Ever since the birth of video games, development costs have grown as hardware and technology have progressively advanced in complexity. As more processing power has been made available games have continued to take advantage of it to produce more detailed simulations and experiences. Is this a new paper of am I stealing around this sentence?
This trend of hardware advancement has led to development costs increasing to match the improved ability of available technology as team sizes increase to produce additional content. Any visual rendering progress can only be effectively utilized if higher quality art content can also be created.
One method that can aid the production of ...view middle of the document...
“Can texture synthesis controlled by genetic programming create textures that share characteristics with a sample texture?”
Aims and Objectives
This project aims to explore the use of evolutionary techniques to generate images using information derived from a sample image.
A system would be created that would make use of genetic programming to evolve populations which individually describe a single image with the aim of containing one or more features that are part of the sample image.
This could be done with a fitness function that, using a range of pattern recognition techniques, would identify a pattern or feature in a sample image and evaluate any generated textures and rate their fitness. The actual techniques or features could be as simple as identifying a dominant colour or more complicated, for example finding specific high or low frequency patterns or textures.
Content generation is a growing issue in the games industry, as currently there is no solution which does not significantly increase development costs. A solution or research into an area, which would help ease this problem, could clearly be valuable. Finding a way to reduce the amount of time spent on recreating repetitive patterned textures and freeing time to be spent specifically on the detailing or style of the image would increasing the visual quality of assets overall.
The use of genetic programming to find a range of solutions instead of an ultimate or single fittest solution is an interesting approach that does not seem to have been explored in great depth. It could be useful to evaluate the effectiveness of this method for further use through other criteria where a fittest solution may vary subjectively, such as artistic merit.
Developing machine intelligence able to create art that is similar to another piece, but unique and stands on its own, is an exciting idea, as the ability to do so is a human trait that has not yet been successful captured.
Methodology and Project Execution
In order to demonstrate and explore the evolution of textures, a system must be built that can facilitate the evolution of images, the procedural generation of images, and the evaluation of fitness of these images. While these three requirements are related in their need to interact with each other, the specific implementation of their function differs enough to consider them independently.
As such, to simplify and help control the scope of this project, these three components of the proposed system will be considered separate and developed as individual applications.
This modular approach will allow sections to be ‘frozen’ once they have met requirements, possibly preventing introduction of bugs or feature creep. This method can also be beneficial in respect to the fitness function as it allows the module to be replaced with another of a different technique easily.
In order to evolve images there must be a notation to...