Blur enhancement of stereoscopic images
View Patent ↗A method and apparatus for selectively blurring parts of an image set according to differences in disparity between successive images. As a result, the one-dimensional ‘ringing’ edge artifacts, intrinsic to autostereoscopic displays, are hidden, thereby improving the perceived image quality of a multi-view autostereoscopic image. The method may involve detecting regions of disparity between adjacent views, and then blurring those regions by weighted local pixel averaging. The relationship between the blur radius and the disparity is preferentially non-linear to maintain sharpness of low disparity regions.
1. A method for enhancing a stereoscopic image, the method comprising:
determining differences in disparity between adjacent image views in an image set, the image views comprising at least two different perspective views of the stereoscopic image; and
blurring at least one portion of at least one of the image views in the image set if one of the differences in disparity exceeds a predetermined value.
2. The method of claim 1 , wherein the blurring at least one portion of at least one of the image views in the image set comprises blurring by local pixel averaging.
3. The method of claim 2 , wherein the blurring by local pixel averaging comprises weighted local pixel averaging.
4. The method of claim 1 , wherein the blurring at least one portion of at least one of the image views in the image set comprises two dimensional blurring.
5. The method of claim 1 , wherein the blurring at least one portion of at least one of the image views in the image set comprises blurring multiple portions of a plurality of the image views in the image set.
6. A method for enhancing a stereoscopic image, the method comprising:
inputting image views in an image set, the image views comprising at least two perspective views of the stereoscopic image;
blurring at least one portion of at least one of the image views in the image set; and
outputting the modified image views of the image set.
7. The method of claim 6 , wherein the blurring at least one portion of at least one of the image views in the image sets comprises: deriving disparity data based on the differences in disparity between adjacent image views in the image set; when a first view of the adjacent image views is the last view in the image set, replacing the pixel values of a second view of the adjacent image views with the average of the local pixel values; and when a first view of the adjacent image views is not the last view in the image set, replacing the pixel values of the last view in the image set and the pixel values of the next to last view in the image set with the average of the local pixel values.
8. The method of claim 7 , wherein the deriving disparity data based on the differences in disparity between adjacent image views in the image set comprises using a block-based motion estimation.
9. The method of claim 7 , wherein the deriving disparity data based on the differences in disparity between adjacent image views in the image set comprises generating a disparity map, the disparity map comprising an array of numbers describing an estimated shift of features between adjacent image views.
10. The method of claim 9 , further comprising smoothing the disparity map by filtering.
11. The method of claim 10 , wherein the smoothing the disparity map by filtering comprises using median filtering.
12. The method of claim 10 , further comprising blurring the adjacent image views as a function of the disparity data.
13. The method of claim 12 , wherein the blurring the adjacent image views comprises averaging pixels over a square area of the adjacent image views.
14. The method of claim 12 , wherein the blurring the adjacent image views comprises averaging pixels over a substantially circular area of the adjacent image views.
15. The method of claim 12 , wherein the blurring the adjacent image views comprises using Gaussian blurring.
16. The method of claim 12 , wherein the blurring the adjacent image views further comprises: blurring only the portions of the image adjacent image views for which the disparity data exceeds a predetermined threshold.
17. A system for enhancing a stereoscopic image, the system including:
a disparity determining module operable to determine differences in disparity between adjacent image views in an image set, the image views comprising at least two different perspective views of the stereoscopic image; and
a blurring module operable to blur at least one portion of at least one of the image views in the image set if one of the differences in disparity exceeds a predetermined value.
18. The system of claim 17 , wherein the blurring module is operable to blur using weighted local pixel averaging.
19. The system of claim 17 , wherein the stereoscopic image is for an autostereoscopic system.
20. A system for enhancing a stereoscopic image, the system comprising:
an input module operable to receive image views in an image set, the image views comprising at least two perspective views of the stereoscopic image;
a disparity deriving module operable to derive disparity data based on the differences in disparity between adjacent image views in the image set;
a replacing pixels module operable to replace the pixel values of a second view of the adjacent image views with the average of the local pixel values when a first view of the adjacent image views is the last view in the image set, and operable to replace the pixel values of the last view in the image set and the pixel values of the next to last view in the image set with the average of the local pixel values when a first view of the adjacent image views is not the last view in the image set; and
an output module operable to output the modified image views of the image set.
21. The system of claim 20 , wherein the disparity deriving module is operable to use a block-based motion estimation.
22. The system of claim 20 , wherein the disparity deriving module is operable to generate a disparity map comprising an array of numbers describing an estimated shift of features between adjacent image views.
23. The system of claim 22 , further comprising a smoothing module operable to smooth the disparity map with median filtering.
24. A system for enhancing a stereoscopic image, the system comprising:
an input module operable to receive an image set comprising image views, the image views comprising at least two perspective views of the stereoscopic image;
a disparity deriving module operable to derive disparity data based on the differences in disparity between adjacent image views in the received image set;
a blurring module operable to blur at least one portion of at least one of the image views in the received image set; and
an output module operable to output the modified image views.