In this paper, we propose a algorithm to tracking target using point feature. The point feature is extracted from the pixels in the first frame and used to label the pixels in the next frame as belonging to either target or background. The positive and negative samples are extracted from the pixels of target and surrounding background, and used to train several weak classifiers, which combine to build a strong classifier using AdaBoost algorithm. The negative samples are given the greater weights than positive samples, which is to avoid that a large number of pixels in background are labeled incorrectly. To efficiently learn a large number of samples, the adopted weak classifier is a linear perceptron model, which is trained and updated using stochastic gradient descent. Only the dot-product between matrices and the sum of matrix elements need to be calculated. To distinguish the similar targets, the histogram-based mean shift algorithm is applied to eliminate those wrong image patches. The histogram of target will be updated over the time. The experiment results show that the proposed algorithm can estimate scale better when scale change, posture change and occlusion occurs.
The traditional algorithms of image super-resolution reconstruction are not effective enough to be used in reconstructing high-frequency information of an image. In order to improve the quality of image reconstruction and restore more high-frequency information, the residual dictionary is introduced which can capture the high-frequency information of images such as the edges, angles and corners. The common dictionary is generated by training and learning pairs of low-resolution and high-resolution images. The dictionary combined by common dictionary and residual dictionary is obtained in which more high-frequency information of the images can be restored while the spatial structure of images can be preserved well. The processing of training and testing dictionary is conducted by Support Vector Regression (SVR). Compared with other algorithms in experiments, the proposed method improves its PSNR and SSIM by 3% ~ 4% and 2% ~ 3% on some different images respectively.
Due to the limitations of image capture device and imaging environments in traditional imaging process, high-resolution (HR) images are difficult to be obtained. The method of digital image processing can be used in image super-resolution with one or an image sequence in original conditions to reconstruct HR images which over the range of imaging system. Traditional learning-based super-resolution algorithm need to run through the sample library with a high computing complexity, and a high recognition rate in the scene with small shifts. This dissertation is mainly about color image SR and parallel implementation of the SR algorithm. An algorithm based on SVM classified learning is proposed in this paper.
In recent years, we have witnessed the prosperity of the face image super-resolution (SR) reconstruction, especially the learning-based technology. In this paper, a novel super-resolution face reconstruction framework based on support vector regression (SVR) about a single image is presented. Given some input data, SVR can precisely predict output class labels. We regard the SR problem as the estimation of pixel labels in its high resolution version. It’s effective to put local binary pattern (LBP) codes and partial pixels into input vectors during training models in our work, and models are learnt from a set of high and low resolution face image. By optimizing vector pairs which are used for learning model, the final reconstructed results were advanced. Especially to deserve to be mentioned, we can get more high frequency information by exploiting the cyclical scan actions in the process of both training and prediction. A large number of experimental data and visual observation have shown that our method outperforms bicubic interpolation and some stateof- the-art super-resolution algorithms.
Wet paper code is a complex model which mainly used in the field of image coding. This paper is based on the wet paper
code model and human visual system, and constructs a new wet paper code steganographic method. According to the
regional complexity and other characteristics of the host image, the secret bits are adaptively embedded into wavelet
coefficients of image subbands with wet paper code. Secret information receivers do not need to know the specific
method of secret writing, just do a simple matrix multiplication operation and can extract the secret information, which
in many ways to improve the security of the steganographic algorithm. The experiments show that the method has good
visual invisibility and resistance of active steganalysis attacks.
In this paper, a new data hiding method - writing on wet paper using multi-pixel differencing is presented in order
to provide large embedding capacity and improve further steganographic embedding efficiency for the stego-image. It
takes into consideration four pixels of a block, and the differences between the lowest gray-value pixel and its
surrounding pixels are used to embed the secret data. The receivers can extract secret bits from carrier images only by
some matrix multiplications no matter how we embed the secret message. Thus, the invisibility of this steganographic
method is achieved in process of secret information transmission. The experimental results show that our scheme has a
larger embedding capacity and better robust performance against the methods of active attacks such as noise addition.
This letter bases on the researches of LSB (least significant bit, i.e. the last bit of a binary pixel value) matching
steganographic method and the steganalytic method which aims at histograms of cover images, and proposes a
modification to LSB matching. In the LSB matching, if the LSB of the next cover pixel matches the next bit of secret
data, do nothing; otherwise, choose to add or subtract one from the cover pixel value at random. In our improved
method, a steganographic information table is defined and records the changes which embedded secrete bits introduce in.
Through the table, the next LSB which has the same pixel value will be judged to add or subtract one dynamically in
order to ensure the histogram's change of cover image is minimized. Therefore, the modified method allows embedding
the same payload as the LSB matching but with improved steganographic security and less vulnerability to attacks
compared with LSB matching. The experimental results of the new method show that the histograms maintain their
attributes, such as peak values and alternative trends, in an acceptable degree and have better performance than LSB
matching in the respects of histogram distortion and resistance against existing steganalysis.