i want to extract silhouette from a a video. but i am not not able to decide which filter i should use. Is there any rule to select the best possible filter for a image. As when i choose a filter for one image it work well but for the second video i have to change the filter size and some time filter too .
Please suggest
In my opinion, filtering in frequency domain is better. In this case you may use a dynamic (law, band or high) pass filter. The term 'dynamic' means: it is capable of changing the selected band of frequencies (allowed frequencies) for each iteration according to some parameters calculated from the input image, like contrast, brightness and so on....it is easy and practical idea.
you can check the attached paper.
Contrast of the image varies for each image.so the results vary. No need to different filters for different images. S
If attempting to use a single filter, the "best" filter is the one that exactly fits your requirements. As Hayde Peregrina-Barreto already stated (and you found out too), filter performance may differ from image to image due to quite a number of parameters.
The best single filter for any application will be the one that delivers the best average results for a given (big) set of test images/videos. It may still show poor performance for some inputs (depends on whether this is considered acceptable -> requirements), but will be the one with the best average results on the test set.
Hope this helps
There is not an unique solution, it depends on a lot of parameters like illumination conditions, capture and so on.
I think that a high pass filter can capture correctly a silhouette from a video sequence, but you can also use a discrete wavelet and capture the details information.
- Badges
- Science topic
- Similar topics
- Engineering
- Computer Engineering
- Computer