Xinggang Lin

Fast intra prediction algorithm in H.264-AVC

Spatial intra prediction is introduced to achieve higher coding efficiency in H.264/AVC. Unfortunately, this comes at the cost in considerably increased complexity when RDO is enabled. In this paper, we propose a fast intra prediction algorithm including reducing the redundancy between the chroma mode decision and luma mode decision, improving the mode decision scheme and heuristically selecting block type. Experimental results show that the proposed algorithm can drastically reduce coding complexity with only negligible coding efficiency loss.

Three-level GPU accelerated Gaussian mixture model for background subtraction

Gaussian Mixture Model (GMM) for background subtraction (BGS) is widely used for detecting and tracking objects in video sequences. Although the GMM can provide good results, low processing speed has become its bottleneck for realtime applications. We propose a novel method to accelerate the GMM algorithm based on graphics processing unit (GPU). As GPU excels at performing massively parallel operations, the novelty lies in how to adopt various optimization strategies to fully exploit GPUs resources. The parallel design consists of three levels. On the basis of first-level implementation, we employ techniques such as memory access coalescing and memory address saving to the secondlevel optimization and the third-level modification, which reduces the time cost and increases the bandwidth greatly. Experimental results demonstrate that the proposed method can yield performance gains of 145 frames per second (fps) for VGA (640*480) video and 505 fps for QVGA (320*240) video which outperform their CPU counterparts by 24X and 23X speedup respectively. The resulted surveillance system can process five VGA videos simultaneously with strong robustness and high efficiency.

Threshold selection using cross-entropy and fuzzy divergence

Image segmentation consists of dividing an image into non- intersecting and dissimilar but meaningful regions (object and background). Thresholding is a commonly employed technique for segmenting image. Many methods for automatic selection of thresholds use optimization process in which some specific criterion functions are defined. Recently, several thresholding methods based on minimizing the cross- entropy function of images have been proposed. Cross-entropy measures the information discrepancy between two probability distributions. Derived from cross-entropy, fuzzy divergence measures the dissimilarity between two fuzzy sets. In this paper, we present four new algorithms for optimal threshold selection based on different criteria integrating cross entropy and fuzzy divergence. The first one is a minimum cross entropy algorithm based on the hypothesis of uniform probability distribution. The second one is a maximum between-class cross entropy algorithm using a posterior probability. The third one is a modified version of existing method based on maximum between-class fuzzy divergence. The last one is a minimum fuzzy divergence algorithm. According to the requirement of image thresholding, we construct a new fuzzy membership function to take into account the gray level probability distribution of object pixels and background pixels about their mean values for the last two algorithms. The effectiveness and generality of these proposed algorithms have been compared with some recent techniques based on related principles, and evaluated by using uniformity measure and shape measure with real images. Results showing the superiority of the proposed algorithms are presented.

Face recognition using a modified line segment hausdorff distance

The automatic recognition of human faces presents a significant challenge to the pattern recognition research community. Typically, human faces are very similar in structure with minor differences from person to person. Furthermore, lighting condition changes, facial expressions, and pose variations further complicate the face recognition task as one of the difficult problems in pattern analysis. In this paper, we proposed a modified line segment hausdorff distance measure for face recognition. The distance measure is weighted according to a weight function derived from the spatial information of the human face. This distance measure, namely Spatially Weighted Line Hausdorff Distance (SWLHD), is further improved by limit the line pairs in a given neighborhood. Experiment showed encouraging result and the results also implicate that line segments could provide sufficient information for face recognition.

Identification of pelagic eggs by image analysis

A pelagic egg identification system by image analysis is proposed. The egg image is obtained using a camera coupled with a microscope. In order to extract the meaningful structures of egg, the edge map of the egg image is used. Our edge detection method consists of two steps: 1. Mark on four edge maps the potential edge positions corresponding to the local maxima of gradients in four directions; 2. Identify the true edge positions in accordance with the local facet orientation and aggregate them into an edge map with some spatial pattern of regularity. Since most of these structures (except the embryo) have circular shapes, a circle detection method using the Hough transform is proposed, where the problem of circle over-detection in a possible site is resolved. An embryo segmentation method is also developed. Based on some criteria of identification defined on the parameters extracted from the structures detected, the pelagic eggs are then classified into the corresponding categories and the corresponding stages of development. Experimental results are also presented in the paper.

Stereo matching using both orthogonal and multiple image pairs

In the stereo matching process, a difficult task is to eliminate the false match caused by the repetitive patterns and smooth patterns. A new stereo matching technique that uses multiple image pairs in orthogonal-directions has been proposed. First, it computes the sum of squared-difference (SSD) values for each stereo pair in orthogonal-directions. Then it simply adds individual SSD to produce the sum of SSDs. The resulting function is called the OSSSD. We have proved that the OSSSD function possesses a unique and clear minimum at the true matching position, so it can easily remove the false match caused by both repetitive and smooth patterns. A number of experiments with synthetic and real image sequences are performed, and the results are presented to demonstrate the effectiveness of the technique.

L-∞ constrained micronoise filtering for high-fidelity image compression

Imaging apparatus inevitably impose undesirable noises onto acquired images during real imaging process. Usually these noises are too faint to cause unpleasing visual effects, however, they degrade image fidelity and significantly lower the compression ratio of lossless coding. More baffling, in this case, there leaves little room for traditional noise filtering methods to work. This paper will introduce some of our efforts trying to weaken the effect of such Micro Noise during near-lossless compression. Experimental results on ISO test images and micro Gaussian noises demonstrate that with potentiality of filtering micro noise, an improved near-lossless coder can not only achieve obviously higher compression ratio but also provide better image fidelity (measured by mean squared error) than lossless coding.