End-to-End Learned ROI Image Compression

Hiroaki Akutsu, Takahiro Naruko; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2019, pp. 0-0

In this paper, we present the effectiveness of image compression based on a convolutional auto encoder (CAE) with region of interest (ROI) for quality control. We use road images used to check damaged parts in the road. Our evaluation reveals that BPG does not provide adequate quality for the road damaged parts at a low bit rate (1.0 bpp or less). We propose a method that adapts image quality for prioritized parts and non-prioritized parts for CAE-based compression. The proposed method uses annotation information for the distortion weights of the MS-SSIM-based loss function. Experimental results show that the proposed method implemented for CAE-based compression from F. Mentzer et al. learns the characteristics of the road damaged parts by end-to-end training with the weighted loss function and reduces bpp by 31% compared to the original method while meeting quality requirements that an average weighted MS-SSIM for the road damaged parts be larger than 0.97 and an average weighted MS-SSIM for the other parts be larger than 0.95.