APPARATUS AND METHOD FOR COMPRESSING IMAGE

The present invention discloses a device and a method for compressing an image. According to an embodiment of the present invention, the device for compressing an image comprises: an input unit which is inputted with an image; a down-sampling unit which conducts down-sampling of the inputted image;...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: CHOI, HYUNG WON, SUNG, NAK HO, SHIN, JIN WOO
Format: Patent
Sprache:eng ; kor
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The present invention discloses a device and a method for compressing an image. According to an embodiment of the present invention, the device for compressing an image comprises: an input unit which is inputted with an image; a down-sampling unit which conducts down-sampling of the inputted image; a first compressing unit which generates a first compressed image from the down-sampled image by using at least one first autoencoder formed of a convolution neural network; an up-sampling unit which conducts up-sampling of the first compressed image; a subtracting unit which subtracts the up-sampled image from the inputted image; a second compressing unit which generates a second compressed image from the subtracted image by using at least one second autoencoder formed of the convolution neural network; and an adding unit which generates a result image by adding the up-sampled image and the second compressed image. Therefore, the device and the method for compressing an image can compress an image by using a convolution neural network. 영상 압축 장치 및 방법이 개시된다. 본 발명의 일 실시예에 따른 영상 압축 장치는, 영상을 입력 받는 입력부, 상기 입력된 영상을 다운 샘플링(down-sampling)하는 다운 샘플링부, 컨볼루션 신경망으로 구성된 적어도 하나의 제1 오토 인코더(auto encoder)를 이용하여, 상기 다운 샘플링된 영상으로부터 제1 압축 영상을 생성하는 제1 압축부, 상기 제1 압축 영상을 업 샘플링(up-sampling)하는 업 샘플링부, 상기 입력된 영상에서 상기 업 샘플링된 영상을 감산하는 감산부, 컨볼루션 신경망으로 구성된 적어도 하나의 제2 오토 인코더를 이용하여, 상기 감산된 영상으로부터 제2 압축 영상을 생성하는 제2 압축부, 및 상기 업 샘플링된 영상 및 상기 제2 압축 영상을 가산하여 결과 영상을 생성하는 가산부를 포함한다.