Saliency Driven Perceptual Image Compression
This work addresses the inadequacy of traditional metrics like MS-SSIM and PSNR for image compression by aligning with human perception, offering potential improvements for applications in media and computer vision.
The paper tackles the problem of lossy image compression by proposing a new end-to-end trainable model that incorporates saliency and a learned perceptual similarity metric, resulting in visually better images and superior performance for computer vision tasks like object detection and segmentation compared to existing methods.
This paper proposes a new end-to-end trainable model for lossy image compression, which includes several novel components. The method incorporates 1) an adequate perceptual similarity metric; 2) saliency in the images; 3) a hierarchical auto-regressive model. This paper demonstrates that the popularly used evaluations metrics such as MS-SSIM and PSNR are inadequate for judging the performance of image compression techniques as they do not align with the human perception of similarity. Alternatively, a new metric is proposed, which is learned on perceptual similarity data specific to image compression. The proposed compression model incorporates the salient regions and optimizes on the proposed perceptual similarity metric. The model not only generates images which are visually better but also gives superior performance for subsequent computer vision tasks such as object detection and segmentation when compared to existing engineered or learned compression techniques.