Automatic Depth-Map Colorization
نویسندگان
چکیده
We present a system for automatically generating custom, structured image-maps for input depth-maps. Our system thus allows quick fitting of masses of objects with tailor-made image-maps. Given a depth-map of a novel 3D object, our method tiles it with intensities from similar, pre-collected, textured objects. These are seamlessly merged to form the new image-map. This process is performed by optimizing a well defined target likelihood function, via a hard-EM procedure. We present results for varied object classes including human figures, and fish.
منابع مشابه
Automatic Colorization of Grayscale Images Using Generative Adversarial Networks
Automatic colorization of gray scale images poses a unique challenge in Information Retrieval. The goal of this field is to colorize images which have lost some color channels (such as the RGB channels or the AB channels in the LAB color space) while only having the brightness channel available, which is usually the case in a vast array of old photos and portraits. Having the ability to coloriz...
متن کاملAutomatic Colorization with Deep Convolutional Generative Adversarial Networks
We attempt to use DCGANs (deep convolutional generative adversarial nets) to tackle the automatic colorization of black and white photos to combat the tendency for vanilla neural nets to ”average out” the results. We construct a small feed-forward convolutional neural network as a baseline colorization system. We train the baseline model on the CIFAR-10 dataset with a per-pixel Euclidean loss f...
متن کاملImage Colorization with Generative Adversarial Networks
Over the last decade, the process of automatic colorization had been studied thoroughly due to its vast application such as colorization of grayscale images and restoration of aged and/or degraded images. This problem is highly ill-posed due to the extremely large degrees of freedom during the assignment of color information. Many of the recent developments in automatic colorization involved im...
متن کاملBuilding 3D Shapes from Parts
The research described in this thesis has focused on the composition of novel 3D shapes (e.g., depth maps and triangulated meshes) by reusing parts of existing shapes. Specifically, two methods were designed [28, 27]. The first is a novel tool for easy composition of existing 3D models (i.e., triangulated meshes) into novel models [28]. This tool was designed with the intention of providing a m...
متن کاملLearning Representations for Automatic Colorization
We develop a fully automatic image colorization system. Our approach leverages recent advances in deep networks, exploiting both low-level and semantic representations. As many scene elements naturally appear according to multimodal color distributions, we train our model to predict per-pixel color histograms. This intermediate output can be used to automatically generate a color image, or furt...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2006