A compressed-domain corner detection method for a DCT-based compressed image
Jongseok Lee, Hyunjae Lee, Dongkyu Lee, Seoung‐Jun Oh · 2017
Researchers have developed compressed-domain computer vision algorithms. Despite the recent progress, the demands for interesting point extraction in compressed-domain still exist. In this paper, we propose a compressed-domain corner detection for a DCT (Discrete Cosine Transform)-based compressed image. It partially decodes the image to obtain DCT data and then split the data into the high precision data. The edge map is then estimated by gradient patterns and coefficients to detect corners. Experimental results show that the proposed method has about 6 pixels of average distance and about 2% miss rate compared to a pixel-domain method with low computational complexity.