Downsampling light field images

摘要

According to various embodiments of the invention, improved downsampling techniques are employed, which can be applied to light field images and which preserve the ability to refocus (and otherwise manipulate) such images. Groups of pixels, rather than individual pixels, are downsampled; such groups of pixels can be defined, for example, as disks of pixels. Such downsampling is accomplished, for example, by aggregating values for pixels having similar relative positions within adjacent disks (or other defined regions or pixel groups) of the image. When applied to light field images, the downsampling techniques of the present invention reduce spatial resolution without sacrificing angular resolution. This ensures that the refocusing capability of the resulting light field image is not reduced and/or adversely impacted.

主权项

1. A method of generating a downsampled image from a first image, comprising:
at a processor, receiving a first image, the first image comprising a plurality of regions, each region comprising a plurality of pixels; at the processor, combining pixel values of pixels within a plurality of adjacent regions of the first image to generate downsampled pixel values; in a storage device, storing the downsampled pixel values; and at the processor, generating a downsampled image using the downsampled pixel values as a region of the downsampled image;wherein combining pixel values from pixels within a plurality of adjacent regions of the first image comprises:
at the processor, identifying a pixel within each of the adjacent regions, each identified pixel having a common relative location with respect to its region; and at the processor, aggregating the identified pixels to generate a pixel value for the downsampled image.