A 2-D Histogram Representation of Images for Pooling
نویسندگان
چکیده
Designing a suitable image representation is one of the most fundamental issues of computer vision. There are three steps in the popular Bag of Words based image representation: feature extraction, coding and pooling. In the final step, current methods make an M × K encoded feature matrix degraded to a K-dimensional vector (histogram), where M is the number of features, and K is the size of the codebook: information is lost dramatically here. In this paper, a novel pooling method, based on 2-D histogram representation, is proposed to retain more information from the encoded image features. This pooling method can be easily incorporated into stateof-the-art computer vision system frameworks. Experiments show that our approach improves current pooling methods, and can achieve satisfactory performance of image classification and image reranking even when using a small codebook and costless linear SVM.
منابع مشابه
A General Framework for 1-D Histogram-baesd Image Contrast Enhancement
In this paper, a general framework for image contrast enhancement algorithm based on an optimization problem is presented. Through this optimization, the intensities can be better distributed. The algorithm is based on the facts that the histogram of the enhanced image is close to the input image histogram and uniform distribution, simultaneously. Based on this fact, we obtain a closed form opt...
متن کاملA Convolutional Neural Network based on Adaptive Pooling for Classification of Noisy Images
Convolutional neural network is one of the effective methods for classifying images that performs learning using convolutional, pooling and fully-connected layers. All kinds of noise disrupt the operation of this network. Noise images reduce classification accuracy and increase convolutional neural network training time. Noise is an unwanted signal that destroys the original signal. Noise chang...
متن کاملColor scene transform between images using Rosenfeld-Kak histogram matching method
In digital color imaging, it is of interest to transform the color scene of an image to the other. Some attempts have been done in this case using, for example, lαβ color space, principal component analysis and recently histogram rescaling method. In this research, a novel method is proposed based on the Resenfeld and Kak histogram matching algorithm. It is suggested that to transform the color...
متن کاملPCANet-II: When PCANet Meets the Second Order Pooling
PCANet, as one noticeable shallow network, employs the histogram representation for feature pooling. However, there are three main problems about this kind of pooling method. First, the histogram-based pooling method binarizes the feature maps and leads to inevitable discriminative information loss. Second, it is difficult to effectively combine other visual cues into a compact representation, ...
متن کاملEarly detection of MS in fMRI images using deep learning techniques
Introduction & Objective:MS is a disease of the central nervous system in which the body makes a defensive attack on its tissues. The disease can affect the brain and spinal cord, causing a wide range of potential symptoms, including balance, movement and vision problems. MRI and fMRI images are a very important tool in the diagnosis and treatment of MS. The aim of this study was to provide...
متن کامل