Applied Mechanics and Materials Vols. 511-512

Abstract: A texture feature extraction method based on Gabor wavelet transform is presented to avoid the limitations of the shape structure characteristics extraction in the palm vein image. First acquire the palm vein images by independent structures of vein acquire device then extracts ROI area of vein images by determining the centerline of the middle finger and ring finger interphalangeal. Correcting the vein image by Gamma operator will improve the contrast of the processing image. Finally, transform the image based on Gabor wavelet to extract texture feature.

Abstract: This paper proposed a new facial expression recognition algorithm based on gabor texture features and Adaboost feature selection via SRC(sparse representation classification). Five scales and eight orientations of Gabor wavelet filters were used in this paper to extract gabor features. For an image of size , the number of gabor features is 163840, In order to extract the most effective features for FER(facial expression recognition), Adaboost algorithm is used for feature selection. This paper divided 7 facial expressions into two categories, where the neutral expression as the first class and the remaining six expressions as the second class. In each size and orientation 110 features are selected. At last 4400 features are selected combined SRC algorithm for FER. Test experiments were performed on Japanese female JAFFE facial expression database. Compared with the traditional expression recognition algorithms such as 2DPCA+SVM, LDA+SVM, the new algorithm achieved a better recognition rate, which shows the effectiveness of the proposed new algorithm.

Abstract: A new method based on Monogenic Binary Coding (MBC) is proposed for facial expression feature extraction and representation. Firstly, monogenic signal analysis is used to extract multi-scale magnitude, orientation and phase components. Secondly, Monogenic Binary Coding (MBC) is used to encode the monogenic local variation and intensity in local regions of each extracted component in each scale and local histograms are built. Then Blocked Fisher Linear Discrimination (BFLD) is used to reduce the dimensionality of histogram features and to enhance discrimination. Finally the three complementary components are fused for more effective facial expression recognition (FER). Experiment results on Japanese female expression database (JAFFE) show that the performance of the fusion method is even better than state-of-the-art local feature based FER methods such as Local Binary Pattern (LBP)+Sparse Representation (SRC), Local Phase Quantization (LPQ)+SRC ,etc.

Abstract: Image coding and compression is one of the most key techniques in the area of image signal processing, However, most of the existing coding methods such as JPEG, employ the similar hybrid architecture to compress images and videos. After many years of development, it is difficult to further improve the coding performance. In addition, most of the existing image compression algorithms are designed to minimize difference between the original and decompressed images based on pixel wise distortion metrics, such as MSE, PSNR which do not consider the HVS features and is not able to guarantee good perceptual quality of reconstructed images, especially at low bit-rate scenarios. In this paper, we propose a novel scheme for low bit-rate image compression. Firstly, the original image is quantized to a binary image based on heat transfer theory. Secondly, the bit sequence of the binary image is divided into several sub-sets and each one is designated a priority based on the rate-distortion principle. Thirdly, the sub-sets with high priorities are selected based on the given bit-rate. Finally, the context-based binary arithmetic coding is employed to encode the sub-sets selected to produce the final compressed stream. At decoder, the image is decoded and reconstructed based on anisotropic diffusion. Experiments are conducted and provide convincing results.

Abstract: As we all know, there is conspicuous connectivity shared by neighboring image block attributes between consecutive frames. In this paper, we used particle filter to estimate the motion curve for every macro blocks during a certain time period, and then developed a video reconstruction computing method by doing the mean and interpolation processes. Compared with MVs, the video constructed by our method can remain more structural features existed in the original image sequences, and has great potential in increasing coding efficiency.

Abstract: An effective multi-feature extraction method for well-log curve line plotted manually. Several log curve line features are used to extract the well log curve information in the proposed method: gray value similarity feature, connected region Length feature, projection feature and distance and angular feature. The experimental results show the good effect of the method for the well log curve information extraction.

Abstract: The underwater images are disturbed with low signal to noise ratio and edge blur, because there are the light scattering and absorption effects. If the traditional thresholding method is used directly to segment underwater images, it will usually lead to be less effective to process underwater images. An image segmentation method of underwater target based on active contour model was proposed in this paper. Firstly, using Canny edge detection algorithm to detect the edges of the original image to obtain the information of a crude outline, then the algorithm based on C-V active contour model to segment underwater target images was addressed. The images processing results based on threshold segmentation method and C-V model method were compared. Experiments demonstrate the effectiveness of the proposed algorithm for underwater targets images segmentation.

Abstract: A natural color fusion method for infrared and low-light-level image is proposed. This method utilizes image fusion and color transfer. The fused image uses sparse representation to merge the source images information to be assigned to the Y channel. And then the I and Q channel is combined using Toets method, which extracts the common component from the source images. Finally, the false-color image is obtained by using color transfer technology to the prior pseudo-color YIQ image. Experiments show that the result of our method is information that is more salient, has a higher color contrast, and a more natural color appearance when compared with those produced by the traditional coloration algorithm.

Abstract: Feature selection plays an important role in terrain classification for outdoor robot navigation. For terrain classification, the image data usually have a large number of feature dimensions. The better selection of features usually results in higher labeling accuracy. In this work, a novel approach for terrain perception using Importance Factor based I-Relief algorithm and Feature Weighted Support Vector Machines (IFIR-FWSVM) is put forward. Firstly, the weight of each feature for classification is computed by using Importance Factor based I-Relief algorithm (IFIR) and the irrelevant features are eliminated. Then the weighted features are used to compute the kernel functions of SVM and trained the classifier. Finally, the trained SVM is employed to predict the terrain label in the far-field regions. Experimental results based on DARPA datasets show that the proposed method IFIR-FWSVM is superior over traditional SVM.

Abstract: The Retinex model is mainly used to removal of unfavorable illumination effects from images. In this paper, the Retinex model combined with the total variation regularization (TV-Retinex) is presented to removal of glass reflection that can be solved by a fast computational approach based on the split Bregman iteration. Experiments demonstrated that the proposed method can effectively reduce this kind of artifact as well as preserve the edge and detailed information.