International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395 -0056
Volume: 04 Issue: 02 | Feb -2017
p-ISSN: 2395-0072
www.irjet.net
Survey on Local Color Image Descriptors Sneha Benny1, Cinita Mary Mathew2 P.G. Student, Department of Computer Engineering,VJCET,Vazhakulam, Kerala, India Assistant Professor, Department of Computer Engineering,VJCET, Vazhakulam, Kerala, India ---------------------------------------------------------------------***--------------------------------------------------------------------1
2
Abstract - Image descriptor is an important topic in the
(5) QR has quaternion algebra theories which can be used in color image processing.
field of image processing. Image descriptors are used to have the characteristics of an image. Global and local descriptors can be extracted according to the needs. Global descriptors consider image as a whole whereas a local descriptor describes a patch within an image. Multiple local descriptors are used to match an image. Traditional descriptors are extracted from each color channel separately or from vector representations. Color characteristics can be included using the Quaternionic representation. A detailed study in the field of local descriptors and Quaternionic representation has been done. Keywords: Quaternionic Representation, QLRBP, QWLD, QMD, SIFT
2. LITERATURE SURVEY The multiresolution approach to gray-scale and rotation invariant texture classification based on local binary patterns [1] is very simple, and efficient. The method recognizes uniform local binary patterns, which are fundamental properties of local image texture. For detecting the uniform patterns for any quantization of the angular space and for any spatial resolution it derives a generalized gray-scale and rotation invariant operator presentation. Thus a method is proposed to combine multiple operators for multiresolution analysis. Advantage of the system is computational simplicity as the operator can be realized with a lookup table and a few operations in a small neighborhood. In terms of gray-scale variations, the system is very robust since the operator is, by definition, invariant against any monotonic transformation of the gray scale, e.g., by changes in illumination intensity. If the gray-scale properties of the training and testing data are different, Gray-scale invariance is also necessary. The basic 3x3 LBP operator provided better performance in experiments. System presents a rotation invariant and grayscale texture operator based on local binary patterns. Rotation invariance is achieved by recognizing the gray-scale invariant operator that incorporates a fixed set of rotation invariant patterns. The uniform appearance of the local binary pattern is referred by the term uniform, i.e., in the circular presentation of the pattern, there are a limited number of discontinuities or transitions. The advantage of the multiresolution approach is that, it is very robust in terms of gray-scale variations and it is invariant against monotonic transformations. But whole color information is not considered as it is converted in to grayscale. The evaluation of Color Descriptors for Object and Scene Recognition [2] is based on the access visual information. So far, at salient points, for feature extraction, intensity-based descriptors have been widely used. Color descriptors have been proposed to increase illumination invariance and discriminative power. The system studies the distinctiveness and invariance properties of color descriptors in a structured way. Histograms: The RGB histogram is composed of three 1D histograms which are of R, G, and B channels of the RGB color space. Hue becomes unstable near the gray axis for HSV color histogram. The certainty of the hue is inversely proportional to the saturation. The hue histogram is made more robust. In
1.INTRODUCTION In the fields of image processing, computer vision, and pattern recognition, image descriptor is an active research topic. From the earlier geometric moment descriptor and Fourier descriptor to the later wavelet descriptor, orthogonal moment invariants, projection descriptor, and all kinds of local descriptors, a large number of image descriptors have been proposed. Its applications are in areas, such as image retrieval and matching, texture classification, face recognition, and many others. With the introduction of modern imaging equipment and devices, its applications in daily life has increased. This has increased the demand of color image descriptors as well. To extract a descriptor from the color image, two methods are there, that is either converts the color image into the corresponding gray-scale image or simply use one color component for further processing. Here, some color information has been ignored as we are not considering the color image as a whole. Extract the descriptor from each color component of a color image respectively or from a vector representation of the color image, in order to address this problem. Thus, a color image representation, called Quaternionic representation (QR), has been developed and used in order to represent the three separate color channel values into a single quaternion. QR encodes all color components of a pixel using a quaternion number. The advantages of QR over RGB representation are: (1) QR combines all color channels of an image, (2) Relatively lower computational complexity is achieved compared to other vector approaches, (3) QR has an implementation of vector cross correlation, (4) Transformations in 3D or 4D space is performed
Š 2017, IRJET
|
Impact Factor value: 5.181
|
ISO 9001:2008 Certified Journal
|
Page 810