BOG: An extension of HOG by interpreting it as bag of features
نویسندگان
چکیده
Histogram of orientated gradient (HOG) is widely used as a local feature descriptor in bag of features (BOF) method, whereas, few studies are conducted to discover the relationship between them. In this paper, we exploit this relationship and reveal that the construction method of descriptor in blocks in HOG can be treated as a variant of BOF method. Based on this interpretation, we propose a new descriptor termed as bag of gradient (BOG), which can be viewed as an extension of HOG, by incorporating principles used in BOF, such as the preservation of locality. Experiment results show that BOG significantly reduces the error rate in comparison to HOG in pedestrian detection.
منابع مشابه
An Analysis of Achievement of the Philosophical Sense of “Extension” in Music, with Interpretaion of Ibn-e Sina’s Explanation an Extension
This research can be considered as one of the studies that seek to explore, in an argumentative way, subtle and solid philosophical concepts in the field of art. The paper provides an analysis of the concept of “extension” in music as one of the most thought-provoking philosophical concepts. The analysis is carried out by interpreting Ibn-Sina’s special conception of musical extension to answer...
متن کاملVisual Tracking using Learning Histogram of Oriented Gradients by SVM on Mobile Robot
The intelligence of a mobile robot is highly dependent on its vision. The main objective of an intelligent mobile robot is in its ability to the online image processing, object detection, and especially visual tracking which is a complex task in stochastic environments. Tracking algorithms suffer from sequence challenges such as illumination variation, occlusion, and background clutter, so an a...
متن کاملPerformance evaluation of block-based copy- move image forgery detection algorithms
Copy-move forgery is a particular type of distortion where a part or portions of one image is/are copied to other parts of the same image. This type of manipulation is done to hide a particular part of the image or to copy one or more objects into the same image. There are several methods for detecting copy-move forgery, including block-based and key point-based methods. In this paper, a method...
متن کاملRecognizing Handwritten Characters with Local Descriptors and Bags of Visual Words
In this paper we propose the use of several feature extraction methods, which have been shown before to perform well for object recognition, for recognizing handwritten characters, These methods are the histogram of oriented gradients (HOG), a bag of visual words using pixel intensity information (BOW), and a bag of visual words using extracted HOG features (HOG-BOW). These feature extraction a...
متن کاملA Novel Method for Content Base Image Retrieval Using Combination of Local and Global Features
Content-based image retrieval (CBIR) has been an active research topic in the last decade. In this paper we proposed an image retrieval method using global and local features. Firstly, for local features extraction, SURF algorithm produces a set of interest points for each image and a set of 64-dimensional descriptors for each interest points and then to use Bag of Visual Words model, a cluster...
متن کامل