Chun-Guang Li

Pairwise Rotation Invariant Co-Occurrence Local Binary Pattern

Designing effective features is a fundamental problem in computer vision. However, it is usually difficult to achieve a great tradeoff between discriminative power and robustness. Previous works shown that spatial co-occurrence can boost the discriminative power of features. However the current existing co-occurrence features are taking few considerations to the robustness and hence suffering from sensitivity to geometric and photometric variations. In this work, we study the Transform Invariance (TI) of co-occurrence features. Concretely we formally introduce a Pairwise Transform Invariance (PTI) principle, and then propose a novel Pairwise Rotation Invariant Co-occurrence Local Binary Pattern (PRICoLBP) feature, and further extend it to incorporate multi-scale, multi-orientation, and multi-channel information. Different from other LBP variants, PRICoLBP can not only capture the spatial context co-occurrence information effectively, but also possess rotation invariance. We evaluate PRICoLBP comprehensively on nine benchmark data sets from five different perspectives, e.g., encoding strategy, rotation invariance, the number of templates, speed, and discriminative power compared to other LBP variants. Furthermore we apply PRICoLBP to six different but related applications-texture, material, flower, leaf, food, and scene classification, and demonstrate that PRICoLBP is efficient, effective, and of a well-balanced tradeoff between the discriminative power and robustness.

HCL2000 - A Large-scale Handwritten Chinese Character Database for Handwritten Character Recognition

In this paper, we present a large scale off-line handwritten Chinese character database-HCL2000 which will be made public available for the research community. The database contains 3,755 frequently used simplified Chinesecharacters written by 1,000 different subjects. The writers’ information is incorporated in the database to facilitate testing on grouping writers with different background such as age, occupation, gender, and education etc. We investigate some characteristics of writing styles from different groups of writers. We evaluate HCL2000 database using three different algorithms as a baseline. We decide to publish the database along with this paper and make it free for a research purpose.

Local Sparse Representation Based Classification

In this paper, we address the computational complexity issue in Sparse Representation based Classification (SRC). In SRC, it is time consuming to find a global sparse representation. To remedy this deficiency, we propose a Local Sparse Representation based Classification (LSRC) scheme, which performs sparse decomposition in local neighborhood. In LSRC, instead of solving the L1-norm constrained least square problem for all of training samples we solve a similar problem in a local neighborhood for each test sample. Experiments on face recognition data sets ORL and Extended Yale B demonstrated that the proposed LSRC algorithm can reduce the computational complexity and remain the comparative classification accuracy and robustness.

Dynamic texture and scene classification by transferring deep image features

Dynamic texture and scene classification are two fundamental problems in understanding natural video content. Extracting robust and effective features is a crucial step towards solving these problems. However, the existing approaches suffer from the sensitivity to either varying illumination, or viewpoint changes, or even camera motion, and/or the lack of spatial information. Inspired by the success of deep structures in image classification, we attempt to leverage a deep structure to extract features for dynamic texture and scene classification. To tackle with the challenges in training a deep structure, we propose to transfer some prior knowledge from image domain to video domain. To be more specific, we propose to apply a well-trained Convolutional Neural Network (ConvNet) as a feature extractor to extract mid-level features from each frame, and then form the video-level representation by concatenating the first and the second order statistics over the mid-level features. We term this two-level feature extraction scheme as a Transferred ConvNet Feature (TCoF). Moreover, we explore two different implementations of the TCoF scheme, i.e., the spatial TCoF and the temporal TCoF. In the spatial TCoF, the mean-removed frames are used as the inputs of the ConvNet; whereas in the temporal TCoF, the differences between two adjacent frames are used as the inputs of the ConvNet. We evaluate systematically the proposed spatial TCoF and the temporal TCoF schemes on three benchmark data sets, including DynTex, YUPENN, and Maryland, and demonstrate that the proposed approach yields superior performance.

Multi-scale Joint Encoding of Local Binary Patterns for Texture and Material Classification.

