Recently, the Steering Committee of the Digital Video Broadcasting Project (DVB) officially approved AVS3, a video codec standard originating in China, to become one of the next generation video co... In 2022, the PCI-SIG developer conference is being held in full swing, and the standards committee PCI-SIG, which hides behind the ubiquitous PCIe interface standard, announced that the goal of PCI... Scottsdale, Arizona – May 10, 2022 – MediaTek today unveiled its new Genio platform for AIoT devices and introduced the first chip in the Genio family, the Genio 1200 designed for premium AIoT prod... |Analysis of domestic semiconductor quartz products marketIn the past month, quartz products, a key component in the semiconductor industry, attracted the attention of industry and capital. In Marc...
Embrace the open-source ecosystem of openharmony and jointly build the future of innovative technolo
2022-03-28
As we all know, openharmony has broad prospects in the interconnection of all things, distribution and OS diversity. Mr. Cui Yuntao mentioned in the sharing: "Quanzhi technology is a manufacturer f...
2021-12-27
[Shenzhen, China, December 27, 2021] Shanghai Hisilicon has long had insight into the business characteristics in the field of visual perception, continuously innovated around image processing and ... MediaTek pentonic smart TV platform uses advanced technology to promote the breakthrough and innovation of smart TV marketNovember 22, 2021 - MediaTek released the new 8K flagship smart TV chip pen...
The first silicon-based OLED evaporation machine was successfully introduced into the mass productio
2021-10-27
Recently, the first silicon-based OLED evaporation machine developed by Hefei xinyihua Intelligent Machine Co., Ltd. (hereinafter referred to as "Hefei xinyihua"), an intelligent equipment manufact...
Ruixin micro toybrick launched tb-rk3568x and tb-rv1126d development boards to support massive appli
2021-07-06
Recently, Ruixin micro toybrick officially released the new tb-rk3568x and tb-rv1126d development boards to support massive scenes with multi-dimensional technical advantages.1、 Tb-rk3568x developm... Ruixin micro video bridge chip rk628d supports sharing of three advantages, six application scenarios, interface performance and resolution. Quanzhi technology announced the launch of "D1" processor, which is the world's first mass-produced application processor equipped with flat head Ge xuantie 906 risc-v, providing a new intelligent key March 3, 2021 - MediaTek released the latest 4K smart TV chip mt9638, which integrates an independent AI processor APU, gives full play to AI's high energy efficiency, supports AI technologies such... With the gradual layout of the Chinese mainland's high generation line and the pursuit of consumers' immersion experience and high quality, large size and high resolution are the future trend of la... According to the company's annual development plan and deployment, and through the continuous research and development efforts of the company's developers, five computer software copyright certificate |
Hisilicon Yueying: a new generation of AI ISP technology, setting a new benchmark for image qualityIssuing time:2021-12-27 14:41 [Shenzhen, China, December 27, 2021] Shanghai Hisilicon has long had insight into the business characteristics in the field of visual perception, continuously innovated around image processing and intelligent analysis technology, and is committed to providing leading technical solutions for industry partners. In 2021, Hisilicon is committed to introducing new generation of intelligent processing technology and solutions to the industry. In 2021, Hisilicon launched a new generation of ISP for intelligent terminals of the Internet of things - Yueying, which introduced AI ® Intelligent image processing engine breaks through the limit of traditional ISP image processing and pushes the image quality benchmark of the industry to a new height. Looking back on the development of the visual industry in the past decade, it has gone through the intelligent stage of "seeing" and "seeing clearly" to the current "understanding". The image resolution continues to improve and the image effect continues to be optimized. However, due to the fragmented scene of visual perception, it is still a recognized problem in the industry to output high-quality images in dark light, backlight, motion and other scenes ISP is the first link of image processing for the original data input by the image sensor. It is not only an important guarantee of "being able to see clearly", but also the direct input of subsequent "being able to understand". The image quality output by ISP determines the ceiling of "being able to understand". ISP covers dozens of image signal processing algorithms. In order to adapt to different lenses and sensors, hundreds of parameters need to be optimized and adjusted to ensure the versatility and flexibility of ISP. At the same time, the algorithms of many modules in ISP affect each other, and algorithm cooperation requires a lot of debugging work. This is an extremely complex process and a very challenging task. In the face of increasing scene complexity, it is more and more difficult to significantly improve the image quality based on traditional image processing technologies such as filtering in dark light, wide dynamic and other scenes. Moreover, as the parameter library becomes larger and larger, debugging becomes more and more difficult, the development cycle is gradually lengthened, and the return on investment for image quality improvement is getting lower and lower. In recent years, image enhancement using AI has gradually become a new research hotspot in the industry and made remarkable progress. In particular, the optimization of still photos using AI has been widely used in high-end mobile phone cameras and achieved good results. The characteristics of the visual industry require that AI can be used for real-time tuning of high-resolution and high frame rate video, which has higher requirements for algorithm and computing power. In particular, it is required to efficiently realize the function of AI ISP in the end-to-side computing environment, so as to obtain better results than traditional ISP. Taking advantage of its long-term accumulation and experience in traditional ISP image quality technology and its advantages in end-to-side AI architecture and optimization, Shanghai Hisilicon released a new generation of intelligent image processing engine - Yueying during this Expo ® AI ISP deeply integrates the traditional image quality optimization experience with AI technology, bringing a qualitative leap in the image quality of real-time HD video. Compared with the previous generation of