Accurate segmentation of the pancreas and its lesions in CT scans is crucial for the precise diagnosis and treatment of pancreatic cancer. However, it remains a highly challenging task due to several factors such as low tissue contrast with surrounding organs, blurry anatomical boundaries, irregular organ shapes, and the small size of lesions. To tackle these issues, we propose DB-MSMUNet (Dual-Branch Multi-scale Mamba UNet), a novel encoder-decoder architecture designed specifically for robust pancreatic segmentation. The encoder is constructed using a Multi-scale Mamba Module (MSMM), which combines deformable convolutions and multi-scale state space modeling to enhance both global context modeling and local deformation adaptation. The network employs a dual-decoder design: the edge decoder introduces an Edge Enhancement Path (EEP) to explicitly capture boundary cues and refine fuzzy contours, while the area decoder incorporates a Multi-layer Decoder (MLD) to preserve fine-grained details and accurately reconstruct small lesions by leveraging multi-scale deep semantic features. Furthermore, Auxiliary Deep Supervision (ADS) heads are added at multiple scales to both decoders, providing more accurate gradient feedback and further enhancing the discriminative capability of multi-scale features. We conduct extensive experiments on three datasets: the NIH Pancreas dataset, the MSD dataset, and a clinical pancreatic tumor dataset provided by collaborating hospitals. DB-MSMUNet achieves Dice Similarity Coefficients of$89.47 \%, 87.59 \%$, and 89.02 %, respectively, outperforming most existing state-of-the-art methods in terms of segmentation accuracy, edge preservation, and robustness across different datasets. These results demonstrate the effectiveness and generalizability of the proposed method for real-world pancreatic CT segmentation tasks.

Example-based texture synthesis is a powerful tool to produce high-quality textures of arbitrary sizes essential in many computer graphic applications. The quality of the input texture is of vital importance to the success of example-based texture synthesis. In this paper, we propose a new texture compaction framework that adaptively synthesizes small texture samples by abstracting the patterns of given large textures according to an importance map. This map is computed from the original texture based on visual saliency to emphasize the important parts of the image that are to be reserved better during compaction. With such enhanced texture compaction, we can re-synthesize textures that maintain the feature patterns of the original textures better compared with previous inverse texture synthesis.

The framework of Texture-by-numbers (TBN) synthesizes images of global-varying patterns with intuitive user control. Previous TBN synthesis methods have difficulties in achieving high-quality synthesis results and efficiency simultaneously. This paper proposes a fast TBN synthesis method based on texture optimization, which uses global optimization to solve the controllable non-homogeneous texture synthesis problem. Our algorithm produces high quality synthesis results by combining texture optimization into TBN framework with two improvements. The initialization process is adopted to generate the initial output of the global optimization algorithm, which speeds up the algorithm's convergence rate and ensures synthesis quality. Besides different metrics to measure image similarity are defined to match human visual perception better. To further improve the synthesis speed, the algorithm is entirely implemented on GPU based on CUDA architecture. The experimental results show that this method synthesizes realistic images with high efficiency, which is not only applicable to the traditional TBN application, but also suitable for other applications including non- photorealistic rendering and image in-painting.

In forest dynamics models, the intensive computation and load involved in the simulation of seed dispersal can become unbearably huge for large-scale forest analysis. To solve this problem, we propose a multi-resolution algorithm to compute seed dispersal on GPU. By exploiting the computation parallelism of seed dispersal, the computation of the whole forest plot is divided into multiple small plot cells, which are computed independently by parallel threads on GPU. To further improve the calculation efficiency with limited threads scale for GPU computation, we propose a hierarchical method to cluster the plot cells into a multi-resolution form according to the biological curves of tree seed dispersal. Experimental results show that our algorithm not only greatly reduces computational time but also obtains comparably correct results as compared to the naive GPU algorithm, which makes it especially suitable for large-scale forest modeling.

Due to the problem that the forest growth simulation system has relatively fixed system architecture and difficulty in reconstruction, this paper presents a rapid reconfiguration platform of dynamic forest growth simulation system based on service component. The key technologies of the rapid reconfiguration platform such as the flowchart of reconfiguration, the design of service component and knowledge representation methods are described. This platform adopts the basic idea of service-oriented and component-oriented technology, and it is composed of supporting environment layer, simulation resource layer, simulation runtime supporting service layer, simulation integration layer and application layer. In this rapid reconfiguration platform, different simulation systems can be reconfigured rapidly according to the simulation tasks and some simulation systems have been reconfigured successfully.

In this paper, we present an importance-driven texture encoding algorithm based on samples. Our algorithm determines a set of samples from source texture based on combined criteria which include compression ratio, visual attention and parameterization distortion. The sample set is used to encode the majority parts of the texture. The remaining regions are then encoded by traditional compression algorithm such as vector quantization. Our method can preserve details of important areas and be extended to dynamic textures. The decoding procedure is performed entirely in programmable graphics hardware, yielding real-time frame rates. Experimental results demonstrate the efficiency and performance of our algorithm.

Presents a new efficient texture mapping algorithm for polygonal meshes. Texture coordinates are assigned to the vertices of the mesh using the variational interpolation technique. For each patch to be mapped, the algorithm first samples equidistantly the boundaries of the patch. In order to decrease distortion, the patch is further sampled as uniformly as possible while its texture domain are correspondingly sampled. The variational interpolation technique is then invoked to yield a parameterization of the patch. The experimental results demonstrate that the algorithm has a great number of potential applications in texture mapping and surface modeling.