Color Object Retrieval Using Local Features Based on Opponent-Process Theory

Finding the Best Lighting Mode for Daguerreotype, Ambrotype, and Tintype Photographs and Their Deterioration on the Cruse Scanner Based on Selected Methods

“…Our idea was to relax from the connected visual descriptor or spanner (MST, RNG) by connecting the dots with the “Hungarian” edges [ 25 , 26 ]. For adding dots, we experimented with searching the gray level Harris corners [ 18 ] and concurrent color channels coSIFT [ 27 ]. The latter performed better on Cruse scans and already the density of Harris or coSIFT corners indicated “interesting parts” in scanned wooden decors.…”

“…This detector suffered from a low count of Harris corners for certain hard-to-scan images. Our goal in coSIFT [ 27 ] research was to improve the gray level method [ 21 ] by a more color-sensitive corner detector, which we describe in detail below. These data will be used to compute the coSITF heatmaps, which will be confronted with the measured heatmaps from the evaluation of eye-tracking sessions.…”

“…Although the color is perceived as an irreplaceable element describing the world around us, the techniques for extracting the local features are mostly based on the description of the shape, while the color information is entirely ignored. The pipeline of the proposed Co-SIFT [ 27 ] algorithm is illustrated in Figure 4 . The key idea of our solution is to incorporate color information from the image into the SIFT method by replacing grayscale information so that key points are detected on two separate chromatic channels separately (red–green ( ) and a yellow–blue ( )) and the achromatic brightness channel.…”