Machine Learning for Underwater Hazy Image

Aditya Patel

ISBN 10: 6203869473 / ISBN 13: 9786203869477

Verlag: LAP LAMBERT Academic Publishing Jun 2021, 2021

Beschreibung

Neuware -Image performance in underwater robots is one of the most challenging problems for autonomous underwater robotics due to light transmission in water. Although image restoration techniques can effectively remove a haze from a damaged image, they require multiple images from the same location making it difficult to use in real time. Considering the positive effects of in-depth learning strategies on other image processing problems such as coloring or finding objects, a deeper learning solution is proposed. The convolutional neural network is trained in image retrieval techniques to capture one image better than other image enhancement techniques. The proposed method is capable of producing high quality image restoration images with a single image as input. The neural network is verified using images from various locations and signals to prove the power of normal action.Books on Demand GmbH, Überseering 33, 22297 Hamburg 72 pp. Englisch.

Bestandsnummer des Verkäufers 9786203869477

Bibliografische Details

• Titel: Machine Learning for Underwater Hazy Image
• Verlag: LAP LAMBERT Academic Publishing Jun 2021
• Erscheinungsjahr: 2021
• Sprache: Englisch
• ISBN-10: 6203869473
• ISBN-13: 9786203869477
• Einband: Taschenbuch
• Zustand: Neu

Über diesen Titel

Inhaltsangabe

Image performance in underwater robots is one of the most challenging problems for autonomous underwater robotics due to light transmission in water. Although image restoration techniques can effectively remove a haze from a damaged image, they require multiple images from the same location making it difficult to use in real time. Considering the positive effects of in-depth learning strategies on other image processing problems such as coloring or finding objects, a deeper learning solution is proposed. The convolutional neural network is trained in image retrieval techniques to capture one image better than other image enhancement techniques. The proposed method is capable of producing high quality image restoration images with a single image as input. The neural network is verified using images from various locations and signals to prove the power of normal action.