In the current multi-scale LBP (MS-LBP) on texture and material classification, each scale is encoded into histograms individually. This strategy ignores the correlation between different scales, and loses a lot of discriminative information. In this paper , we propose a novel and effective multi-scale joint encoding of local binary patterns (MSJLBP) for texture and material classification. In MSJ-LBP, the joint encoding strategy can capture the correlation between different scales and hence depict richer local structures. In addition, the proposed MSJ-LBP is computationally simple and rotation invariant. Extensive experiments on four challenging databases (Outex_TC_00012, Brodatz, KTH-TIPS, KTH-TIPS2a) show that the proposed MSJ-LBP significantly outperforms the classical MS-LBP and achieves the state-of-the-art performance.

Learning Semi-Supervised Representation Towards a Unified Optimization Framework for Semi-Supervised Learning

State of the art approaches for Semi-Supervised Learning (SSL) usually follow a two-stage framework -- constructing an affinity matrix from the data and then propagating the partial labels on this affinity matrix to infer those unknown labels. While such a two-stage framework has been successful in many applications, solving two subproblems separately only once is still suboptimal because it does not fully exploit the correlation between the affinity and the labels. In this paper, we formulate the two stages of SSL into a unified optimization framework, which learns both the affinity matrix and the unknown labels simultaneously. In the unified framework, both the given labels and the estimated labels are used to learn the affinity matrix and to infer the unknown labels. We solve the unified optimization problem via an alternating direction method of multipliers combined with label propagation. Extensive experiments on a synthetic data set and several benchmark data sets demonstrate the effectiveness of our approach.

Exploring Cross-Channel Texture Correlation for Color Texture Classification.

This paper proposes a novel approach to encode cross-channel texture correlation for color texture classification task. Firstly, we quantitatively study the correlation between different color channels using Local Binary Pattern (LBP) as the texture descriptor and using Shannon’s information theory to measure the correlation. We find that (R, G) channel pair exhibits stronger correlation than (R, B) and (G, B) channel pairs. Secondly, we propose a novel descriptor to encode the cross-channel texture correlation. The proposed descriptor can capture well the relative variance of texture patterns between different channels. Meanwhile, our descriptor is computationally efficient and robust to image rotation. We conduct extensive experiments on four challenging color texture databases to validate the effectiveness of the proposed approach. The experimental results show that the proposed approach significantly outperforms its mostly relevant counterpart (Multichannel color LBP), and achieves the state-of-the-art performance.

An evaluation on different graphs for semi-supervised learning

Graph-based Semi-Supervised Learning (SSL) has been an active topic in machine learning for about a decade. It is well-known that how to construct the graph is the central concern in recent work since an efficient graph structure can significantly boost the final performance. In this paper, we present a review on several different graphs for graph-based SSL at first. And then, we conduct a series of experiments on benchmark data sets in order to give a comprehensive evaluation on the advantageous and shortcomings for each of them. Experimental results shown that: a) when data lie on independent subspaces and the number of labeled data is enough, the low-rank representation based method performs best, and b) in the majority cases, the local sparse representation based method performs best, especially when the number of labeled data is few.

Bases sorting: Generalizing the concept of frequency for over-complete dictionaries

We propose an algorithm, called Bases Sorting, to sort the bases of over-complete dictionaries used in sparse representation according to the magnitudes of coefficients when representing the training samples. Then the bases are considered to be ordered from low to high frequencies, thus generalizing the traditional concept of frequency for over-complete dictionaries. Applications are also shown.

Dimensionality reduction by low-rank embedding

We consider the dimensionality reduction task under the scenario that data vectors lie on (or near by) multiple independent linear subspaces. We propose a robust dimensionality reduction algorithm, named as Low-Rank Embedding(LRE). In LRE, the affinity weights are calculated via low-rank representation and the embedding is yielded by spectral method. Owing to the affinity weight induced from low-rank model, LRE can reveal the subtle multiple subspace structure robustly. In the virtual of spectral method, LRE transforms the subtle multiple subspaces structure into multiple clusters in the low dimensional Euclidean space in which most of the ordinary algorithms can perform well. To demonstrate the advantage of the proposed LRE, we conducted comparative experiments on toy data sets and benchmark data sets. Experimental results confirmed that LRE is superior to other algorithms.