Due to the instability of network training, the image super-resolution reconstruction based on Generative Adversarial Network (GAN) has a mode collapse phenomenon. To solve this problem, a Spherical double Discriminator Super-Resolution Generative Adversarial Network (SDSRGAN) based on spherical geometric moment matching and feature discrimination was proposed, and the stability of network training was improved by adopting geometric moment matching and discrimination of high-frequency features. First of all, the generator was used to produce a reconstructed image through feature extraction and upsampling. Second, the spherical discriminator was used to map image features to high-dimensional spherical space, so as to make full use of higher-order statistics of feature data. Third, a feature discriminator was added to the traditional discriminator to extract high-frequency features of the image, so as to reconstruct both the characteristic high-frequency component and the structural component. Finally, game training between the generator and double discriminator was carried out to improve the quality of the image reconstructed by the generator. Experimental results show that the proposed algorithm can effectively converge, its network can be stably trained, and has Peak Signal-to-Noise Ratio (PSNR) of 31.28 dB, Structural SIMilarity (SSIM) of 0.872. Compared with Bicubic, Super-Resolution Residual Network (SRResNet), Fast Super-Resolution Convolutional Neural Network (FSRCNN), Super-Resolution using a Generative Adversarial Network (SRGAN), and Enhanced Super-Resolution Generative Adversarial Network (ESRGAN) algorithms, the reconstructed image of the proposed algorithm has more precise structural texture characteristics. The proposed algorithm provides a double discriminant method for spherical moment matching and feature discrimination for the research of image super-resolution based on GAN, which is feasible and effective in practical applications.

Aiming at the problems of low accuracy and even object lost in moving object tracking under occlusion, deformation, rotation, and illumination changes and poor real-time performance of the traditional tracking algorithm, a target tracking algorithm based on foreground discrimination and Circle Search (CS) was proposed. The image perceptual hashing technique was used to describe and match tracked object, and the tracking process was based on the combination of the above was tracking strategies, which could effectively solve the above problems. Firstly, because the direction of motion uncertain and the inter-frame motion was slow, CS algorithm was used to search the local best matching position (around the tracked object) in the current frame. Then, the foreground discrimination PBAS (Pixel-Based Adaptive Segmenter) algorithm was adopted to search for the global optimal object foreground in the current frame. Finally, the one with higher similarity with the object template was selected as the tracking result, and whether to update the target template was determined according to the matching threshold. The experimental results show that the proposed algorithm is better than the MeanShift algorithm in precision, accuracy, and has a better tracking advantage when the target is not moving fast.

A joint time-frequency audio feature extraction algorithm based on Gabor dictionary and weight tensor of sparse representation was proposed to describe the characteristic of non-stationary audio signal. Conventional sparse representation uses a predefined dictionary to encode the audio signal as sparse weight vector. In this paper, the elements in the weight vector were reorganized into tensor format. Each order of the tensor respectively characterized time, frequency and duration property of signal, making it the joint time-frequency-duration representation of the signal. The frequency factors and duration factors were concatenated as audio features through tensor decomposition. To solve the over-fitting problem of sparse tensor factorization, an automatic-adjust-penalty-coefficient factorization algorithm was proposed. The experimental results show that the proposed feature outperforms MFCC (Mel-Frequency Cepstrum Coefficient) feature, MFCC+MP feature concatenated by MFCC and Matching Pursuit (MP) features, and nonuniform scale-frequency map feature by 28.0%, 19.8% and 6.7% respectively, in 15-category audio classification.

Most existing detection algorithms of video frame copy-move forgery in time domain were designed for the copy-move forgery of video sequence containing 20 frames at least, and are difficult to detect single frame forgery. While according to the characteristics of human visual perception, 15 frames at least were needed to modify the video meaning. So when goal in vision was made by the tampering, continuous operation and many times were needed. In order to detect the tampering, a detection algorithm based on quantized Discrete Cosine Transform (DCT) coefficients for continuous and repeated single frame copy-move forgery in videos was proposed. Firstly, the video was converted into images, and quantized DCT coefficients were taken as the feature vector of a frame image. Then, the similarity between frames was measured by calculating Bhattacharyya coefficient, and threshold was set to judge the abnormal similarity between two adjacent frames. Finally, whether the video was tampered and the tampered positions were determined by the continuity of frames with abnormal similarity and the number of continuous frames. The experimental results show that the proposed algorithm can detect the video with different scenarios, it possesses fast detection speed, and is not affected by further compression factors, but also is of high accuracy and low omission ratio.