For example, we have Checkerboard, Baboon, Cameraman and Lena etc etc some are rich in texture and some are more smooth than other. I am interested in indexing them according to the spatial activity are texture.
In general, one obvious thing would be to use frequency analysis like Fourier transform or Cosine transform on the images, and then make a score on the amount of frequencies somehow. But I guess there are many options for this. And also depends on what it would be used for.
Thank you all.
@Alexandru Isar:
the normalized local standard deviation, nlv, in the attached file is exactly what I was looking. Interesting thresholds to separate homogeneous zones, textured regions and zones of contours. Is the proposed method and thresholds work equally good for noisy as well as noise free images?
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing