Resolution and image quality improvements for small image sensors
View Patent ↗An image having higher resolution and/or better color is generated from multiple images of a subject taken by a handheld imaging device. The motion of the device, even when a user holds the device steady, gives the images perspectives that differ by an amount ranging from a fraction of a pixel to a few pixels. The differences in perspective provide different information about the subject that can be combined to create a better image. In particular, an image shifted by a non-integer number of pixels relative to another image provides information about portions of the subject that are between the pixels of the other image and usable for increasing resolution. Images shifted relative to each other by an integer number of pixels have aligned pixels that may provide different color information for the same portion of the subject when the aligned pixels correspond sensor elements having different color filters.
1. A process for generating a combined image, comprising:
taking a plurality of images of a subject using a handheld device;
analyzing the images to determine relative motion of the images; and
for each pixel in the combined image,
identifying one or more nearby pixel that after correcting for the determined relative motion of the images are less than a threshold distance from a location corresponding to the pixel in the combined image; and
determining a pixel value for the pixel in the combined images from one or more pixel values of the nearby pixels.
2. The process of claim 1 , wherein the combined image has a resolution that is greater than a resolution of the images of the subject.
3. The process of claim 1 , wherein the pixel value for at least one pixel in the combined image has a first color component measured during taking of a first of the plurality of images and a second color component measured during taking of a second of the plurality of images.
4. The process of claim 1 , wherein taking the plurality of images of the subject comprises holding the handheld device steady in a user's hand while taking the plurality of images.
5. The process of claim 4 , wherein the images are taken in series at a rate greater than about 15 per second.
6. The process of claim 1 , wherein analyzing the images comprises for each of the plurality of images, constructing a continuous motion field that maps pixels in the image to locations in a reference image.
7. The process of claim 6 , wherein constructing the continuous motion field for an image comprises:
expanding the image to create a representation of an expanded image having a resolution higher than a resolution of an image sensor in the handheld device;
finding at least one motion vector identifying an area in the expanded image that best matches an area in the reference image; and
constructing the continuous motion field from the at least one motion vector for the expanded image.
8. The process of claim 6 , wherein determining a pixel value for a pixel in the combined images comprises identifying in one of the plurality of images a source pixel that maps to a location of the pixel in the combined image, and using a pixel value associated with the source pixel in the pixel value of the pixel in the combined image.
9. The process of claim 8 , wherein the pixel value is a color value.
10. The process of claim 8 , wherein the pixel value is a directly measured color component.