Image Analysis and Computer Vision Applied to a Baby Monitor and Color Printing

<p dir="ltr">This dissertation mainly has two topics. One is for a motion detection algorithm designed for a baby monitor device. The other is for color printing, which actually contains two techniques: content-color-dependent screening for printing continuous-tone content pages and dual screening dual frequency screening for printing text content pages.</p><p dir="ltr">The motion detection for a baby monitor uses a statistics method, the kernel density estimation, to acquire the motions over the whole image frame. With the additional saliency map extracted from saliency detection neural network providing the foreground and background separations, the motion detection algorithm results in an overall motion value which emphasizes foreground meaningful motions.</p><p dir="ltr">Content-color-dependent screening (CCDS) determines the best screen assignments for either regular or irregular halftones for each image segment, which minimizes the perceived error compared to the continuous-tone digital image. The model first detects smooth areas of the image and applies a spatiochromatic HVS-based model for the superposition of the four halftones to find the best screen assignment for these smooth areas.</p><p dir="ltr">Similar to CCDS, dual frequency dual screening assigns a set of high frequency screens to text areas and a set of low frequency screens to constant tone background areas for reaching the best printing quality on both regions. Low frequency screens generally render smooth constant tone color patches with less texture, while high frequency screens have the ability to demonstrate more details for the text areas, which contain high frequency information.</p>