Video stream compression, using lossy algorithms, is performed to reduce the bandwidth required for transmission. To improve the video quality, either for human view or for automatic video analysis, videos are post-processed to eliminate the introduced compression artifacts. Generative Adversarial Network have been shown to obtain extremely high quality results in image enhancement tasks; however, to obtain top quality results high capacity large generators are usually employed, resulting in high computational costs and processing time. In this paper we present an architecture that can be used to reduce the cost of generators, paving a way towards real-time frame enhancement with GANs. With the proposed approach, enhanced images appear natural and pleasant to the eye. Locally high frequency patterns often differ from the raw uncompressed images. A possible application is to improve video conferencing, or live streaming. In these cases there is no original uncompressed video stream available. Therefore, we report results using popular no-reference metrics showing high naturalness and quality even for efficient networks.
Towards Real-Time Image Enhancement GANs / Galteri L.; Seidenari L.; Bertini M.; Del Bimbo A.. - ELETTRONICO. - 11678:(2019), pp. 183-195. (Intervento presentato al convegno 18th International Conference on Computer Analysis of Images and Patterns, CAIP 2019 tenutosi a ita nel 2019) [10.1007/978-3-030-29888-3_15].
Towards Real-Time Image Enhancement GANs
Galteri L.;Seidenari L.;Bertini M.;Del Bimbo A.
2019
Abstract
Video stream compression, using lossy algorithms, is performed to reduce the bandwidth required for transmission. To improve the video quality, either for human view or for automatic video analysis, videos are post-processed to eliminate the introduced compression artifacts. Generative Adversarial Network have been shown to obtain extremely high quality results in image enhancement tasks; however, to obtain top quality results high capacity large generators are usually employed, resulting in high computational costs and processing time. In this paper we present an architecture that can be used to reduce the cost of generators, paving a way towards real-time frame enhancement with GANs. With the proposed approach, enhanced images appear natural and pleasant to the eye. Locally high frequency patterns often differ from the raw uncompressed images. A possible application is to improve video conferencing, or live streaming. In these cases there is no original uncompressed video stream available. Therefore, we report results using popular no-reference metrics showing high naturalness and quality even for efficient networks.I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.