3D modeling on a smartphone is becoming increasingly attractive due to the performance enhancements in smartphone hardware and its embedded computer vision algorithms. No matter what kind of algorithms we use, an image set input is always required. For research purposes, we may capture an image set of good quality with ease, but in practice, when using modeling applications, ordinary people usually cannot produce an image set with sharp edges and capture all perspectives of the object, which may result in an output having poor visualization.

To harness the smartphone’s computing power and make those modeling apps adaptive to the ordinary people, we designed and implemented a pipeline to produce image sets with better quality. This quality can be reflected in the set’s completen...[ Read more ]

3D modeling on a smartphone is becoming increasingly attractive due to the performance enhancements in smartphone hardware and its embedded computer vision algorithms. No matter what kind of algorithms we use, an image set input is always required. For research purposes, we may capture an image set of good quality with ease, but in practice, when using modeling applications, ordinary people usually cannot produce an image set with sharp edges and capture all perspectives of the object, which may result in an output having poor visualization.

To harness the smartphone’s computing power and make those modeling apps adaptive to the ordinary people, we designed and implemented a pipeline to produce image sets with better quality. This quality can be reflected in the set’s completeness, as well as the sharpness of each image. By utilizing the embedded computer vision algorithms, we can also store valuable information inside each image, which can be further used for supplementary or verification purposes in the reconstruction process.

Furthermore, the implementation of the pipeline is done by utilizing the smartphone CPU’s vector computing capabilities. This significantly reduced its run time and increased overall performance. Meanwhile, by having a cache management system, it is also convenient to extend the pipeline to add more computation stages.