Image-to-image translation is an important and challenging problem in computer vision image processing. Diffusion models (DM) have shown great potentials for high-quality synthesis, gained competitive performance on the task of image-to-image translation. However, most existing diffusion treat as conditional generation processes, suffer heavily from gap between distinct domains. In this paper, a novel method based Brownian Bridge Model (BBDM) proposed, which stochastic process, learns two...

In this paper, we propose a novel correlation particle filter (CPF) for robust visual tracking. Instead of simple combination and filter, exploit complement the strength each one. Compared with existing tracking methods based on filters filters, proposed tracker has four major advantages: 1) it is to partial total occlusions, can recover from lost tracks by maintaining multiple hypotheses; 2) effectively handle large-scale variation via sampling strategy; 3) efficiently maintain modes in...

The video highlight detection task is to localize key elements (moments of user's major or special interest) in a video. Most existing approaches extract features from the segment as whole without considering difference local both temporally and spatially. Due complexity content, this kind mixed will impact final prediction. In temporal extent, not all frames are worth watching because some them only contain background environment human other moving objects. spatial it similar that regions...

Replacement of least-significant bit plane is one the popular steganography techniques in digital images because its extreme simplicity. But it more difficult to precisely estimate rate secret message embedded by replacement multiple (MLSB) planes a carrier object. In order model MLSB embedding, lemma introduced prove transition relationships among some trace subsets. Then, based on these relationships, two novel steganalysis frameworks are designed detect kinds distinct embedding methods. A...

Contrast enhancement regarded as a significant branch of image processing, not only enhances human perception details hidden in the scene, but also improves rapid recognition interested targets. In this paper, non-overlapped sub-blocks and local histogram projection based contrast (NOSHP) is presented. First, original segmented into numbers where (HP) then executed individually. Subsequently, each sub-block related to its adjacent three ones by certain weights, so that integral can be both...

In fringe projection profilometry, the relationship between phase and height is of utmost significance to perform accurate measurement. To enhance accuracy robustness, a calibration method presented based on pattern geometric constraints. By analyzing distribution in image coordinate, this establishes one-to-one mapping directly, lens distortion camera considered involved mathematical description, so as eliminate error propagation. The (x, y) coordinates are calculated from with known...

Researches on breast cancer histopathological image classification have achieved a great breakthrough using deep backbones of Convolutional Neural Networks (CNNs) in recent years. However, due to the inductive bias locality, CNNs are unable effectively extract global feature information images, limiting improvement results. To overcome this shortcoming, paper reasonably introduces an extra backbone stream pure transformer, which consists self-attention mechanism capture receptive fields...

We present an optical encryption scheme for hyperspectral images using improved binary tree structure (IBTS) and phase-truncated discrete multiple-parameter fractional Fourier transform (PTdmpFrFT). In the IBTS, modules based on represented branch nodes, while bands of plain were considered as leaf nodes. All pairs encoded into intermediate data with subsequent asymmetric ciphertexts PTdmpFrFT. The proposed approach generated different secret keys, decryption paths. Compared state-of-the-art...

Text-driven image editing aims to manipulate images with the guidance of natural language description. Text is much more and intuitive than many other interaction modes, attracts attention recently. However, compared classical supervised learning tasks, there no standard benchmark dataset for text-driven interactive up now. Therefore, it hard train an end-to-end model pixel-aligned driven by text. Some methods follow paradigm text-to-image models incorporating target into process generation....

Most of the existing 3D talking face synthesis methods suffer from lack detailed facial expressions and realistic head poses, resulting in unsatisfactory experiences for users. In this paper, we propose a novel pose-aware method with geometry-guided audio-vertices attention. To capture more expression, such as subtle nuances mouth shape eye movement, to build hierarchical audio features including global attribute feature series vertex-wise local latent movement features. Then, order fully...

An infrared image detail enhancement method based on local adaptive gamma correction (LAGC) is proposed.The values are designed the Weber curve to enhance effectively details.Subsequently, active grayscale range of processed by LAGC further extended using our proposed histogram statistical stretching.The experimental results show that algorithm could considerably increase details and improve contrast entire image.Thus, it has significant potential for practical applications.

Clothes-invariant feature extraction is critical to the clothes-changing person re-identification (CC-ReID). It can provide discriminative identity features and eliminate negative effects caused by confounder--clothing changes. But we argue that there exists a strong spurious correlation between clothes human identity, restricts common likelihood-based ReID method P(Y|X) extract clothes-irrelevant features. In this paper, propose new Causal Clothes-Invariant Learning (CCIL) achieve...

Automatic anime sketch colorization aims to generate a color image from image, which is challenging due limited structure and semantic understanding, leading constrained style, inconsistency. In this paper, we introduce diffusion model with prompt learning (SPL), better prompts stimulate the powerful understanding capabilities of large-scale multi-modal models, effectively bridging gap between color. We two distillation strategies for prompts: one prediction-level by optimizing global...

A novel blind compressed video stream steganalysis algorithm employing the inter-frame correlation (IFCS) is proposed. The process executed wholly in domain for real time requirement. Statistical features of correlation, AC coefficients and DC DCT are extracted out as classifier using an collusion principle. window adjusted according to correlation. And one three layers feedforward neural network capable nonlinear feature mapping employed classifier. Experimental results verify availability...