Ambarella has been granted a patent for an apparatus that processes pixel data from a capture device to generate fused maps for disparity and optical flow. The apparatus also generates regenerated image frames and performs classification of sample image frames, updating parameters based on the accuracy of the classification. GlobalData’s report on Ambarella gives a 360-degree view of the company including its patenting strategy. Buy the report here.
Smarter leaders trust GlobalData
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
According to GlobalData’s company profile on Ambarella, cloud gaming was a key innovation area identified from patents. Ambarella's grant share as of September 2023 was 87%. Grant share is based on the ratio of number of grants to total number of patents.
The patent is granted for an apparatus for video frame processing and classification
A recently granted patent (Publication Number: US11775614B2) describes an apparatus and method for unsupervised multi-scale disparity/optical flow fusion. The apparatus includes an interface to receive pixel data from a capture device and a processor to process the pixel data. The processor is configured to extract features from video frames, generate fused maps for disparity and optical flow based on the extracted features, and generate regenerated image frames through warping. The apparatus also performs classification of sample image frames and updates parameters based on the correctness of the classification.
The apparatus can be used with a stereo camera setup, where the first subset of video frames is captured by the first camera and the second subset is captured by the second camera. The regenerated image frames can be used as a training dataset for disparity calculations. Alternatively, the first subset can represent video frames captured at an earlier time in a sequence, while the second subset represents frames captured at a later time. In this case, the regenerated image frames serve as a training dataset for optical flow information.
The classification of sample image frames is based on second parameters, and the first and second parameters are updated depending on the correctness of the classification. The regenerated image frames are used as a training dataset when the probability of correct classification exceeds a threshold amount.
The apparatus includes a processor that implements a first neural network based on the first parameters and a second neural network based on the second parameters. The first neural network extracts features, generates fused maps, and generates regenerated image frames. The second neural network generates scaled maps for disparity and optical flow and performs the classification of sample image frames.
The generative neural network model and the discriminative neural network model can be configured as an unsupervised generative adversarial network. In a training mode of operation, both models are used to generate regenerated image frames. In a data generation mode of operation, the discriminative neural network model is disabled, and only the generative neural network model is used.
Overall, this patent describes an apparatus and method for unsupervised multi-scale disparity/optical flow fusion, which can be used in various applications such as computer vision, image processing, and machine learning.
To know more about GlobalData’s detailed insights on Ambarella, buy the report here.
Data Insights
From
The gold standard of business intelligence.
Blending expert knowledge with cutting-edge technology, GlobalData’s unrivalled proprietary data will enable you to decode what’s happening in your market. You can make better informed decisions and gain a future-proof advantage over your competitors.
GlobalData, the leading provider of industry intelligence, provided the underlying data, research, and analysis used to produce this article.
GlobalData Patent Analytics tracks bibliographic data, legal events data, point in time patent ownerships, and backward and forward citations from global patenting offices. Textual analysis and official patent classifications are used to group patents into key thematic areas and link them to specific companies across the world’s largest industries.