Hisilicon flagship ISP, AI ISP has made significant progress in five aspects. Advantages and features of Hisilicon Yueying image processing engine [super sensitive noise reduction] intelligent noise reduction based on deep learning, with signal-to-noise ratio increased by 4 times Denoising has always been an important function of ISP, especially in the field of video perception. For example, in the streets or parks with insufficient light at night, various forms of noise will appear in the recorded video, seriously affecting the picture definition, making it difficult to "see clearly" and even more difficult to "understand". The traditional NR technology adopts multistage time-domain / spatial filtering, and the filter design integrates a variety of heterogeneous types, and the benefit has been gradually reduced. Hisilicon uses neural network to deeply learn the noise and signal distribution characteristics of massive low illumination scene pictures, and trains a set of intelligent noise reduction algorithm model to make the more shadow ® AI ISP can intelligently distinguish the signal and noise in the image and realize intelligent noise reduction in low illumination scenes. According to the test, Yueying ® Compared with Hisilicon's flagship ISP of the previous generation, the image signal-to-noise ratio can be increased by 4 times in low illumination scenes. Hisilicon Yueying super sensitive noise reduction, signal-to-noise ratio increased by 4 times, fully covering low illumination scenes [multispectral fusion] it adopts special hardware acceleration to realize 0.01LUX low illumination full-color In low light scenario, the sensitivity of RGB sensor decreases sharply, resulting in the rapid decline of color image definition, which can not take into account both image detail and color at the same time, which brings serious challenges to all-weather video perception. In the night traffic management scene, it is often necessary to install flash lights at the gate to supplement the lighting, which is easy to cause light pollution and even affect the driving safety of drivers. By fusing the perceptual information of visible light and infrared light, multispectral fusion technology solves the problem of detail loss caused by insufficient sensitivity of visible light in low illumination environment without increasing light pollution, realizes the best balance between image color and detail, and finally makes the image effect reach 0.01LUX full-color dark light and rich detail texture. Hisilicon multispectral fusion technology fuses common spectra through infrared supplementary light to realize dark full-color [super wide dynamic] adaptive light complex changes, dynamic range increased by 12dB Wide dynamic technology is an important technology to improve image quality in backlight scenes. For example, when driving in and out of the tunnel, due to the huge contrast between internal and external light intensity, how to quickly track the changing light and accurately make adaptive exposure to the target is a very important technology for automatic and assisted driving safety. Based on the intelligent recognition of the target scene, the super wide dynamic technology carries out adaptive multiple exposure to the value area, and integrates the image data of long frame, medium frame and short frame with different exposure time through the intelligent algorithm, so that the image can obtain the perfect integration of dark and bright areas: the bright part is not exposed, the dark part details are not lost, the backlight scene can be seen clearly, and the dark area is full of color and rich in details. According to the test, Yueying ® Compared with Hisilicon's flagship ISP of the previous generation, the dynamic range can be increased by 12dB in the backlight scene, and the scene tracking is faster, which can effectively avoid the instantaneous over darkness or over exposure of the image, and converge to a stable state within 3 frames. Hisilicon's new generation ISP intelligent processing exposure can still present a clear picture when entering and leaving the tunnel [super image stabilization] 6-axis digital anti shake, greatly improving the stability of video imaging Consumer cameras will be widely used in shooting moving scenes. Because the vibration of the camera in motion will affect the final imaging effect and seriously affect the viewing experience. When visual perception devices are installed in automatic or assisted driving vehicles, the image blur caused by such jitter may even affect driving safety. Yueying ® The super image stabilization technology intelligently analyzes the motion attitude based on IMU 6-axis sensing information, and modifies the dynamic video through accurate compensation, which can effectively reduce the impact on video imaging caused by shaking or shaking of the external environment, and greatly improve the video stability in various vibration environments. Hisilicon Yueying 6-axis digital anti shake greatly improves the imaging stability [multi-dimensional perception] multi perception cross-border integration to stimulate the new potential of intelligent vision Vision is the main input source of perceptual information, but the demand for information in production and life is multi-dimensional. Yueying ® It supports the integration of RGB and multi-dimensional perception capabilities such as radar, infrared thermal imaging and 3D structured light, and matches the needs of a variety of products such as infrared temperature measurement, intelligent payment, floor sweeping robot and radar video all-in-one machine. It can be widely used in a variety of business scenarios such as traffic management, financial payment and industrial vision. This multi-dimensional perception ability can also be used by various industry segments to stimulate the innovation of more new categories of intelligent terminals. Hisilicon Yueying supports multi-dimensional perception and activates more intelligent terminal innovation Vision and AI are key technologies in many fields of new infrastructure, and they are also the most important basic capabilities of intelligent terminals. Based on the vision industry, Shanghai Hisilicon has deep experience and industry-leading core competence in the field of end-to-side visual perception and AI computing. It adheres to investment and innovation to provide industry partners with the most advanced technologies and solutions. Haisi Yueying ® AI ISP deeply integrates vision and AI capabilities, which will provide more excellent visual capabilities for Qianxing Baiye intelligent terminals, and create more value for society and people in many aspects, such as urban management, industrial production, smart life, driving safety and so on. |