge orientation and motion. In our approach to create a saliency model which also contains cues for region and object segregation, we therefore start by using colour information, as this provides the most important input for attention (van de Weijer et al, 2006), in order to build a colour conspicuity map which will later be combined with a texture map. But before using colour features the input images must be corrected because a same object will look different when illuminated by different light sources, i.e., the number, power and spectra of these.…”

“…In this paper we concentrate on three aspects: (1) the construction of a saliency map for FoA on the basis of colour, which was shown to be very effective in attracting attention (van de Weijer et al, 2006), also texture (du Buf, 2007), (2) a first region segregation by employing low-level geometry in terms of blobs, bars and corners, and (3) using low-level geometry allows us to reduce significantly the dimensionality of texture features. We note that our approach is not based on the cortical multi-scale keypoint representation as recently proposed by Rodrigues and du Buf (2006), who built saliency maps which work very well for the detection of facial landmarks and for invariant object recognition on homogeneous backgrounds (Rodrigues and du Buf, 2007), but may lead to enormous amounts of local peaks in natural scenes.…”