Co-occurrence Histograms Of Oriented Gradients Research Articles

Boosting the Performance of the BoVW Model Using SURF–CoHOG-Based Sparse Features with Relevance Feedback for CBIR

The content-based image retrieval (CBIR) based on sparse visual features is a challenging research problem to categorize images into semantically meaningful classes. Different robust feature representation methods have been proposed to address the semantic gap issue of CBIR. In this article, we propose a novel robust image representation for CBIR, which is based on complementary visual words integration of speeded up robust features (SURF) and co-occurrence histograms of oriented gradients (CoHOG). The SURF is a local feature descriptor, while CoHOG is a global feature descriptor. The local features give better performance for images belonging to different semantic classes and having close visual appearance among their visual contents, while global features give better performance to retrieval images at large scale. To ensure image retrieval accuracy, the proposed method build two smaller size dictionaries, each containing visual words of SURF and CoHOG descriptors, which are assimilated to form one larger size dictionary. In the CBIR system, a dictionary of smaller size produces better sensitivity, while a dictionary of larger sizes produces better specificity. The dictionary of the larger sizes also yields an overfitting problem to affect CBIR performance, which is addressed by introducing a linear discriminant analysis (LDA) and relevance feedback with the proposed method. The comparative performance analysis of the proposed method is performed with the competitor CBIR methods (i.e., feature integration of SURF–CoHOG descriptors’ based on the bag-of-visual-words (BoVW), a single feature of SURF–BoVW and CoHOG–BoVW methods). The quantitative and qualitative analysis carried out on four standard image databases proves the robust performance of the proposed method as compared to its competitor and recent CBIR methods.

Segmentation Based Word Spotting Method for Handwritten Documents

In this paper, we present a segmentation based word spotting method for handwritten document images using Co-occurrence Histograms of Oriented Gradients (Co-HOG) descriptor. The drawback of Histogram of Oriented Gradients (HOG) is that HOG ignores spatial information of adjacent pixels where as the Co-HOG take into account spatial contextual information by capturing the co-occurrence of orientation pairs of neighbouring pixels. In order to construct Co-HOG descriptor for word spotting, we divide a word image into blocks and Co-HOG features are extracted from each block and finally concatenate them to form a feature descriptor. The proposed method is evaluated using precision and recall metrics through experimentation conducted on popular GW dataset and confirmed that our method outperforms for this dataset.

Principal curvatures based rotation invariant algorithms for efficient texture classification

The histograms of oriented gradients (HOG) and co-occurrence HOG (CoHOG) algorithms are simple and intuitive descriptors. However, the HOG and CoHOG algorithms based on gradient computation still have some shortcomings: they ignore meaningful textural properties and are unstable to noise. In this paper, two new efficient HOG and CoHOG methods are proposed. The proposed algorithms are based on the Gaussian derivative filters, and the feature vectors are obtained by means of principal curvatures. The feature vectors are rotation invariant by means of the rotation invariance characteristic of principal curvatures (i.e. eigenvalues). The experimental results on the CUReT, KTH-TIPS, KTH-TIPS2-a, UIUC, Brodatz album, Kylberg and Xu datasets confirm that the developed algorithms have higher classification rates than state-of-the-art texture classification methods. The classification results also demonstrate that the developed algorithms are more stable to noise and rotation than the original HOG and CoHOG algorithms.

Continuous rotation invariant features for gradient-based texture classification

Extracting rotation invariant features is a valuable technique for the effective classification of rotation invariant texture. The Histograms of Oriented Gradients (HOG) algorithm has been proved to be theoretically simple, and has been applied in many areas. Also, the co-occurrence HOG (CoHOG) algorithm provides a unified description including both statistical and differential properties of a texture patch. However, HOG and CoHOG have some shortcomings: they discard some important texture information and are not invariant to rotation. In this paper, based on the original HOG and CoHOG algorithms, four novel feature extraction methods are proposed. The first method uses Gaussian derivative filters named GDF-HOG. The second and the third methods use eigenvalues of the Hessian matrix named Eig(Hess)-HOG and Eig(Hess)-CoHOG, respectively. The fourth method exploits the Gaussian and means curvatures to calculate curvatures of the image surface named GM-CoHOG. We have empirically shown that the proposed novel extended HOG and CoHOG methods provide useful information for rotation invariance. The classification results are compared with original HOG and CoHOG algorithms methods on the CUReT, KTH-TIPS, KTH-TIPS2-a and UIUC datasets show that proposed four methods achieve best classification result on all datasets. In addition, we make a comparison with several well-known descriptors. The experiments of rotation invariant analysis are carried out on the Brodatz dataset, and promising results are obtained from those experiments.

Multi-Cascade of Complementary Features for Fast and Accurate Pedestrian Detection

We propose a cascade of two complementary features to detect pedestrians from static images quickly and accurately. Co-occurrence Histograms of Oriented Gradients (CoHOG) descriptors have a strong classification capability but are extremely high dimensional. On the other hand, Haar-like features are computationally efficient but not highly discriminative for extremely varying texture and shape information such as pedestrians with different clothing and stances. Therefore, the combination of both features enables fast and accurate pedestrian detection. Our framework comprises a cascade of Haar based weak classifiers followed by a CoHOG-SVM classifier. The experimental results on the DaimlerChrysler and INRIA benchmark datasets show that we can reach very close accuracy to the most accurate CoHOG-only classifier but in less than 1/200 of its computational cost. Additionally, we show that by integrating two of our proposed cascades: one full body with another upper body detectors, we can reach higher accuracy than the standalone full body CoHOG-only in about 1/100 of its computational cost.

Robust CoHOG Feature Extraction in Human-Centered Image/Video Management System

Many human-centered image and video management systems depend on robust human detection. To extract robust features for human detection, this paper investigates the following shortcomings of co-occurrence histograms of oriented gradients (CoHOGs) which significantly limit its advantages: 1) The magnitudes of the gradients are discarded, and only the orientations are used; 2) the gradients are not smoothed, and thus, aliasing effect exists; and 3) the dimensionality of the CoHOG feature vector is very large (e.g., 200,000). To deal with these problems, in this paper, we propose a framework that performs the following: 1) utilizes a novel gradient decomposition and combination strategy to make full use of the information of gradients; (2) adopts a two-stage gradient smoothing scheme to perform efficient gradient interpolation; and (3) employs incremental principal component analysis to reduce the large dimensionality of the CoHOG features. Experimental results on the two different human databases demonstrate the effectiveness of the proposed method.

Co-occurrence Histograms of Oriented Gradients for Human Detection

The purpose of the work reported in this paper is to detect humans from images. This paper proposes a method for extracting feature descriptors consisting of co-occurrence histograms of oriented gradients (CoHOG). Including co-occurrence with various positional offsets, the feature descriptors can express complex shapes of objects with local and global distributions of gradient orientations. Our method is evaluated with a simple linear classifier on two well-known human detection benchmark datasets: “DaimlerChrysler pedestrian classification benchmark dataset” and “INRIA person data set”. The results show that our method reduces the miss rate by half compared with HOG, and outperforms the state-of-the-art methods on both datasets. Furthermore, as an example of a practical application, we applied our method to a surveillance video eight hours in length. The result shows that our method reduces false positives by half compared with HOG. In addition, CoHOG can be calculated 40% faster than HOG.