BACKGROUND

 Recently, the method of creating 3D models directly from photographs has received much interest. This method has an advantage in producing photo-realistic rendering  because real images are used as input in the rendering process. In this project, we propose to implement a user-friendly application that is capable of extracting 3D models from a set of digitized photographs. The program will use various computer vision techniques such as photogrammetry and stereo algorithms to recover the 3D models. After the 3D models are constructed, photo-realistic images can be used as texture for the models to complete the rendering process. To reduce the complexity of the system, we will limit ourselves to recovering objects with simple geometric shapes such as cubes, planes and pyramids.

PRODUCT DESCRIPTION

In its final state, this program should allow the user to input a set of images profiling a given structure and a set of points defining important edges on the images. The program will use this information to create an accurate 3D model of the structure(s) in the images. The backend portion of our program can be divided into two main engines: the model extraction engine and the texture extraction engine. The model extraction engine uses user-selected points on the various 2D images to create the 3D model. The rendering engine then searches the 2D images to find the appropriate textures for each face of the newly created 3D model. These textures are then applied to the model to create a realistic computer model of the structure(s) in the images.

FEATURES

In addition to the general purpose of PhotoRealistic, we will also implement advance rendering and animation features. Users will be able to generate 3D-animation from rendered 3D models. This feature will require the user to manipulate a scene of models to generate individual key frames.

By rendering realistic 3D models from flat, 2D images, our product provides users from a wide range of backgrounds, the ability to create 3D worlds, buildings, etc. This method of modeling allows users to create realistic models without any previous 3D modeling experience. This product opens up new opportunities to users that have only been limited to the realm of 2D imaging.

Users of our product would include but not be limited to the following:

	City planners, housing developers, and architects would be able to create realistic models to accurately simulate a future building or development before the actual construction begins.
	Real estate brokers, travel agencies, and tourist would be able to see sneak previews of properties and destinations.
	Game developers can generate 3D models and environments with realistic textures to further enhance the realism necessary in today's gaming industry.
	Online shoppers would be able to participate with enhanced customer interaction with products like automobiles, furniture, and appliances as a result of PhotoRealistic 3D models.
	3D artists can render PhotoRealistic models and objects
	Interior designers can provide their customers with multiple previews of new and different interior colors, furniture, carpets, etc by creating 3D interiors for simulated walk-throughs.

MAJOR HURDLES
There are several hurdles that are foreseen in the process of implementing this program. The first hurdle involves the ability to accurately create the 3D model using the user-selected points in the 2D image. The algorithm used to extract the 3D model may be slightly inaccurate and need tuning. Also, the computational time for the algorithm may be rather long. Another hurdle involves the ability to develop a graphical user interface that is simple to use and understand. A final hurdle might be the ability for the program to take 2-D images in multiple formats, i.e. *.jpg, *.bmp, *.tiff, *.gif, as a basis for the conversion.

PLATFORM & TOOLS
PhotoRealistic will be developed on Windows NT for the Win32 environment. The tools involved will include MS Visual C++, MFC, and OpenGL. Presentation tools will include Adobe Premier and possibly MS PowerPoint.