Generating a disparity map having reduced over-smoothing
This disclosure describes systems and techniques for generating a disparity map having reduced over-smoothing. To achieve the reduction in over-smoothing that might otherwise occur, a third image is captured in addition to stereo reference and search images. Information from the third image, which may be of the same or different type as the stereo images (e.g., RGB, grey-scale, infra-red (IR)) and which may have the same or different resolution as the stereo images, can help better define the object edges.
1. A method of generating a disparity map comprising:
acquiring first and second stereo images, wherein a first one of the stereo images serves as a reference image and a second one of the stereo images serves as a search image;
acquiring a third image from a vantage point different from the stereo images;
determining a first matching cost between respective pixels in the reference and search images;
determining a smoothness term for pixels in the third image;
determining disparity values between respective pixels in the reference and search images,
wherein the disparity values minimize a sum of the first matching cost and the smoothness term, and
wherein the disparity values collectively define a disparity map for each point in the reference image; and
mapping the smoothness term onto the reference image.
2. The method of claim 1 wherein the stereo images and the third image are of the same type, and wherein the type is infra-red, color, or grey-scale.
3. The method claim 1 wherein the third image has a higher resolution than the stereo images.
4. The method claim 1 wherein determining a first matching cost includes calculating a difference in intensities between the respective pixels in the reference and search images.
5. The method of claim 1 wherein determining a smoothness term includes comparing intensity or color values of adjacent pixels in the third image.
6. The method of claim 1 further including: determining a second matching cost between respective pixels in the reference image and the third image, wherein the disparity values minimize a sum of the first matching cost, the second matching cost and the smoothness term.
7. The method of claim 6 wherein determining a second matching cost includes calculating a difference in intensities between the respective pixels in the reference image and the third image.
8. The method of claim 6 wherein the stereo images and the third image are of the same type, and wherein the type is infra-red, color, or grey-scale.
9. The method of claim 6 wherein the third image has a higher resolution than the stereo images.
10. The method of claim 1 wherein a type of the stereo images and a type of the third image are different, wherein the types are infra-red, color, or grey-scale, and wherein the method further comprises: mapping a color space of the third image onto a color space of the reference image.
11. The method of claim 10 wherein the third image has a higher resolution than the stereo images.
12. The method claim 1 wherein the stereo images and the third image are acquired, respectively, using different image sensors, and wherein different depths or ranges of depth represented by the disparity map are displayed on a display screen using different visual indicators.