Fast CU Encoding Schemes Based on Merge Mode and Motion Estimation for HEVC Inter Prediction

( Jinfu Wu ),( Baolong Guo ),( Jie Hou ),( Yunyi Yan ),( Jie Jiang ) 한국인터넷정보학회 2016 KSII Transactions on Internet and Information Syst Vol.10 No.3

The emerging video coding standard High Efficiency Video Coding (HEVC) has shown almost 40% bit-rate reduction over the state-of-the-art Advanced Video Coding (AVC) standard but at about 40% computational complexity overhead. The main reason for HEVC computational complexity is the inter prediction that accounts for 60%-70% of the whole encoding time. In this paper, we propose several fast coding unit (CU) encoding schemes based on the Merge mode and motion estimation information to reduce the computational complexity caused by the HEVC inter prediction. Firstly, an early Merge mode decision method based on motion estimation (EMD) is proposed for each CU size. Then, a Merge mode based early termination method (MET) is developed to determine the CU size at an early stage. To provide a better balance between computational complexity and coding efficiency, several fast CU encoding schemes are surveyed according to the rate-distortion-complexity characteristics of EMD and MET methods as a function of CU sizes. These fast CU encoding schemes can be seamlessly incorporated in the existing control structures of the HEVC encoder without limiting its potential parallelization and hardware acceleration. Experimental results demonstrate that the proposed schemes achieve 19%-46% computational complexity reduction over the HEVC test model reference software, HM 16.4, at a cost of 0.2%-2.4% bit-rate increases under the random access coding configuration. The respective values under the low-delay B coding configuration are 17%-43% and 0.1%-1.2%.

A Sparse Target Matrix Generation Based Unsupervised Feature Learning Algorithm for Image Classification

( Dan Zhao ),( Baolong Guo ),( Yunyi Yan ) 한국인터넷정보학회 2018 KSII Transactions on Internet and Information Syst Vol.12 No.6

Unsupervised learning has shown good performance on image, video and audio classification tasks, and much progress has been made so far. It studies how systems can learn to represent particular input patterns in a way that reflects the statistical structure of the overall collection of input patterns. Many promising deep learning systems are commonly trained by the greedy layerwise unsupervised learning manner. The performance of these deep learning architectures benefits from the unsupervised learning ability to disentangling the abstractions and picking out the useful features. However, the existing unsupervised learning algorithms are often difficult to train partly because of the requirement of extensive hyperparameters. The tuning of these hyperparameters is a laborious task that requires expert knowledge, rules of thumb or extensive search. In this paper, we propose a simple and effective unsupervised feature learning algorithm for image classification, which exploits an explicit optimizing way for population and lifetime sparsity. Firstly, a sparse target matrix is built by the competitive rules. Then, the sparse features are optimized by means of minimizing the Euclidean norm (L₂) error between the sparse target and the competitive layer outputs. Finally, a classifier is trained using the obtained sparse features. Experimental results show that the proposed method achieves good performance for image classification, and provides discriminative features that generalize well.

Image Retrieval Method Based on IPDSH and SRIP

( Xu Zhang ),( Baolong Guo ),( Yunyi Yan ),( Wei Sun ),( Meng Yi ) 한국인터넷정보학회 2014 KSII Transactions on Internet and Information Syst Vol.8 No.5

At present, the Content-Based Image Retrieval (CBIR) system has become a hot research topic in the computer vision field. In the CBIR system, the accurate extractions of low-level features can reduce the gaps between high-level semantics and improve retrieval precision. This paper puts forward a new retrieval method aiming at the problems of high computational complexities and low precision of global feature extraction algorithms. The establishment of the new retrieval method is on the basis of the SIFT and Harris (APISH) algorithm, and the salient region of interest points (SRIP) algorithm to satisfy users` interests in the specific targets of images. In the first place, by using the IPDSH and SRIP algorithms, we tested stable interest points and found salient regions. The interest points in the salient region were named as salient interest points. Secondary, we extracted the pseudo-Zernike moments of the salient interest points` neighborhood as the feature vectors. Finally, we calculated the similarities between query and database images. Finally, We conducted this experiment based on the Caltech-101 database. By studying the experiment, the results have shown that this new retrieval method can decrease the interference of unstable interest points in the regions of non-interests and improve the ratios of accuracy and recall.

A Fast TU Size Decision Method for HEVC RQT Coding

( Jinfu Wu ),( Baolong Guo ),( Yunyi Yan ),( Jie Hou ),( Dan Zhao ) 한국인터넷정보학회 2015 KSII Transactions on Internet and Information Syst Vol.9 No.6

The emerging high efficiency video coding (HEVC) standard adopts the quadtree-structured transform unit (TU) in the residual quadtree (RQT) coding. Each TU allows to be split into four equal sub-TUs recursively. The RQT coding is performed for all the possible transform depth levels to achieve the highest coding efficiency, but it requires a very high computational complexity for HEVC encoders. In order to reduce the computational complexity requested by the RQT coding, in this paper, we propose a fast TU size decision method incorporating an adaptive maximum transform depth determination (AMTD) algorithm and a full check skipping - early termination (FCS-ET) algorithm. Because the optimal transform depth level is highly content-dependent, it is not necessary to perform the RQT coding at all transform depth levels. By the AMTD algorithm, the maximum transform depth level is determined for current treeblock to skip those transform depth levels rarely used by its spatially adjacent treeblocks. Additionally, the FCS-ET algorithm is introduced to exploit the correlations of transform depth level between four sub-CUs generated by one coding unit (CU) quadtree partitioning. Experimental results demonstrate that the proposed overall algorithm significantly reduces on average 21% computational complexity while maintaining almost the same rate distortion (RD) performance as the HEVC test model reference software, HM 13.0.

STAR-24K: A Public Dataset for Space Common Target Detection

Chaoyan Zhang,Baolong Guo,Nannan Liao,Qiuyun Zhong,Hengyan Liu,Cheng Li,Jianglei Gong 한국인터넷정보학회 2022 KSII Transactions on Internet and Information Syst Vol.16 No.2

The target detection algorithm based on supervised learning is the current mainstream algorithm for target detection. A high-quality dataset is the prerequisite for the target detection algorithm to obtain good detection performance. The larger the number and quality of the dataset, the stronger the generalization ability of the model, that is, the dataset determines the upper limit of the model learning. The convolutional neural network optimizes the network parameters in a strong supervision method. The error is calculated by comparing the predicted frame with the manually labeled real frame, and then the error is passed into the network for continuous optimization. Strongly supervised learning mainly relies on a large number of images as models for continuous learning, so the number and quality of images directly affect the results of learning. This paper proposes a dataset STAR-24K (meaning a dataset for Space TArget Recognition with more than 24,000 images) for detecting common targets in space. Since there is currently no publicly available dataset for space target detection, we extracted some pictures from a series of channels such as pictures and videos released by the official websites of NASA (National Aeronautics and Space Administration) and ESA (The European Space Agency) and expanded them to 24,451 pictures. We evaluate popular object detection algorithms to build a benchmark. Our STAR-24K dataset is publicly available at https://github.com/Zzz-zcy/STAR-24K.