At this stage, many of the original IC chip in the video capture chip, integrated with the DSP kernel, to give users more DIY space, but also makes the camera become more and more intelligent. For example, Rockchip, Novatek, and iCatch chips are integrated with CEVA DSP. PI should image the development of the industry, the depth of the original image algorithm adapting CEVA DSP kernel, the algorithm can according to industry needs, more flexible, accelerate product and ensure the stability of the product.
DSP machine splicing, GPU acceleration, according to the embedded chip platform and the built-in DSP kernel performance, Pisoft algorithm can achieve binocular, multi eye, panoramic spelling, 2.5K/30fps~8K/30fps ranging. At present, the real-time panoramic stitching of 2.5K/30fps has been implemented based on CEVA XM4 128K cache, and more image algorithms are being transplanted to DSP, including steady state, identification, tracking and so on.
Single machine panoramic broadcast, limited transmission bandwidth
machine to complete the image stitching, reducing bandwidth usage
front-end resources limited?
DSP machine splicing, one button sharing, without two transcoding, save APP resources, easy to combine with more intelligent algorithms.