Inspired by the unique structure of insect compound eyes,a multi-channel image acquisition system is designed to photograph a cylindrical panorama of its surroundings with one shot. The hardware structure consists of ...Inspired by the unique structure of insect compound eyes,a multi-channel image acquisition system is designed to photograph a cylindrical panorama of its surroundings with one shot. The hardware structure consists of an embedded ARM system and one array of 16 micro-image sensors. The system achieves the synchronization of captured photos in 10 ms,as well as 10 f /s video capture. The software architecture includes the TCP /IP protocol,video capture procedures in"Poll/Read"or"video streaming"modes,thread pool monitoring in multi-threading mutex,synchronization control with the"event""mutex signal"and"critical region"functions,and a synthetic image algorithm characterized by its portability,modularity,and remote transmission. The panoramic imaging system is expected to be a vision sensor for mobile robotics.展开更多
Panoramic images, offering a 360-degree view, are essential in virtual reality(VR) and augmented reality(AR), enhancing realism with high-quality textures. However, acquiring complete and high-quality panoramic textur...Panoramic images, offering a 360-degree view, are essential in virtual reality(VR) and augmented reality(AR), enhancing realism with high-quality textures. However, acquiring complete and high-quality panoramic textures is challenging. This paper introduces a method using generative adversarial networks(GANs) and the contrastive language-image pretraining(CLIP) model to restore and control texture in panoramic images. The GAN model captures complex structures and maintains consistency, while CLIP enables fine-grained texture control via semantic text-image associations. GAN inversion optimizes latent codes for precise texture details. The resulting low dynamic range(LDR) images are converted to high dynamic range(HDR) using the Blender engine for seamless texture blending. Experimental results demonstrate the effectiveness and flexibility of this method in panoramic texture restoration and generation.展开更多
Abstract： Based on digital signal processor（DSP） and field programmable gate array（FPGA） techniques, the architecture of super large view field（SLVF） panoramic night vision image processing hardware platform wa...Abstract： Based on digital signal processor（DSP） and field programmable gate array（FPGA） techniques, the architecture of super large view field（SLVF） panoramic night vision image processing hardware platform was established. The panoramic unwrapping and correcting algorithm, up to a full 360°, based on coordinate rotation digital computer （CORDIC） and night vision image enhancement algorithm, based on histogram equalization theory and edge detection theory, was presented in this paper, with the purpose of processing night vision dynamic panoramic annular image. The annular image can be unwrapped and corrected to conventional rectangular panorama by the panoramic image processing algorithm, which uses the pipelined CORDIC configuration to realize a trigonometric function generator with high speed and high precision. Histogram equalization algorithm can perfectly enhance the contrast of the night vision image. Edge detection algorithm can be propitious to find and detect small dim dynamic targets in night vision circumstances. After abundant experiment, the al- gorithm for panoramic image processing and night vision image enhancement is successfully implemented in FPGA and DSP. The panoramic night vision image system is a compact device, with no external rotating parts. And the system can reliably and dynamically detect 360＊ SLVF panoramic night vision image.展开更多
An effective algorithm of electronic image stabilization （EIS） of catadioptric panoramic imaging system for track robots is presented. The key techniques of this algorithm are as follows：① A model of electronic im...An effective algorithm of electronic image stabilization （EIS） of catadioptric panoramic imaging system for track robots is presented. The key techniques of this algorithm are as follows：① A model of electronic image stabilization is built by analyzing the imaging theory and the principle of EIS, and the image shift function of unwrapped panoramic image is deduced;② The relationship equation between motion estimation parameters of annular panoramic image and motion estimation parameters of unwrapped panoramic image is developed according to the constrained aspect ratio of real objects, motion parameters of annular panoramic image are firstly estimated, and then motion parameters among the image shift function are carried out according to the relationship equation;③ An excessive stabilization threshold is presented to prevent the phenomena of excessive stabilization, and the Kalman filtering is adopted to smooth the image sequences. Numerical experimental results show that this algorithm can effectively smooth out the unwanted motion and follow the intentional camera movement under certain resolutions.展开更多
基金Supported by the National Natural Science Foundation of China(61233014)the China Postdoctoral Science Foundation(2012M5210711,20123218110031)the National Natural Science Major International Cooperation Projects(61161120323)
文摘Inspired by the unique structure of insect compound eyes,a multi-channel image acquisition system is designed to photograph a cylindrical panorama of its surroundings with one shot. The hardware structure consists of an embedded ARM system and one array of 16 micro-image sensors. The system achieves the synchronization of captured photos in 10 ms,as well as 10 f /s video capture. The software architecture includes the TCP /IP protocol,video capture procedures in"Poll/Read"or"video streaming"modes,thread pool monitoring in multi-threading mutex,synchronization control with the"event""mutex signal"and"critical region"functions,and a synthetic image algorithm characterized by its portability,modularity,and remote transmission. The panoramic imaging system is expected to be a vision sensor for mobile robotics.
文摘Panoramic images, offering a 360-degree view, are essential in virtual reality(VR) and augmented reality(AR), enhancing realism with high-quality textures. However, acquiring complete and high-quality panoramic textures is challenging. This paper introduces a method using generative adversarial networks(GANs) and the contrastive language-image pretraining(CLIP) model to restore and control texture in panoramic images. The GAN model captures complex structures and maintains consistency, while CLIP enables fine-grained texture control via semantic text-image associations. GAN inversion optimizes latent codes for precise texture details. The resulting low dynamic range(LDR) images are converted to high dynamic range(HDR) using the Blender engine for seamless texture blending. Experimental results demonstrate the effectiveness and flexibility of this method in panoramic texture restoration and generation.
文摘Abstract： Based on digital signal processor（DSP） and field programmable gate array（FPGA） techniques, the architecture of super large view field（SLVF） panoramic night vision image processing hardware platform was established. The panoramic unwrapping and correcting algorithm, up to a full 360°, based on coordinate rotation digital computer （CORDIC） and night vision image enhancement algorithm, based on histogram equalization theory and edge detection theory, was presented in this paper, with the purpose of processing night vision dynamic panoramic annular image. The annular image can be unwrapped and corrected to conventional rectangular panorama by the panoramic image processing algorithm, which uses the pipelined CORDIC configuration to realize a trigonometric function generator with high speed and high precision. Histogram equalization algorithm can perfectly enhance the contrast of the night vision image. Edge detection algorithm can be propitious to find and detect small dim dynamic targets in night vision circumstances. After abundant experiment, the al- gorithm for panoramic image processing and night vision image enhancement is successfully implemented in FPGA and DSP. The panoramic night vision image system is a compact device, with no external rotating parts. And the system can reliably and dynamically detect 360＊ SLVF panoramic night vision image.
基金Supported by State Key Laboratory of Explosion Science and Technology Foundation(ZDKT08-05)
文摘An effective algorithm of electronic image stabilization （EIS） of catadioptric panoramic imaging system for track robots is presented. The key techniques of this algorithm are as follows：① A model of electronic image stabilization is built by analyzing the imaging theory and the principle of EIS, and the image shift function of unwrapped panoramic image is deduced;② The relationship equation between motion estimation parameters of annular panoramic image and motion estimation parameters of unwrapped panoramic image is developed according to the constrained aspect ratio of real objects, motion parameters of annular panoramic image are firstly estimated, and then motion parameters among the image shift function are carried out according to the relationship equation;③ An excessive stabilization threshold is presented to prevent the phenomena of excessive stabilization, and the Kalman filtering is adopted to smooth the image sequences. Numerical experimental results show that this algorithm can effectively smooth out the unwanted motion and follow the intentional camera movement under certain resolutions.
